1 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
3 * collect2.c (main): Parse the output file early so we can make nicer
4 temp names. Respond to “-save-temps” in the GCC OPTIONS.
5 (maybe_unlink): Don’t print “[Leaving…”] for files we never created
8 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
10 * collect2.c (main): Combine flags from both the command line and
11 COLLECT_GCC_OPTIONS to determine the set in force
13 2018-12-21 Jan Hubicka <hubicka@ucw.cz>
15 * ipa-utils.c (ipa_merge_profiles): Recompute summaries.
17 2018-12-21 Jakub Jelinek <jakub@redhat.com>
21 * omp-expand.c (extract_omp_for_update_vars): Regimplify the condition
23 (expand_omp_for_generic): Don't clobber t temporary for ordered loops.
25 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
28 * ipa-polymorphic-call.c
29 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Handle
30 arguments of thunks correctly.
31 (ipa_polymorphic_call_context::get_dynamic_context): Be ready for
33 * lto-cgraph.c (lto_output_node): Always stream thunk info.
35 2018-12-21 Andreas Krebbel <krebbel@linux.ibm.com>
37 * config/s390/vector.md ("floatv2div2df2", "floatunsv2div2df2")
38 ("fix_truncv2dfv2di2", "fixuns_truncv2dfv2di2"): New pattern
41 2018-12-21 Eric Botcazou <ebotcazou@adacore.com>
43 PR rtl-optimization/87727
44 * combine.c (cant_combine_insn_p): On a LEAF_REGISTERS target, combine
45 again moves from leaf hard registers.
47 * final.c (final_scan_insn_1) <NOTE_INSN_INLINE_ENTRY>: Minor tweak.
49 2018-12-21 Jakub Jelinek <jakub@redhat.com>
52 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask,
53 *avx512pf_gatherpf<mode>df_mask, *avx512pf_scatterpf<mode>sf_mask,
54 *avx512pf_scatterpf<mode>df_mask): Use %X5 instead of %5 for
56 (gatherq_mode): Remove mode iterator.
57 (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2): Use X instead
59 (*avx512f_gatherdi<mode>): Use X instead of <gatherq_mode>.
60 (*avx512f_gatherdi<mode>_2, *avx512f_scattersi<mode>,
61 *avx512f_scatterdi<mode>): Use %X5 for -masm=intel.
63 PR rtl-optimization/88563
64 * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Swap innermode
65 and mode arguments to convert_modes. Likewise swap mode and word_mode
66 arguments. Handle both arguments with VOIDmode before convert_modes
67 of one of them. Formatting fixes.
69 2018-12-21 Uros Bizjak <ubizjak@gmail.com>
72 * internal-fn.def (COSH): New.
75 * optabs.def (cosh_optab): New.
78 * config/i386/i386-protos.h (ix86_emit_i387_sinh): New prototype.
79 (ix86_emit_i387_cosh): Ditto.
80 (ix86_emit_i387_tanh): Ditto.
81 * config/i386/i386.c (ix86_emit_i387_sinh): New function.
82 (ix86_emit_i387_cosh): Ditto.
83 (ix86_emit_i387_tanh): Ditto.
84 * config/i386/i386.md (sinhxf2): New expander.
91 2018-12-21 Jakub Jelinek <jakub@redhat.com>
94 * config/i386/i386.c (ix86_expand_int_sse_cmp): Optimize
95 x > y ? 0 : -1 into min (x, y) == x ? -1 : 0.
97 2018-12-21 Jan Beulich <jbeulich@suse.com>
99 * config/i386/sse.md (vaesdec_<mode>, vaesdeclast_<mode>): Allow
102 2018-12-21 Alexandre Oliva <oliva@adacore.com>
104 * config.gcc (tmake_file): Add name to tmake_file for
105 each @name in --with-multilib-list on arm-*-* targets.
106 * doc/install.texi (with-multilib-list): Document it.
108 2018-12-21 Jakub Jelinek <jakub@redhat.com>
110 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Use
111 DR_TARGET_ALIGNMENT on dr_info rather than dr. Spelling fixes.
113 2018-12-20 Joern Rennecke <joern.rennecke@riscy-ip.com>
115 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Don't do
116 versioning for data accesses with misaligned step.
118 2018-12-20 H.J. Lu <hongjiu.lu@intel.com>
121 * doc/invoke.texi: Document -Wno-address-of-packed-member.
123 2018-12-20 Vladimir Makarov <vmakarov@redhat.com>
126 * ira-color.c (fast_allocation): Choose the best cost hard reg.
128 2018-12-20 Richard Sandiford <richard.sandiford@arm.com>
130 * config/aarch64/iterators.md (SVE_INT_UNARY, fp_int_op): Add abs.
131 (SVE_FP_UNARY): Sort.
133 2018-12-20 Richard Sandiford <richard.sandiford@arm.com>
135 * config/aarch64/aarch64-sve.md (*cond_<optab><mode>_4): Use
136 sve_fmla_op rather than sve_fmad_op for the movprfx alternative.
138 2018-12-20 Martin Jambor <mjambor@suse.cz>
141 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Assert that
144 2018-12-20 Richard Biener <rguenther@suse.de>
146 PR tree-optimization/84362
147 * tree-ssa-loop-im.c: Include alias.h, builtins.h and tree-dfa.h.
148 (struct im_mem_ref): add ref_canonical flag.
149 (struct mem_ref_hasher): Use ao_ref as compare_type.
150 (mem_ref_hasher::equal): Adjust and add variant comparing ao_ref
152 (mem_ref_alloc): Take ao_ref parameter, initialize ref_canonical
154 (gather_mem_refs_stmt): Set up ao_ref early and do the lookup
155 using it. If we have non-equal refs canonicalize the one
156 in the hashtable used for insertion.
157 (tree_ssa_lim_initialize): Adjust.
159 2018-12-20 Jakub Jelinek <jakub@redhat.com>
162 * config/i386/i386.c (ix86_expand_sse_movcc): For maskcmp, try to
163 emit vpmovm2? instruction perhaps after knot?. Reorganize code
164 so that it doesn't have to test !maskcmp in almost every conditional.
166 2018-12-20 Alan Modra <amodra@gmail.com>
168 * config/rs6000/sysv4.h (GNU_USER_DYNAMIC_LINKER): Define.
170 2018-12-19 Andreas Tobler <andreast@gcc.gnu.org>
172 * config.gcc: Enable TARGET_DEFAULT_ASYNC_UNWIND_TABLES on
175 2018-12-19 Uros Bizjak <ubizjak@gmail.com>
177 * config/i386/i386.md (SWI1248_AVX512BWDQ_64): Rename from
178 SWI1248_AVX512BWDQ2_64. Unconditionally enable HImode.
179 (*cmp<mode>_ccz_1): Emit kortest instead of ktest insn.
180 Use SWI1248_AVX512BWDQ_64 mode iterator and enable only for
183 2018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org>
185 * config/arm/elf.h: Update comment about condition that need to
186 match with libgcc/config/arm/lib1funcs.S to also include
187 libgcc/config/arm/t-arm.
188 * doc/sourcebuild.texi (output-exists, output-exists-not): Rename
189 subsubsection these directives are in to "Check for output files".
190 Move scan-symbol to that section and add to it new scan-symbol-not
193 2018-12-19 Tom de Vries <tdevries@suse.de>
195 * config/nvptx/nvptx.c (PTX_CTA_SIZE): Define.
197 2018-12-19 Tom de Vries <tdevries@suse.de>
199 * omp-offload.c (oacc_fn_attrib_level): Remove static.
200 * omp-offload.h (oacc_fn_attrib_level): Declare.
202 2018-12-19 Tom de Vries <tdevries@suse.de>
204 * omp-offload.c (oacc_get_default_dim): New function.
205 * omp-offload.h (oacc_get_default_dim): Declare.
207 2018-12-19 David Malcolm <dmalcolm@redhat.com>
210 * gcc-rich-location.c
211 (maybe_range_label_for_tree_type_mismatch::get_text): Move here from
213 (binary_op_rich_location::binary_op_rich_location): New ctor.
214 (binary_op_rich_location::use_operator_loc_p): New function.
215 * gcc-rich-location.h
216 (class maybe_range_label_for_tree_type_mismatch)): Move here from
218 (struct op_location_t): New forward decl.
219 (class binary_op_rich_location): New class.
220 * tree.h (struct op_location_t): New struct.
222 2018-12-19 David Malcolm <dmalcolm@redhat.com>
226 * convert.c: Include "selftest.h".
227 (preserve_any_location_wrapper): New function.
228 (convert_to_pointer_maybe_fold): Update to handle location
230 (convert_to_real_maybe_fold): Likewise.
231 (convert_to_integer_1): Strip expr when using TREE_OVERFLOW.
232 Handle location wrappers when checking for INTEGER_CST.
233 (convert_to_integer_maybe_fold): Update to handle location
235 (convert_to_complex_maybe_fold): Likewise.
236 (selftest::test_convert_to_integer_maybe_fold): New functions.
237 (selftest::convert_c_tests): New function.
238 * convert.h (preserve_any_location_wrapper): New decl.
239 * fold-const.c (size_binop_loc): Strip location wrappers when
241 (operand_equal_p): Strip any location wrappers.
242 (integer_valued_real_p): Strip any location wrapper.
243 * selftest-run-tests.c (selftest::run_tests): Call
244 selftest::convert_c_tests.
245 * selftest.h (selftest::convert_c_tests): New decl.
246 * tree.c (build_complex): Assert that REAL and IMAG are constants.
247 (integer_zerop): Look through location wrappers.
248 (integer_onep): Likewise.
249 (integer_each_onep): Likewise.
250 (integer_all_onesp): Likewise.
251 (integer_minus_onep): Likewise.
252 (integer_pow2p): Likewise.
253 (integer_nonzerop): Likewise.
254 (integer_truep): Likewise.
255 (fixed_zerop): Likewise.
256 (real_zerop): Likewise.
257 (real_onep): Likewise.
258 (real_minus_onep): Likewise.
259 (tree_int_cst_equal): Likewise.
260 (simple_cst_equal): Treat location wrappers with non-equal source
261 locations as being unequal.
262 (uniform_integer_cst_p): Look through location wrappers.
263 (maybe_wrap_with_location): Don't create wrappers if any
264 auto_suppress_location_wrappers are active.
265 (suppress_location_wrappers): New variable.
266 (selftest::test_predicates): New test.
267 (selftest::tree_c_tests): Call it.
268 * tree.h (CONSTANT_CLASS_OR_WRAPPER_P): New macro.
269 (suppress_location_wrappers): New decl.
270 (class auto_suppress_location_wrappers): New class.
272 2018-12-19 Paul A. Clarke <pc@us.ibm.com>
274 * config/rs6000/tmmintrin.h (_mm_hadds_epi16): Vector lanes swapped.
275 (_mm_hsub_epi32): Likewise.
276 (_mm_shuffle_epi8): Fix reversed interpretation of parameters.
277 (_mm_shuffle_pi8): Likewise.
278 (_mm_addubs_pi16): Likewise.
280 2018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org>
282 * config/arm/arm.h (TARGET_HARD_FLOAT): Restrict to TARGET_32BIT
284 * config/arm/arm.c (output_return_instruction): Only check
285 TARGET_HARD_FLOAT to decide whether FP instructions are available.
287 2018-12-19 Tom de Vries <tom@codesourcery.com>
289 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
290 dump files): Add offload-rtl.
292 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
295 * config/rs6000/vsx.md (*vsx_extract_<P:mode>_<VSX_D:mode>_load):
296 Require TARGET_POWERPC64.
298 2018-12-19 Richard Biener <rguenther@suse.de>
300 PR tree-optimization/88533
302 2018-04-30 Richard Biener <rguenther@suse.de>
304 PR tree-optimization/28364
305 PR tree-optimization/85275
306 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
307 copying first exit test.
309 * tree-ssa-loop-ch.c: Include tree-phinodes.h and
311 (should_duplicate_loop_header_p): Track whether stmt compute
312 loop invariants or values based on IVs. Apart from the
313 original loop header only duplicate blocks with exit tests
314 that are based on IVs or invariants.
316 2018-12-19 Tom de Vries <tdevries@suse.de>
318 * config/nvptx/nvptx.c (nvptx_gen_shared_bcast, shared_prop_gen)
319 (nvptx_goacc_expand_accel_var): Use MAX and ROUND_UP.
320 (nvptx_assemble_value, nvptx_output_skip): Use MIN.
321 (nvptx_shared_propagate, nvptx_single, nvptx_expand_shared_addr): Use
324 2018-12-19 Tom de Vries <tdevries@suse.de>
326 * config/nvptx/nvptx.c (nvptx_gen_wcast): Rename as
327 nvptx_gen_warp_bcast.
328 (nvptx_gen_wcast): Rename to nvptx_gen_shared_bcast, add bool
329 vector argument, and update call to nvptx_gen_shared_bcast.
330 (propagator_fn): Add bool argument.
331 (nvptx_propagate): New bool argument, pass bool argument to fn.
332 (vprop_gen): Rename to warp_prop_gen, update call to
333 nvptx_gen_warp_bcast.
334 (nvptx_vpropagate): Rename to nvptx_warp_propagate, update call to
336 (wprop_gen): Rename to shared_prop_gen, update call to
337 nvptx_gen_shared_bcast.
338 (nvptx_wpropagate): Rename to nvptx_shared_propagate, update call
340 (nvptx_wsync): Rename to nvptx_cta_sync.
341 (nvptx_single): Update calls to nvptx_gen_warp_bcast,
342 nvptx_gen_shared_bcast and nvptx_cta_sync.
343 (nvptx_process_pars): Likewise.
344 (write_worker_buffer): Rename as write_shared_buffer.
345 (nvptx_file_end): Update calls to write_shared_buffer.
346 (nvptx_expand_worker_addr): Rename as nvptx_expand_shared_addr.
347 (nvptx_expand_builtin): Update call to nvptx_expand_shared_addr.
348 (nvptx_get_worker_red_addr): Rename as nvptx_get_shared_red_addr.
349 (nvptx_goacc_reduction_setup): Update call to
350 nvptx_get_shared_red_addr.
351 (nvptx_goacc_reduction_fini): Likewise.
352 (nvptx_goacc_reduction_teardown): Likewise.
354 2018-12-19 Tom de Vries <tdevries@suse.de>
356 * config/nvptx/nvptx.c (worker_bcast_size): Rename as
358 (worker_bcast_align): Rename as oacc_bcast_align.
359 (worker_bcast_sym): Rename as oacc_bcast_sym.
360 (nvptx_option_override): Update usage of oacc_bcast_*.
361 (struct wcast_data_t): Rename as broadcast_data_t.
362 (nvptx_gen_wcast): Update type of data argument and usage of
364 (wprop_gen): Update type of data_ and usage of oacc_bcast_align.
365 (nvptx_wpropagate): Update type of data and usage of
366 oacc_bcast_{sym,size}.
367 (nvptx_single): Update type of data and usage of oacc_bcast_size.
368 (nvptx_file_end): Update usage of oacc_bcast_{sym,align,size}.
370 2018-12-19 Tom de Vries <tdevries@suse.de>
372 * config/nvptx/nvptx.md (nvptx_barsync): Add and handle operand.
373 * config/nvptx/nvptx.c (nvptx_wsync): Update call to gen_nvptx_barsync.
375 2018-12-19 Tom de Vries <tdevries@suse.de>
377 * config/nvptx/nvptx.c (nvptx_single): Always pass false to
379 (nvptx_process_pars): Likewise.
381 2018-12-19 Tom de Vries <tdevries@suse.de>
383 * config/nvptx/nvptx.c (nvptx_previous_fndecl): Declare.
384 (nvptx_set_current_function): New function.
385 (TARGET_SET_CURRENT_FUNCTION): Define.
387 2018-12-19 Shaokun Zhang <zhangshaokun@hisilicon.com>
389 * config/aarch64/aarch64-cores.def (tsv110): Fix architecture. This
390 part is really Armv8.2 with some permitted Armv8.4 extensions.
392 2018-12-19 Jakub Jelinek <jakub@redhat.com>
395 * config/i386/vpclmulqdqintrin.h (_mm256_clmulepi64_epi128): Enable
396 for -mavx -mvpclmulqdq rather than just for -mavx512vl -mvpclmulqdq.
398 2018-12-19 Alexandre Oliva <aoliva@redhat.com>
403 (vr_values::vrp_evaluate_conditional_warnv_with_ops): Extend
404 simplification of overflow tests to cover cases in which we
405 can determine the result of the comparison.
407 2018-12-19 Bin Cheng <bin.cheng@linux.alibaba.com>
409 * auto-profile.c (afdo_indirect_call): Skip generating histogram
410 value if we can't find cgraph_node for then indirected callee. Save
411 profile_id of the cgraph_node in histogram value's first counter.
412 * value-prof.c (gimple_value_profile_transformations): Don't skip
413 for flag_auto_profile.
415 2018-12-18 Vladimir Makarov <vmakarov@redhat.com>
417 PR rtl-optimization/87759
418 * lra-assigns.c (lra_split_hard_reg_for): Recalculate
421 2018-12-18 Jakub Jelinek <jakub@redhat.com>
424 * config/i386/i386-builtin-types.def
425 (VOID_FTYPE_PDOUBLE_QI_V8SI_V4DF_INT,
426 VOID_FTYPE_PFLOAT_QI_V4DI_V8SF_INT,
427 VOID_FTYPE_PLONGLONG_QI_V8SI_V4DI_INT,
428 VOID_FTYPE_PINT_QI_V4DI_V8SI_INT,
429 VOID_FTYPE_PDOUBLE_QI_V4SI_V2DF_INT,
430 VOID_FTYPE_PFLOAT_QI_V2DI_V4SF_INT,
431 VOID_FTYPE_PLONGLONG_QI_V4SI_V2DI_INT,
432 VOID_FTYPE_PINT_QI_V2DI_V4SI_INT): New builtin types.
433 * config/i386/i386.c (enum ix86_builtins): Add
434 IX86_BUILTIN_SCATTERALTSIV4DF, IX86_BUILTIN_SCATTERALTDIV8SF,
435 IX86_BUILTIN_SCATTERALTSIV4DI, IX86_BUILTIN_SCATTERALTDIV8SI,
436 IX86_BUILTIN_SCATTERALTSIV2DF, IX86_BUILTIN_SCATTERALTDIV4SF,
437 IX86_BUILTIN_SCATTERALTSIV2DI and IX86_BUILTIN_SCATTERALTDIV4SI.
438 (ix86_init_mmx_sse_builtins): Fix up names of IX86_BUILTIN_GATHERALT*,
439 IX86_BUILTIN_GATHER3ALT* and IX86_BUILTIN_SCATTERALT* builtins to
440 match the IX86_BUILTIN codes. Build IX86_BUILTIN_SCATTERALTSIV4DF,
441 IX86_BUILTIN_SCATTERALTDIV8SF, IX86_BUILTIN_SCATTERALTSIV4DI,
442 IX86_BUILTIN_SCATTERALTDIV8SI, IX86_BUILTIN_SCATTERALTSIV2DF,
443 IX86_BUILTIN_SCATTERALTDIV4SF, IX86_BUILTIN_SCATTERALTSIV2DI and
444 IX86_BUILTIN_SCATTERALTDIV4SI decls.
445 (ix86_vectorize_builtin_scatter): Expand those new builtins.
447 2018-12-18 Bill Schmidt <wschmidt@linux.ibm.com>
449 * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions):
450 Describe when a typedef name can be used as the type specifier for
451 a vector type, and when it cannot.
453 2018-12-18 Jozef Lawrynowicz <jozef.l@mittosystems.com>
455 * config/msp430/msp430.h: Define TARGET_VTABLE_ENTRY_ALIGN.
457 2018-12-18 Jakub Jelinek <jakub@redhat.com>
461 * optabs.def (vec_pack_sbool_trunc_optab, vec_unpacks_sbool_hi_optab,
462 vec_unpacks_sbool_lo_optab): New optabs.
463 * optabs.c (expand_widen_pattern_expr): Use vec_unpacks_sbool_*_optab
464 and pass additional argument if both input and target have the same
465 scalar mode of VECTOR_BOOLEAN_TYPE_P vectors.
466 * expr.c (expand_expr_real_2) <case VEC_PACK_TRUNC_EXPR>: Handle
467 VECTOR_BOOLEAN_TYPE_P pack where result has the same scalar mode
468 as the operands using vec_pack_sbool_trunc_optab.
469 * tree-vect-stmts.c (supportable_widening_operation): Use
470 vec_unpacks_sbool_{lo,hi}_optab for VECTOR_BOOLEAN_TYPE_P conversions
471 where both wider_vectype and vectype have the same scalar mode.
472 (supportable_narrowing_operation): Similarly use
473 vec_pack_sbool_trunc_optab if narrow_vectype and vectype have the same
475 * config/i386/i386.c (ix86_get_builtin)
476 <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Check for VECTOR_MODE_P
477 rather than non-VOIDmode.
478 * config/i386/sse.md (vec_pack_trunc_qi, vec_pack_trunc_<mode>):
479 Remove useless ()s around "register_operand", formatting fixes.
480 (vec_pack_sbool_trunc_qi, vec_unpacks_sbool_lo_qi,
481 vec_unpacks_sbool_hi_qi): New expanders.
482 * doc/md.texi (vec_pack_sbool_trunc_M, vec_unpacks_sbool_hi_M,
483 vec_unpacks_sbool_lo_M): Document.
485 2018-12-18 Jozef Lawrynowicz <jozef.l@mittosystems.com>
487 * combine.c (update_rsp_from_reg_equal): Only look for the nonzero bits
488 of src in nonzero_bits_mode if the mode of src is MODE_INT and
490 (reg_nonzero_bits_for_combine): Add clarification to comment.
492 2018-12-18 Wei Xiao <wei3.xiao@intel.com>
494 * config/i386/driver-i386.c (host_detect_local_cpu): Detect cascadelake.
495 * config/i386/i386.c (fold_builtin_cpu): Handle cascadelake.
496 * doc/extend.texi: Add cascadelake.
497 2018-12-17 Peter Bergner <bergner@linux.ibm.com>
500 * config/rs6000/vsx.md (nW): New mode iterator.
501 (vsx_mov<mode>_64bit): Use it. Remove redundant GPR 0/-1 alternative.
502 Update length attribute for (<??r>, <nW>) alternative.
503 (vsx_mov<mode>_32bit): Likewise.
505 2018-12-17 Tom de Vries <tdevries@suse.de>
507 * config/nvptx/nvptx.c (PTX_VECTOR_LENGTH, PTX_WORKER_LENGTH,
508 PTX_DEFAULT_RUNTIME_DIM): Move to the top of the file.
510 2018-12-17 Tom de Vries <tdevries@suse.de>
512 * config/nvptx/nvptx.c (PTX_WARP_SIZE): Define.
513 (nvptx_simt_vf): Return PTX_WARP_SIZE instead of PTX_VECTOR_LENGTH.
515 2018-12-17 Tom de Vries <tdevries@suse.de>
517 * config/nvptx/nvptx.c (nvptx_single): Fix whitespace.
518 (nvptx_neuter_pars): Likewise.
520 2018-12-17 Tom de Vries <tdevries@suse.de>
522 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Work around Fortran
523 bug PR72741 by overriding dims parameter for routines.
525 2018-12-17 Tom de Vries <tdevries@suse.de>
527 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Rewrite using
530 2018-12-17 Steve Ellcey <sellcey@cavium.com>
532 * config/aarch64/aarch64-protos.h (aarch64_use_simple_return_insn_p):
534 (aarch64_epilogue_uses): Ditto.
535 * config/aarch64/aarch64.c (aarch64_attribute_table): New array.
536 (aarch64_simd_decl_p): New function.
537 (aarch64_reg_save_mode): New function.
538 (aarch64_function_ok_for_sibcall): Check for simd calls.
539 (aarch64_layout_frame): Check for simd function.
540 (aarch64_gen_storewb_pair): Handle E_TFmode.
541 (aarch64_push_regs): Use aarch64_reg_save_mode to get mode.
542 (aarch64_gen_loadwb_pair): Handle E_TFmode.
543 (aarch64_pop_regs): Use aarch64_reg_save_mode to get mode.
544 (aarch64_gen_store_pair): Handle E_TFmode.
545 (aarch64_gen_load_pair): Ditto.
546 (aarch64_save_callee_saves): Handle different mode sizes.
547 (aarch64_restore_callee_saves): Ditto.
548 (aarch64_components_for_bb): Check for simd function.
549 (aarch64_epilogue_uses): New function.
550 (aarch64_process_components): Check for simd function.
551 (aarch64_expand_prologue): Ditto.
552 (aarch64_expand_epilogue): Ditto.
553 (aarch64_expand_call): Ditto.
554 (aarch64_use_simple_return_insn_p): New function.
555 (TARGET_ATTRIBUTE_TABLE): New define.
556 * config/aarch64/aarch64.h (EPILOGUE_USES): Redefine.
557 (FP_SIMD_SAVED_REGNUM_P): New macro.
558 * config/aarch64/aarch64.md (simple_return): New define_expand.
559 (load_pair_dw_tftf): New instruction.
560 (store_pair_dw_tftf): Ditto.
561 (loadwb_pair<TX:mode>_<P:mode>): Ditto.
562 (storewb_pair<TX:mode>_<P:mode>): Ditto.
564 2018-12-17 Uros Bizjak <ubizjak@gmail.com>
567 * internal-fn.def (ACOSH): New.
570 * optabs.def (acosh_optab): New.
571 (asinh_optab): Ditto.
572 (atanh_optab): Ditto.
573 * config/i386/i386-protos.h (ix86_emit_i387_asinh): New prototype.
574 (ix86_emit_i387_acosh): Ditto.
575 (ix86_emit_i387_atanh): Ditto.
576 * config/i386/i386.c (ix86_emit_i387_asinh): New function.
577 (ix86_emit_i387_acosh): Ditto.
578 (ix86_emit_i387_atanh): Ditto.
579 * config/i386/i386.md (asinhxf2): New expander.
580 (asinh<mode>2): Ditto.
582 (acosh<mode>2): Ditto.
584 (atanh<mode>2): Ditto.
586 2018-12-17 David Edelsohn <dje.gcc@gmail.com>
588 * config.gcc (powerpc-ibm-aix6.*): Delete extra_headers.
589 (powerpc-ibm-aix7.1.*): Same.
590 (powerpc-ibm-aix[789].*): Same.
592 2018-12-17 H.J. Lu <hongjiu.lu@intel.com>
595 * dwarf2out.c (find_AT_string_in_table): Add insert argument
596 defaulting to INSERT and replace INSERT.
597 (find_AT_string): Likewise.
598 (output_macinfo_op): Pass NO_INSERT to find_AT_string.
600 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
602 * coverage.c (struct conts_entry): Add n_counts.
603 (remap_counts_file): Record number of ocunts.
604 (get_coverage_counts): Verify that counts match.
605 * coverage.h (get_coverage_counts): Update prototype.
606 * profile.c (get_exec_counts. compute_value_histograms): Add
609 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
611 PR rtl-optimization/88253
612 * combine.c (combine_simplify_rtx): Test for side-effects before
613 substituting by zero.
615 2018-12-17 Richard Sandiford <richard.sandiford@arm.com>
617 * doc/invoke.texi (-fversion-loops-for-strides): Document
618 (loop-versioning-group-size, loop-versioning-max-inner-insns)
619 (loop-versioning-max-outer-insns): Document new --params.
620 * Makefile.in (OBJS): Add gimple-loop-versioning.o.
621 * common.opt (fversion-loops-for-strides): New option.
622 * opts.c (default_options_table): Enable fversion-loops-for-strides
624 * params.def (PARAM_LOOP_VERSIONING_GROUP_SIZE)
625 (PARAM_LOOP_VERSIONING_MAX_INNER_INSNS)
626 (PARAM_LOOP_VERSIONING_MAX_OUTER_INSNS): New parameters.
627 * passes.def: Add pass_loop_versioning.
628 * timevar.def (TV_LOOP_VERSIONING): New time variable.
629 * tree-ssa-propagate.h
630 (substitute_and_fold_engine::substitute_and_fold): Add an optional
632 * tree-ssa-propagate.c
633 (substitute_and_fold_engine::substitute_and_fold): Likewise.
634 When passed, only walk blocks dominated by that block.
635 * tree-vrp.h (range_includes_p): Declare.
636 (range_includes_zero_p): Turn into an inline wrapper around
638 * tree-vrp.c (range_includes_p): New function, generalizing...
639 (range_includes_zero_p): ...this.
640 * tree-pass.h (make_pass_loop_versioning): Declare.
641 * gimple-loop-versioning.cc: New file.
643 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
645 * ipa-fnsummary.c (remap_edge_change_prob): Do not ICE when changes
648 2018-12-15 H.J. Lu <hongjiu.lu@intel.com>
651 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Revert
654 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
656 * ipa-fnsummary.c (analyze_function_body): Do not loeak conds and
658 (ipa_fn_summary_generate): Add prevails parameter; do not allocate
659 data when symbol is not prevailing.
660 (inline_read_section): Likewise.
662 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
664 * cgraph.h (cgraph_node): Add predicate prevailing_p.
665 (cgraph_edge): Add predicate possible_call_in_translation_unit_p.
666 * ipa-prop.c (ipa_write_jump_function): Optimize streaming of ADDR_EXPR.
667 (ipa_read_jump_function): Add prevails parameter; optimize streaming.
668 (ipa_read_edge_info): Break out from ...
669 (ipa_read_node_info): ... here; optimize streaming.
670 * cgraph.c (cgraph_edge::possibly_call_in_translation_unit_p): New
673 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
675 * ipa-utils.c (ipa_merge_profiles): Do no merging when source function
678 2018-12-15 Jakub Jelinek <jakub@redhat.com>
680 PR tree-optimization/88464
682 * tree-vect-stmts.c (vect_build_gather_load_calls): For NARROWING
683 and mask with integral masktype, don't try to permute mask vectors,
684 instead emit VEC_UNPACK_{LO,HI}_EXPR. Fix up NOP_EXPR operand.
685 (vectorizable_store): Handle masked scatters with decl and integral
687 (permute_vec_elements): Allow scalar_dest to be NULL.
688 * config/i386/i386.c (ix86_get_builtin)
689 <case IX86_BUILTIN_GATHER3ALTDIV16SF>: Use lowpart_subreg for masks.
690 <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Don't assume mask and src have
693 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
695 * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET
697 (cgraph_build_static_cdtor): Update.
698 (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor
701 2018-12-15 Alan Modra <amodra@gmail.com>
703 * config/rs6000/rs6000.c (generate_set_vrsave, rs6000_emit_savres_rtx),
704 (rs6000_emit_prologue, rs6000_call_aix, rs6000_call_sysv),
705 (rs6000_call_darwin_1): Use gen_hard_reg_clobber.
707 2018-12-15 Jakub Jelinek <jakub@redhat.com>
710 * config/i386/sse.md (UNSPEC_SFIXUPIMM): New unspec enumerator.
711 (avx512f_sfixupimm<mode><mask_name><round_saeonly_name>): Use it
712 instead of UNSPEC_FIXUPIMM.
714 PR rtl-optimization/88478
715 * cselib.c (cselib_record_sets): Move sets[i].src_elt tests
716 after REG_P (dest) test.
718 2018-12-14 Martin Sebor <msebor@redhat.com>
720 PR tree-optimization/88372
721 * calls.c (maybe_warn_alloc_args_overflow): Handle function pointers.
722 * tree-object-size.c (alloc_object_size): Same. Simplify.
723 * doc/extend.texi (Object Size Checking): Update.
724 (Other Builtins): Add __builtin_object_size.
725 (Common Type Attributes): Add alloc_size.
726 (Common Variable Attributes): Ditto.
728 2018-12-14 Martin Sebor <msebor@redhat.com>
730 PR rtl-optimization/87096
731 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Avoid
732 folding calls whose bound may exceed INT_MAX. Diagnose bound ranges
733 that exceed the limit.
735 2018-12-14 Martin Sebor <msebor@redhat.com>
738 * doc/extend.texi (attribute const, pure): Clarify.
740 2018-12-14 H.J. Lu <hongjiu.lu@intel.com>
742 * config/i386/i386.c (rest_of_insert_endbranch): Insert ENDBR
743 at the function entry only when -mmanual-endbr isn't used or
744 there is cf_check function attribute.
745 (ix86_attribute_table): Add cf_check.
746 * config/i386/i386.opt: Add -mmanual-endbr.
747 * doc/extend.texi: Document cf_check attribute.
748 * doc/invoke.texi: Document -mmanual-endbr.
750 2018-12-14 Thomas Schwinge <thomas@codesourcery.com>
751 Cesar Philippidis <cesar@codesourcery.com>
753 * omp-offload.c (inform_oacc_loop): New function.
754 (execute_oacc_device_lower): Use it to display loop parallelism.
756 2018-12-14 Jakub Jelinek <jakub@redhat.com>
760 * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument.
761 * expr.c (categorize_ctor_elements_1): Likewise. Compute it like
762 p_nz_elts, except don't multiply it by mult. Adjust recursive call.
763 Fix up COMPLEX_CST handling.
764 (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize
765 it and pass it through to categorize_ctor_elements_1.
766 (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers.
767 * gimplify.c (gimplify_init_constructor): Likewise. Don't force
768 ctor into readonly data section if num_unique_nonzero_elements is
769 smaller or equal to 1/8 of num_nonzero_elements and size is >= 64
772 2018-12-14 Thomas Preud'homme <thomas.preudhomme@arm.com>
774 * config/arm/arm.md (arm_movdi): Split if -mslow-flash-data and
775 source is a constant that would be loaded by literal pool.
776 (movsf expander): Generate a no_literal_pool_sf_immediate insn if
777 -mslow-flash-data is present, targeting hardfloat ABI and source is a
778 float constant that cannot be loaded via vmov.
779 (movdf expander): Likewise but generate a no_literal_pool_df_immediate
781 (arm_movsf_soft_insn): Split if -mslow-flash-data and source is a
782 float constant that would be loaded by literal pool.
783 (softfloat constant movsf splitter): Splitter for the above case.
784 (movdf_soft_insn): Split if -mslow-flash-data and source is a float
785 constant that would be loaded by literal pool.
786 (softfloat constant movdf splitter): Splitter for the above case.
787 * config/arm/constraints.md (Pz): Document existing constraint.
788 (Ha): Define constraint.
790 * config/arm/predicates.md (hard_sf_operand): New predicate.
791 (hard_df_operand): Likewise.
792 * config/arm/thumb2.md (thumb2_movsi_insn): Split if
793 -mslow-flash-data and constant would be loaded by literal pool.
794 * constant/arm/vfp.md (thumb2_movsi_vfp): Likewise and disable constant
795 load in VFP register.
796 (movdi_vfp): Likewise.
797 (thumb2_movsf_vfp): Use hard_sf_operand as predicate for source to
798 prevent match for a constant load if -mslow-flash-data and constant
799 cannot be loaded via vmov. Adapt constraint accordingly by
800 using Ha instead of E for generic floating-point constant load.
801 (thumb2_movdf_vfp): Likewise using hard_df_operand predicate instead.
802 (no_literal_pool_df_immediate): Add a clobber to use as the
803 intermediate general purpose register and also enable it after reload
804 but disable it constant is a valid FP constant. Add constraints and
805 generate a DI intermediate load rather than 2 SI loads.
806 (no_literal_pool_sf_immediate): Add a clobber to use as the
807 intermediate general purpose register and also enable it after
810 2018-12-14 Uros Bizjak <ubizjak@gmail.com>
813 * internal-fn.def (HYPOT): New.
814 * optabs.def (hypot_optab): New.
815 * config/i386/i386.md (hypot<mode>3): New expander.
817 2018-12-14 Jeff Law <law@redhat.com>
819 * target.def (post_cfi_startproc): Fix text.
821 2018-12-14 Sam Tebbs <sam.tebbs@arm.com>
823 * doc/tm.texi (TARGET_ASM_POST_CFI_STARTPROC): Define.
824 * doc/tm.texi.in (TARGET_ASM_POST_CFI_STARTPROC): Define.
825 * dwarf2out.c (dwarf2out_do_cfi_startproc): Trigger the hook.
826 * hooks.c (hook_void_FILEptr_tree): Define.
827 * hooks.h (hook_void_FILEptr_tree): Define.
828 * target.def (post_cfi_startproc): Define.
830 2018-12-14 Tom de Vries <tdevries@suse.de>
832 * lto-cgraph.c (verify_node_partition): New function.
833 (input_overwrite_node, input_varpool_node): Use verify_node_partition.
835 2018-12-14 H.J. Lu <hongjiu.lu@intel.com>
838 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
839 use get_frame_size ().
841 2018-12-14 Andrew Stubbs <ams@codesourcery.com>
843 * gcc/lra-int.h (lra_register_new_scratch_op): Add third parameter.
844 * gcc/lra-remat.c (update_scratch_ops): Pass icode to
845 lra_register_new_scratch_op.
846 * gcc/lra.c (struct sloc): Add icode field.
847 (lra_register_new_scratch_op): Add icode parameter.
848 Use icode to skip insns that have changed beyond recognition.
850 2018-12-14 Eric Botcazou <ebotcazou@adacore.com>
852 * dwarf2out.c (analyze_discr_in_predicate): Simplify.
853 (analyze_variants_discr): Deal with naked boolean discriminants.
855 2018-12-14 Claudiu Zissulescu <claziss@synopsys.com>
857 * config/arc/arc.c (arc_restore_callee_milli) Don't clobber off
860 2018-12-14 Claudiu Zissulescu <claziss@synopsys.com>
862 * config/arc/arc.h (reg_class): Add SIBCALL_REGS.
864 2018-12-14 Richard Biener <rguenther@suse.de>
866 * gimple-ssa-split-paths.c (is_feasible_trace): Remove
867 duplicated code block.
869 2018-12-14 Alan Modra <amodra@gmail.com>
871 PR rtl-optimization/88311
872 * config/rs6000/rs6000.c (rs6000_call_sysv): Do not mask cookie.
873 (rs6000_sibcall_sysv): Likewise.
875 2018-12-14 Segher Boessenkool <segher@kernel.crashing.org>
877 PR rtl-optimization/88001
878 * function.c (match_asm_constraints_1): Don't invalidly share RTL.
880 2018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com>
882 * auto-profile.c (afdo_annotate_cfg): Call update_max_bb_count even
883 if autofdo counts are all zeros.
885 2018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com>
887 * auto-profile.c (afdo_indirect_call): Skip generating histogram
888 value for internal call.
890 2018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com>
892 * auto-profile.c (AFDO_EINFO): New macro.
893 (class edge_info): New class.
894 (is_edge_annotated, set_edge_annotated): Delete.
895 (afdo_propagate_edge, afdo_propagate_circuit, afdo_propagate): Remove
896 parameter. Adjust edge count computation and annotation using class
898 (afdo_calculate_branch_prob, afdo_annotate_cfg): Likewise.
900 2018-12-13 Michael Ploujnikov <michael.ploujnikov@oracle.com>
902 * ipa-cp.c (print_all_lattices): Skip cp clones.
904 * ipa-cp.c: Fix various comment typos.
906 2018-12-13 Jakub Jelinek <jakub@redhat.com>
908 PR tree-optimization/88444
909 * tree-inline.c (fold_marked_statements): Iterate up to
910 last_basic_block_for_fn rather than n_basic_blocks_for_fn.
912 PR rtl-optimization/88470
913 * cfgcleanup.c (outgoing_edges_match): If the function is
914 shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake
915 edge to EXIT, return false.
917 2018-12-13 Martin Sebor <msebor@redhat.com>
919 * doc/extend.texi: Consistently use @code for const and volatile
920 qualifiers, the true and false constants, and asm statements.
922 2018-12-13 Vladimir Makarov <vmakarov@redhat.com>
924 PR rtl-optimization/88414
925 * lra-int.h (lra_asm_error_p): New.
926 * lra-assigns.c (lra_assign): Check lra_asm_error_p for checking
927 call crossed pseudo assignment correctness.
928 (lra_split_hard_reg_for): Set up lra_asm_error_p.
929 * lra-constraints.c (curr_insn_transform): Ditto.
930 * lra.c (lra_asm_error_p): New.
931 (lra): Reset lra_asm_error_p.
933 2018-12-13 Jakub Jelinek <jakub@redhat.com>
935 PR rtl-optimization/88416
936 * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify
937 even if !AUTO_INC_DEC.
939 2018-12-13 David Edelsohn <dje.gcc@gmail.com>
941 * config/rs6000/rs6000.c (rs6000_function_arg): Ensure type is
943 (rs6000_arg_partial_bytes): Same.
945 2018-12-13 Jason Merrill <jason@redhat.com>
947 * gdbinit.in (pp): New macro.
950 2018-12-13 Jakub Jelinek <jakub@redhat.com>
952 PR tree-optimization/88464
953 * tree-vect-stmts.c (vect_build_gather_load_calls): Handle INTEGER_TYPE
954 masktype if mask is non-NULL.
955 (vectorizable_load): Don't reject masked gather loads if masktype
956 in the decl is INTEGER_TYPE.
958 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
960 * config/s390/s390-builtins.def (s390_vec_double_s64): Map to
961 s390_vec_double_s64 instead of s390_vcdgb.
962 (s390_vec_double_u64): Map to s390_vec_double_u64 instead of
965 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
967 * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
968 ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
969 ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.
971 2018-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
973 * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
976 2018-12-13 Wei Xiao <wei3.xiao@intel.com>
978 * common/config/i386/i386-common.c (processor_names): Add cascadelake.
979 (processor_alias_table): Add cascadelake.
980 * config.gcc: Add -march=cascadelake.
981 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
983 * config/i386/i386.c (Add m_CASCADELAKE): New.
984 (processor_cost_table): Add cascadelake.
985 (get_builtin_code_for_version): Handle cascadelake.
986 * config/i386/i386.h (TARGET_CASCADELAKE, PROCESSOR_CASCADELAKE): New.
987 (PTA_CASCADELAKE): Ditto.
988 * doc/invoke.texi: Add -march=cascadelake.
990 2018-12-13 Jakub Jelinek <jakub@redhat.com>
993 * config/i386/i386.md (*movdi_internal, *movsi_internal,
994 *movhi_internal, *movqi_internal): Add alternative(s) to load
995 0 or -1 into k registers using kxor or kxnoq instructions.
998 * config/i386/sse.md (VI1248_AVX512VLBW, AVX512ZEXTMASK): New
1000 (<avx512>_testm<mode>3<mask_scalar_merge_name>,
1001 <avx512>_testnm<mode>3<mask_scalar_merge_name>): Merge patterns
1002 with VI12_AVX512VL and VI48_AVX512VL iterators into ones with
1003 VI1248_AVX512VLBW iterator.
1004 (*<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext,
1005 *<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext_mask,
1006 *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext,
1007 *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext_mask): New
1011 * config/i386/i386.md (*zero_extendsidi2, zero_extend<mode>di2,
1012 *zero_extend<mode>si2, *zero_extendqihi2): Add =*k, *km alternatives.
1014 2018-12-12 Tom de Vries <tdevries@suse.de>
1015 Jakub Jelinek <jakub@redhat.com>
1017 * omp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1018 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1019 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1020 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1021 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1022 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME): Fix up
1023 function types - remove one argument.
1025 2018-12-12 Martin Sebor <msebor@redhat.com>
1027 * doc/extend.texi (attribute copy): Fix typos.
1029 2018-12-12 Martin Sebor <msebor@redhat.com>
1031 * doc/extend.texi (Function Attributes): Clarify C++ aspects.
1032 (Variable Attributes): Same.
1033 (Type Attributes): Same.
1035 2018-12-12 Olivier Hainque <hainque@adacore.com>
1037 * config/aarch64/aarch64.c (aarch64_override_options): Once arch,
1038 cpu and tune were validated, insert SUBTARGET_OVERRIDE_OPTIONS if
1041 2018-12-12 Olivier Hainque <hainque@adacore.com>
1043 * config/aarch64/aarch64.c (PROBE_STACK_FIRST_REG) : Redefine as
1044 R9_REGNUM instead of 9.
1045 (PROBE_STACK_SECOND_REG): Redefine as R10_REGNUM instead of 10.
1047 2018-12-12 Eric Botcazou <ebotcazou@adacore.com>
1050 * config/sparc/sparc.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1051 (speculation_barrier): New instruction for V9.
1053 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
1054 Iain Sandoe <iain@sandoe.co.uk>
1057 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
1058 unless it has been used.
1059 (first_reg_to_save): Remove dead code.
1061 2018-12-12 Iain Sandoe <iain@sandoe.co.uk>
1063 * config/rs6000/rs6000-protos.h (rs6000_call_darwin): New.
1064 (rs6000_sibcall_darwin): New. (macho_call_template): Remove.
1065 * config/rs6000/rs6000.c (get_prev_label): Forward declaration.
1066 (rs6000_call_template_1): Handle Darwin.
1067 (macho_call_template): Remove.
1068 (rs6000_call_sysv): Remove handling for Darwin.
1069 (rs6000_call_darwin_1, rs6000_call_darwin): New
1070 (rs6000_sibcall_darwin): New.
1071 * config/rs6000/rs6000.md (define_expand “call”): Handle Darwin
1072 with its own expander. (define_expand “call_value”): Likewise.
1073 (define_expand “sibcall”): Likewise.
1074 (define_expand “sibcall_value”): Likewise.
1075 (call_nonlocal_sysv): Remove Darwin special-casing.
1076 (call_value_nonlocal_sysv): Likewise.
1078 2018-12-12 Iain Sandoe <iain@sandoe.co.uk>
1080 * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64): Remove.
1081 (call_nonlocal_darwin64): Remove.
1082 (call_value_indirect_nonlocal_darwin64): Remove.
1083 (call_value_nonlocal_darwin64): Remove.
1084 * config/rs6000/rs6000.c (rs6000_call_template_1): Handle Darwin with
1085 the same asm output as AIX/ELFv2. (rs6000_call_sysv): Preserve the
1086 CALL_LONG flag when needed for Darwin.
1087 * config/rs6000/rs6000.md (define expand “call”): Expand Darwin as
1089 (define_expand “call_value”): Likewise.
1090 (define_expand “sibcall”): Likewise.
1091 (define_expand “sibcall_value”): Likewise.
1092 (call_indirect_nonlocal_sysv): Mark the clobber mode P.
1093 (call_nonlocal_sysv): Likewise.
1094 (call_nonlocal_sysv_secure): Likewise.
1095 (call_value_indirect_nonlocal_sysv): Likewise.
1096 (call_value_nonlocal_sysv): Likewise.
1097 (call_value_nonlocal_sysv_secure): Likewise.
1098 (call_local64): Mark the clobber DI.
1099 (call_value_local64): Likewise.
1101 2018-12-11 Jakub Jelinek <jakub@redhat.com>
1103 PR tree-optimization/88444
1104 * tree-vrp.c (register_edge_assert_for_2): Only register assertions
1105 for conversions if rhs1 is a SSA_NAME.
1107 2018-12-11 Dimitar Dimitrov <dimitar@dinux.eu>
1109 * cfgexpand.c (asm_clobber_reg_is_valid): Also produce
1110 error when stack pointer is clobbered.
1111 (expand_asm_stmt): Refactor clobber check in separate function.
1113 2018-12-11 Eric Botcazou <ebotcazou@adacore.com>
1115 * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
1116 accounting for STACK_BOUNDARY 128.
1117 (STACK_DYNAMIC_OFFSET): Likewise.
1119 2018-12-11 Jakub Jelinek <jakub@redhat.com>
1122 * config/i386/i386.md (*x86_mov<SWI48:mode>cc_0_m1_neg_leu<SWI:mode>):
1123 New define_insn_and_split.
1125 2018-12-11 Richard Biener <rguenther@suse.de>
1129 * gimple.c (gimple_assign_set_rhs_with_ops): Revert previous
1131 * tree-complex.c (update_complex_assignment): Properly transfer
1132 or clean EH info around gimple_assign_set_rhs_with_ops.
1134 2018-12-11 Richard Earnshaw <rearnsha@arm.com>
1137 * config/aarch64/iterators.md (sizem1): Add sizes for SFmode and DFmode.
1138 (Vbtype): Add SFmode mapping.
1139 * config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete.
1140 (copysign<GPF:mode>3): New expand pattern.
1141 (copysign<GPF:mode>3_insn): New insn pattern.
1143 2018-12-11 Richard Biener <rguenther@suse.de>
1145 * ccmp.c (ccmp_candidate_p): Use GIMPLE API properly.
1147 2018-12-11 Jakub Jelinek <jakub@redhat.com>
1150 * doc/sourcebuild.texi (lto_incremental): Document new effective
1153 2018-12-11 Segher Boessenkool <segher@kernel.crashing.org>
1156 * config/rs6000/rs6000.c (rs6000_expand_zeroop_builtin): Use
1157 rs6000_isa_flags instead of rs6000_isa_flags_explicit to decide
1158 whether soft float is enabled.
1159 (rs6000_expand_mtfsb_builtin): Ditto.
1160 (rs6000_expand_set_fpscr_rn_builtin): Ditto.
1161 (rs6000_expand_set_fpscr_drn_builtin): Ditto.
1163 2018-12-10 Jeff Law <law@redhat.com>
1165 PR tree-optimization/80520
1166 * gimple-ssa-split-paths.c (is_feasible_trace): Recognize half
1167 diamonds that are likely if convertable.
1169 2018-12-10 Martin Sebor <msebor@redhat.com>
1171 PR tree-optimization/86196
1172 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
1173 base size only of arrays.
1175 2018-12-10 Segher Boessenkool <segher@kernel.crashing.org>
1177 * config.gcc (Obsolete configurations): Delete powerpc*-*-*spe*.
1178 (Unsupported targets): Add powerpc*-*-*spe*.
1179 (powerpc*-*-*spe*): Delete.
1180 (powerpc-*-eabispe*): Delete.
1181 (powerpc-*-rtems*spe*): Delete.
1182 (powerpc*-*-linux*spe*): Delete.
1183 (powerpc*-*-linux*): Do not handle the linux*spe* targets.
1184 (powerpc-wrs-vxworks*spe): Delete.
1185 (with_cpu setting code): Delete powerpc*-*-*spe* handling.
1186 * config.host (target powerpc*-*-*spe*): Delete.
1187 * doc/invoke.texi (PowerPC SPE Options): Delete.
1188 (PowerPC SPE Options): Delete.
1189 * config/powerpcspe: Delete.
1191 2018-12-10 Uros Bizjak <ubizjak@gmail.com>
1194 * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes,
1195 check operand 1 with vector_operand predicate.
1196 (ix86_expand_sse_movcc): For vector modes, check op_true with
1197 vector_operand, not nonimmediate_operand.
1199 2018-12-10 Richard Biener <rguenther@suse.de>
1201 * tree-affine.c (tree_to_aff_combination): Remove unreachable
1203 (aff_combination_expand): Cache on SSA names, not possibly
1204 on conversion trees. Avoid expanding cases we do not handle.
1206 2018-12-10 Richard Biener <rguenther@suse.de>
1208 PR tree-optimization/88427
1209 * vr-values.c (vr_values::extract_range_from_phi_node):
1210 Handle symbolic ranges conservatively when trying to drop
1213 2018-12-10 Martin Jambor <mjambor@suse.cz>
1216 * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure
1217 we check pointers against pointers.
1219 2018-12-10 Richard Biener <rguenther@suse.de>
1222 * gimple.c (gimple_assign_set_rhs_with_ops): Transfer EH
1223 info to a newly allocated stmt.
1225 2018-12-10 Jerome Lambourg <lambourg@adacore.com>
1227 * config/vxworksae.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): Define.
1229 2018-12-10 Fredrik Nyström <fredrik@lysator.liu.se>
1232 * config/sol2.h: Only use libgcc-unwind.map if
1233 ENABLE_SHARED_LIBGCC.
1235 2018-12-10 Bin Cheng <bin.cheng@linux.alibaba.com>
1237 * auto-profile.c (afdo_calculate_branch_prob): Convert profile_count
1238 to afdo precision quality.
1239 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1240 (afdo_annotate_cfg): Set profile_count for exit basic block.
1242 2018-12-10 Bin Cheng <bin.cheng@linux.alibaba.com>
1244 * auto-profile.c (string_table::get_index_by_decl)
1245 (function_instance::get_function_instance_by_decl): Avoid infinite
1246 recursion by using DECL_FROM_INLINE.
1248 2018-12-09 John David Anglin <danglin@gcc.gnu.org>
1250 * config.gcc (hppa*-*-linux*): Add pa/t-pa to tmake_file. Define
1252 * config/pa/pa-protos.h (pa_d_target_versions): New prototype.
1253 * config/pa/pa.h (TARGET_D_CPU_VERSIONS): Define.
1254 * config/pa/pa-d.c: New file.
1255 * config/pa/t-pa: New file.
1257 2018-12-08 Bin Cheng <bin.cheng@linux.alibaba.com>
1259 * ipa-cp.c (update_profiling_info): Call adjust_for_ipa_scaling for
1262 2018-12-08 Jakub Jelinek <jakub@redhat.com>
1265 * tree-nested.c (convert_local_reference_stmt): Handle clobbers where
1266 lhs is not a decl normally, don't call use_pointer_in_frame on that
1269 2018-12-08 Eric Botcazou <ebotcazou@adacore.com>
1271 PR rtl-optimization/88390
1272 * dwarf2cfi.c (struct dw_cfi_row): Add window_save field.
1273 (cfi_row_equal_p): Compare it.
1274 (dwarf2out_frame_debug_cfa_window_save): Add FAKE parameter.
1275 If FAKE is false, set window_save of the current row.
1276 (dwarf2out_frame_debug) <REG_CFA_TOGGLE_RA_MANGLE>: Call above
1277 function with FAKE parameter set to true.
1278 <REG_CFA_WINDOW_SAVE>: Likewise but with FAKE parameter set to false.
1279 (change_cfi_row): Emit a DW_CFA_GNU_window_save if necessary.
1281 2018-12-07 Peter Bergner <bergner@linux.ibm.com>
1284 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1285 -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
1286 Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
1287 * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
1288 require -mlong-double-128.
1290 2018-12-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1292 * config/aarch64/aarch64.c (aarch64_tuning_override_functions): Add
1294 (aarch64_parse_sve_width_string): Define.
1296 2018-12-07 Jeff Law <law@redhat.com>
1299 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Add
1300 m_update_global_ranges member. Add corresponding argument to ctor.
1301 * gimple-ssa-evrp-analyze.c
1302 (evrp_range_analyzer::evrp_range_analyzer): Add new argument and
1303 initialize m_update_global_ranges.
1304 (evrp_range_analyzer::set_ssa_range_info): Assert that we are
1305 updating global ranges.
1306 (evrp_range_analyzer::record_ranges_from_incoming_edge): Only
1307 update global ranges if explicitly requested.
1308 (evrp_range_analyzer::record_ranges_from_phis): Similarly.
1309 (evrp_range_analyzer::record_ranges_from_stmt): Similarly.
1310 * gimple-ssa-evrp.c (evrp_dom_walker): Pass new argument to
1311 evrp_range_analyzer ctor.
1312 * gimple-ssa-sprintf.c (sprintf_dom_walker): Similarly.
1313 * tree-ssa-dom.c (dom_opt_dom_walker): Similarly.
1315 2018-12-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1317 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum):
1318 Add SVE_NOT_IMPLEMENTED value.
1319 * config/aarch64/aarch64-protos.h (struct tune_params): Add sve_width
1321 * config/aarch64/aarch64.c (generic_tunings,cortexa35_tunings,
1322 cortexa53_tunings, cortexa57_tunings, cortexa72_tunings,
1323 cortexa73_tunings, exynosm1_tunings, thunderx_tunings,
1324 thunderx_tunings, tsv110_tunings, xgene1_tunings, qdf24xx_tunings,
1325 saphira_tunings, thunderx2t99_tunings, emag_tunings):
1327 (aarch64_estimated_poly_value): Define.
1328 (TARGET_ESTIMATED_POLY_VALUE): Define.
1330 2018-12-07 Paul A. Clarke <pc@us.ibm.com>
1333 * config/rs6000/mmintrin.h (_mm_packs_pu16): Correctly use "__vector".
1335 2018-12-07 Vladimir Makarov <vmakarov@redhat.com>
1337 PR rtl-optimization/88349
1338 * ira-costs.c (record_operand_costs): Check bigger reg class on
1341 2018-12-07 Richard Sandiford <richard.sandiford@arm.com>
1343 * config/aarch64/aarch64-sve.md (*mul<mode>3, *v<optab><mode>3):
1344 Split the patterns after reload if we don't need the predicate
1346 (*post_ra_mul<mode>3, *post_ra_v<optab><mode>3): New patterns.
1348 2018-12-07 Richard Sandiford <richard.sandiford@arm.com>
1350 * config/aarch64/iterators.md (SVE_UNPRED_FP_BINARY): New code
1352 (sve_fp_op): Handle minus and mult.
1353 * config/aarch64/aarch64-sve.md (*add<mode>3, *sub<mode>3)
1354 (*mul<mode>3): Split the patterns after reload if we don't
1355 need the predicate operand.
1356 (*post_ra_<sve_fp_op><mode>3): New pattern.
1358 2018-12-07 Bin Cheng <bin.cheng@linux.alibaba.com>
1360 * profile-count.h (profile_count::oeprator>=): Fix typo by inverting
1361 return condition when *this is precise zero.
1363 2018-12-07 Jakub Jelinek <jakub@redhat.com>
1366 * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
1367 for functions with naked attribute.
1370 * tree-vrp.c (extract_range_from_binary_expr): For POINTER_PLUS_EXPR
1371 with -fno-delete-null-pointer-checks, set_nonnull only if the pointer
1372 is non-NULL and offset is known to have most significant bit clear.
1373 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): For ADDR_EXPR
1374 of MEM_EXPR, return true if the MEM_EXPR has non-zero offset with
1375 most significant bit clear. If offset does have most significant bit
1376 set and -fno-delete-null-pointer-checks, don't return true even if
1377 the base pointer is non-NULL.
1379 2018-12-06 Alexandre Oliva <aoliva@redhat.com>
1381 * cselib.c (cselib_record_sets): Skip strict low part sets
1384 2018-12-06 Paul A. Clarke <pc@us.ibm.com>
1387 * config/rs6000/smmintrin.h: New file.
1388 * config.gcc: Add smmintrin.h to extra_headers for powerpc*-*-*.
1390 2018-12-06 Paul A. Clarke <pc@us.ibm.com>
1393 * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian.
1394 (_mm_unpacklo_pi8): Likewise.
1395 (_mm_mulhi_pi16): Likewise.
1396 (_mm_packs_pi16): Fix for big-endian. Use preferred API.
1397 (_mm_packs_pi32): Likewise.
1398 (_mm_packs_pu16): Likewise.
1399 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian.
1400 (_mm_cvtss_si64): Likewise.
1401 (_mm_cvtpi32x2_ps): Likewise.
1402 (_mm_shuffle_ps): Likewise.
1403 (_mm_movemask_pi8): Likewise.
1404 (_mm_mulhi_pu16): Likewise.
1405 (_mm_sad_pu8): Likewise.
1406 (_mm_sad_pu8): Likewise.
1407 (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API.
1408 (_mm_cvtpu8_ps): Likewise.
1409 (_mm_movemask_ps): Better #else case for big-endian (no functional
1411 (_mm_shuffle_pi16): Likewise.
1412 * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian.
1413 Better #else case for big-endian (no functional change).
1414 (_mm_movemask_epi8): Likewise.
1415 (_mm_shufflehi_epi16): Likewise.
1416 (_mm_shufflelo_epi16): Likewise.
1417 (_mm_shuffle_epi32): Likewise.
1418 (_mm_mul_epu32): Fix for big-endian.
1419 (_mm_bsrli_si128): Likewise.
1420 (_mm_cvtps_pd): Better #else case for big endian.
1421 (_mm_mulhi_epi16): Likewise.
1422 (_mm_mul_epu32): Likewise.
1423 (_mm_slli_si128): Likewise.
1424 (_mm_sll_epi16): Likewise.
1425 (_mm_sll_epi32): Likewise.
1426 (_mm_sra_epi16): Likewise.
1427 (_mm_sra_epi32): Likewise.
1428 (_mm_srl_epi16): Likewise.
1429 (_mm_srl_epi32): Likewise.
1430 (_mm_mulhi_epu16): Likewise.
1431 (_mm_sad_epu8): Likewise.
1432 * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian.
1433 (_mm_sub_ps): Likewise.
1434 * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode.
1435 * gcc/config/rs6000/tmmintrin.h (_mm_alignr_epi8): Use ENDIAN
1436 macros consistently (no functional changes).
1437 (_mm_alignr_pi8): Likewise.
1439 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1442 * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
1444 * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
1446 2018-12-06 Vladimir Makarov <vmakarov@redhat.com>
1449 * ira.c (ira_init_register_move_cost): Use info from
1450 hard_regno_mode_ok instead of contains_reg_of_mode.
1451 * ira-costs.c (contains_reg_of_mode): Don't use cost from bigger
1452 hard register class for some fixed hard registers.
1454 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1456 * doc/extend.texi (Using Assembly Language with C): Document asm inline.
1457 (Size of an asm): Fix typo. Document asm inline.
1458 * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
1459 * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
1460 (gimple_asm_set_volatile): Fix typo.
1461 (gimple_asm_inline_p): New.
1462 (gimple_asm_set_inline): New.
1463 * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
1465 * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
1466 gimple_asm_inline_p flag, too.
1467 * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
1469 * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
1470 a minimum size for an asm.
1471 * tree.h (ASM_INLINE_P): New.
1473 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1476 * doc/extend.texi (Basic Asm): Update grammar.
1477 (Extended Asm): Update grammar.
1479 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1482 * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
1483 * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
1484 stack alignment in non-leaf functions.
1486 2018-12-06 Jakub Jelinek <jakub@redhat.com>
1489 * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
1490 call output_operand_lossage on VOIDmode CONST_INTs. After
1491 output_operand_lossage do return false.
1493 2018-12-06 Richard Biener <rguenther@suse.de>
1495 * df-problems.c (df_rd_local_compute): Use bitmap_release.
1496 (df_live_free): Likewise.
1497 (df_md_local_compute): Likewise.
1498 (df_md_free): Release df_md_scratch bitmap.
1499 * loop-invariant.c (calculate_loop_reg_pressure): Use
1501 * sched-deps.c (true_dependency_cache, output_dependency_cache,
1502 anti_dependency_cache, control_dependency_cache,
1503 spec_dependency_cache): Use bitmap instead of bitmap_head *.
1504 * sched-ebb.c (schedule_ebbs_init): Initialize non-GTY
1505 dont_calc_deps as bitmap allocated from obstack not GC.
1506 (schedule_ebbs_finish): Use bitmap_release.
1507 * sched-rgn.c (schedule_insns): Initialize non-GTY
1508 not_in_df as bitmap allocated from obstack not GC.
1510 * sel-sched.c (_forced_ebb_heads): Remove premature optimization.
1511 (sel_region_init): Allocate forced_ebb_heads.
1512 (sel_region_finish): Free forced_ebb_heads.
1514 2018-12-06 Richard Biener <rguenther@suse.de>
1516 * bitmap.c (bitmap_head::crashme): Define.
1517 * bitmap.h (bitmap_head): Add constexpr default constructor
1518 poisoning the obstack member.
1519 (bitmap_head::crashme): Declare.
1520 (bitmap_release): New function clearing a bitmap and poisoning
1522 * gengtype.c (main): Make it recognize CONSTEXPR.
1523 * lra-constraints.c (lra_inheritance): Use bitmap_release
1524 instead of bitmap_clear.
1525 * ira.c (ira): Work around class-memaccess warning.
1526 * regrename.c (create_new_chain): Likewise.
1528 2018-12-06 Jakub Jelinek <jakub@redhat.com>
1530 PR tree-optimization/85726
1531 * generic-match-head.c (optimize_successive_divisions_p): New function.
1532 * gimple-match-head.c (optimize_successive_divisions_p): Likewise.
1533 * match.pd: Don't combine successive divisions if they aren't exact
1534 and optimize_successive_divisions_p is false.
1537 * tree-nested.c (convert_nonlocal_reference_stmt): Remove clobbers
1538 for non-local automatic decls.
1540 2018-12-05 David Edelsohn <dje.gcc@gmail.com>
1542 * config/rs6000/aix72.h (ASM_DEFAULT_SPEC): Match Power7 processor
1545 2018-12-05 Uros Bizjak <ubizjak@gmail.com>
1547 * dwarf2out.c (field_byte_offset): Remove
1548 #ifdef PCC_BITFIELD_TYPE_MATTERS check.
1550 2018-12-05 Richard Biener <rguenther@suse.de>
1551 Martin Sebor <msebor@redhat.com>
1554 * calls.c (get_attr_nonstring_decl): Avoid setting *REF to
1556 * gcc/gimple-low.c (lower_stmt): Fold builtin calls here.
1557 * gimplify (maybe_fold_stmt): Avoid folding builtin calls.
1559 2018-12-05 Iain Sandoe <iain@sandoe.co.uk>
1561 * configure.ac (gcc_cv_otool): Set.
1562 * configure: Regenerate.
1564 2018-12-05 Uros Bizjak <ubizjak@gmail.com>
1566 * config/i386/cygming.h (PCC_BITFIELD_TYPE_MATTERS): Remove.
1567 (GROUP_BITFIELDS_BY_ALIGN): Ditto.
1569 2018-12-05 Richard Sandiford <richard.sandiford@arm.com>
1571 PR tree-optimization/88064
1572 * tree-vect-patterns.c (vect_convert_input): Convert the result of
1573 an existing cast if it has the right width but the wrong sign.
1574 Do not test the signedness of the required result when
1575 considering whether to split an existing cast; instead split to
1576 a type with the same signedness as the source of the cast, then
1577 convert it to the opposite signedness where necessary.
1578 (vect_recog_over_widening_pattern): Handle sign changes between
1579 the final PLUS_EXPR and the RSHIFT_EXPR.
1580 (vect_recog_average_pattern): Use an unsigned operation when
1581 truncating an addition, subtraction or multiplication. Cast the
1582 result back to the "real" signedness before promoting.
1584 2018-12-05 Richard Biener <rguenther@suse.de>
1586 PR tree-optimization/86637
1587 * tree-vectorizer.c (pass_slp_vectorize::execute): Reset
1588 vect_location at the end.
1590 2018-12-05 Jakub Jelinek <jakub@redhat.com>
1592 PR tree-optimization/87360
1593 * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
1594 data dependencies, don't return false, just continue. Formatting
1596 (merge_loop_tree, bb_prevents_fusion_p, unroll_jam_possible_p,
1597 fuse_loops): Formatting fixes.
1600 * cfgexpand.c (expand_stack_vars): If asan_vec is empty, start with
1601 aligning frame offset to ASAN_RED_ZONE_SIZE bytes.
1603 2018-12-05 Xianmiao Qu <xianmiao_qu@c-sky.com>
1605 * config.gcc (csky-*-linux-gnu*): Force .init_array support.
1607 2018-12-05 Segher Boessenkool <segher@kernel.crashing.org>
1609 * config/rs6000/rs6000.c (rs6000_function_arg): Only do the special
1610 aggregate handling on actual AIX, not on somewhat similar systems.
1611 (rs6000_arg_partial_bytes): Ditto.
1613 2018-12-04 Vladimir Makarov <vmakarov@redhat.com>
1615 PR rtl-optimization/88317
1616 * lra-constraints.c (split_reg): Don't set up check_only_regs if
1617 we are outside the inheritance pass.
1619 2018-12-04 Jan Hubicka <hubicka@ucw.cz>
1621 * ipa-prop.c (jump_function_useful_p): New.
1622 (ipa_read_node_info): Do not allocated useless jump functions.
1624 2018-12-04 David Malcolm <dmalcolm@redhat.com>
1627 * spellcheck.c (selftest::test_suggestions): Add another assertion.
1629 2018-12-04 Martin Sebor <msebor@redhat.com>
1631 * doc/extend.texi (__builtin_has_attribute): Fix typo.
1633 2018-12-04 Jakub Jelinek <jakub@redhat.com>
1636 * config/rs6000/rs6000.c (ccr_bit): Return -1 instead of assertion
1638 (print_operand): Use REG_P instead of GET_CODE == REG.
1639 <case 'D'>: Also check CR_REGNO_P (REGNO (x)).
1640 <case 't'>: Likewise. Remove GET_MODE check.
1641 * config/rs6000/rs6000.md (scc patterns): Assert ccr_bit didn't
1644 2018-12-04 Martin Sebor <msebor@redhat.com>
1646 * doc/extend.texi (attribute aligned): Expand.
1648 2018-12-04 David Edelsohn <dje.gcc@gmail.com>
1651 * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates
1653 (rs6000_arg_partial_bytes): Same.
1655 2018-12-04 Vladimir Makarov <vmakarov@redhat.com>
1658 * ira-costs.c (exec): Try bigger class to use smaller register
1661 2018-12-04 Michael Ploujnikov <michael.ploujnikov@oracle.com>
1664 * ipa-cp.c (create_specialized_node): Track clone counters by
1665 node assembler names.
1666 (ipcp_driver): Change type of clone_num_suffixes key to const char*.
1668 2018-12-04 Richard Biener <rguenther@suse.de>
1670 * tree-vect-loop.c (vect_transform_loop): Fix epilogue iterations
1671 for epilogue vectorization.
1673 2018-12-04 Richard Biener <rguenther@suse.de>
1675 PR tree-optimization/88301
1676 * tree-vrp.c (register_edge_assert_for_2): Fix sign-conversion
1677 issues in last commit.
1679 2018-12-04 Richard Biener <rguenther@suse.de>
1681 PR tree-optimization/88315
1682 * tree-vect-loop.c (get_initial_defs_for_reduction): Simplify
1683 and fix initialization vector for SAD and DOT_PROD SLP reductions.
1685 2018-12-03 Sandra Loosemore <sandra@codesourcery.com>
1688 * doc/extend.texi (Nonlocal gotos): New section.
1690 2018-12-03 Uros Bizjak <ubizjak@gmail.com>
1692 * config/i386/sse.md (vec_concatv2df): Change (v,xm,C)
1693 alternative to (v,vm,c).
1695 2018-12-03 Segher Boessenkool <segher@kernel.crashing.org>
1697 * gcc/config.gcc (Unsupported targets): Add powerpc*-*-linux*paired*.
1698 (powerpc*-*-linux*): Don't handle it here.
1700 2018-12-03 Jeff Law <law@redhat.com>
1702 * config/h8300/h8300.md (<code>qi3_1): Add gcc_unreachable to keep
1705 * config/h8300/h8300.md (low byte masking peepholes): Only accept
1706 nonimmediate_operand for the to-be-masked operand.
1708 2018-12-03 Richard Biener <rguenther@suse.de>
1710 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1711 allow expected function-pointer cast re-instantiation.
1713 2018-12-03 Ilya Leoshkevich <iii@linux.ibm.com>
1715 * common/config/s390/s390-common.c (s390_option_init_struct):
1716 Use section anchors by default.
1717 * config/s390/s390.c (s390_check_symref_alignment): Handle
1719 (TARGET_MAX_ANCHOR_OFFSET): Use short displacement.
1720 * output.h (assemble_align): Pass `align' as unsigned int, so
1721 that the value 0x80000000, which corresponds to `aligned(1 <<
1722 28)', would pass the `align > BITS_PER_UNIT' check.
1723 * varasm.c (assemble_align): Likewise.
1725 2018-12-03 Julian Brown <julian@codesourcery.com>
1727 * tree-pretty-print.c (dump_omp_clause): Make default case
1730 2018-12-03 Richard Biener <rguenther@suse.de>
1732 PR tree-optimization/88301
1733 * tree-vrp.c (register_edge_assert_for_2): Handle conversions
1734 that do not change the value by registering the same assert
1737 2018-12-03 Richard Biener <rguenther@suse.de>
1739 * tree-vect-loop.c (vect_transform_loop): Properly compute
1740 upper bound for the epilogue when doing epilogue vectorization.
1742 2018-12-03 Ilya Leoshkevich <iii@linux.ibm.com>
1745 * cfgcleanup.c (class pass_postreload_jump): New pass.
1746 (pass_postreload_jump::execute): Likewise.
1747 (make_pass_postreload_jump): Likewise.
1748 * passes.def: Add pass_postreload_jump before
1749 pass_postreload_cse.
1750 * tree-pass.h (make_pass_postreload_jump): New pass.
1752 2018-12-02 Uros Bizjak <ubizjak@gmail.com>
1754 * config/i386/i386.c (inline_memory_move_cost): Correctly
1755 compute number of TFmode GPR moves.
1757 2018-12-02 Jakub Jelinek <jakub@redhat.com>
1760 * config/i386/sse.md (*vec_concatv4sf_0, *vec_concatv4si_0): New insns.
1762 2018-12-02 Jeff Law <law@redhat.com>
1764 * config/h8300/h8300.md (call, call_value): Drop mode from
1765 operand containing number of bytes pushed.
1766 (call_insn, call_value_insn): Use mode iterator for number
1767 of bytes pushed operand. Make pattern name vary appropriately.
1769 2018-12-02 Jan Hubicka <jh@suse.cz>
1771 * bb-reorer.c (find_rarely_executed_basic_blocks_and_crossin_edges):
1772 trust precise profiles.
1774 2018-12-02 Jakub Jelinek <jakub@redhat.com>
1776 * tree-nested.c (convert_nonlocal_omp_clauses,
1777 convert_local_omp_clauses): Handle OMP_CLAUSE_IN_REDUCTION,
1778 OMP_CLAUSE_TASK_REDUCTION and OMP_CLAUSE__SIMT_ clauses.
1779 (convert_nonlocal_reference_stmt, convert_local_reference_stmt):
1780 Convert clauses for GIMPLE_OMP_TASKGROUP.
1782 * omp-low.c (check_omp_nesting_restrictions): Allow cancel or
1783 cancellation point with taskgroup clause inside of taskloop. Consider
1784 a taskloop construct without nogroup clause as implicit taskgroup for
1785 diagnostics if cancel/cancellation point with taskgroup clause is
1786 closely nested inside of taskgroup region.
1789 * asan.c (asan_clear_shadow): Move assert that len is multiple of 4
1790 to the start of the function.
1791 (asan_emit_stack_protection): When emitting clearing sequence for
1792 epilogue, align offset down to ASAN_RED_ZONE_SIZE granularity,
1793 add last_size_aligned which is last_size padded to multiples of
1794 ASAN_RED_ZONE_SIZE and emit asan_clear_shadow always on 4 byte
1797 2018-12-01 Jakub Jelinek <jakub@redhat.com>
1800 * asan.c (asan_redzone_buffer::flush_redzone_payload): Fix up
1801 an off-by-one for BYTES_BIG_ENDIAN.
1804 * combine.c (find_split_point): For invalid memory address
1805 nonobj + obj + const, if reg + obj + const is valid addressing
1806 mode, split at nonobj. Use if rather than else if for the
1807 fallback. Comment fixes.
1809 2018-11-30 Indu Bhagat <indu.bhagat@oracle.com>
1811 * coverage.c (get_coverage_counts): Use from_function_decl for precise
1813 * profile-count.c (profile_count::dump): Add handling for precise
1815 * profile.c (compute_branch_probabilities): Rely on exec_counts instead
1816 of profile_info to set profile_status of function.
1817 (branch_prob): Do not set profile_status of function based on
1818 profile_info. Done above based on exec_counts.
1820 2018-11-30 Jakub Jelinek <jakub@redhat.com>
1822 PR tree-optimization/88274
1823 * tree-ssa-reassoc.c (optimize_range_tests_xor,
1824 optimize_range_tests_diff): If type has smaller precision than
1825 corresponding mode or if it has non-standard min/max, compute
1826 everything in a standard type for the precision.
1829 * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
1830 * tree-ssa-ifcombine.c (ifcombine_ifandif): If
1831 --param logical-op-non-short-circuit is present, override
1832 LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
1833 * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
1835 2018-11-30 Jeff Law <law@redhat.com>
1837 * optabs.c (expand_binop): Use "machine_mode" rather than
1838 "enum machine mode" in most recent change.
1840 2018-11-30 Wilco Dijkstra <wdijkstr@arm.com>
1843 * builtins.c (expand_builtin_longjmp): Use a temporary when restoring
1845 (expand_builtin_nonlocal_goto): Likewise.
1847 2018-11-30 David Malcolm <dmalcolm@redhat.com>
1849 * diagnostic-core.h (emit_diagnostic): New decl.
1850 * diagnostic.c (emit_diagnostic): New overload, taking a
1853 2018-11-30 David Malcolm <dmalcolm@redhat.com>
1855 * pretty-print.c (class selftest::test_pretty_printer): New
1856 subclass of pretty_printer.
1857 (selftest::test_prefixes_and_wrapping): New test.
1858 (selftest::pretty_print_c_tests): Call it.
1860 2018-11-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
1862 Minimize clone counter memory usage in create_virtual_clone.
1863 * cgraph.h (clone_function_name): Add a variant that takes a
1865 * cgraph.h (cgraph_node::create_virtual_clone): Add a new
1866 argument: num_suffix.
1867 * cgraphclones.c (cgraph_node::create_virtual_clone): Pass
1868 num_suffix to clone_function_name.
1869 (clone_function_name): Add a variant that takes a tree decl.
1870 * ipa-cp.c (create_specialized_node): Keep track of clone
1871 counters in clone_num_suffixes hash map.
1872 (ipcp_driver): Free the counter hash map.
1873 * ipa-hsa.c (process_hsa_functions): Creates at most one hsa
1876 2018-11-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
1877 Make function assembly more independent.
1879 This is achieved by having clone_function_name assign unique clone
1880 numbers for each function independently.
1882 * cgraphclones.c: Replaced clone_fn_id_num with clone_fn_ids;
1884 (clone_function_name_numbered): Use clone_fn_ids.
1886 2018-11-30 Vladimir Makarov <vmakarov@redhat.com>
1888 PR rtl-optimization/88179
1889 * lra-constraints.c (address_eliminator::address_eliminator):
1890 Don't eleminate regs in illegitimate address.
1892 2018-11-30 David Malcolm <dmalcolm@redhat.com>
1894 PR preprocessor/88257
1895 * input.c (get_substring_ranges_for_loc): Fix indentation.
1896 Bulletproof against getting a different files back from
1897 linemap_resolve_location and expand_location_to_spelling_point.
1899 2018-11-30 Alexander Monakov <amonakov@ispras.ru>
1901 PR gcov-profile/88279
1902 * tree-profile.c (init_ic_make_global_vars): Call
1903 decl_default_tls_model for variable and not it's type.
1905 2018-11-30 Sam Tebbs <sam.tebbs@arm.com>
1907 * config/arm/arm.h (TARGET_DOTPROD): Add TARGET_VFP5 constraint.
1909 2018-11-30 Martin Liska <mliska@suse.cz>
1912 * asan.c (asan_shadow_cst): Remove, partially transform
1913 into flush_redzone_payload.
1914 (RZ_BUFFER_SIZE): New.
1915 (struct asan_redzone_buffer): New.
1916 (asan_redzone_buffer::emit_redzone_byte): Likewise.
1917 (asan_redzone_buffer::flush_redzone_payload): Likewise.
1918 (asan_redzone_buffer::flush_if_full): Likewise.
1919 (asan_emit_stack_protection): Use asan_redzone_buffer class
1920 that is responsible for proper aligned stores and flushing
1921 of shadow memory payload.
1922 * asan.h (ASAN_MIN_RED_ZONE_SIZE): New.
1923 (asan_var_and_redzone_size): Likewise.
1924 * cfgexpand.c (expand_stack_vars): Use smaller alignment
1925 (ASAN_MIN_RED_ZONE_SIZE) in order to make shadow memory
1926 for automatic variables more compact.
1928 2018-11-30 Alan Modra <amodra@gmail.com>
1930 * config/rs6000/predicates.md (easy_fp_constant): Avoid long
1931 dependent insn sequences.
1932 * config/rs6000/rs6000.c (num_insns_constant): Support long
1934 * config/rs6000/rs6000.md (mov<mode>_softfloat <FMOVE128>): Adjust
1937 2018-11-30 Alan Modra <amodra@gmail.com>
1939 * config/rs6000/rs6000.c (num_insns_constant_gpr): Renamed from
1940 num_insns_constant_wide. Make static. Revise comment.
1941 (num_insns_constant_multi): New function.
1942 (num_insns_constant): Formatting. Correct CONST_WIDE_INT
1943 calculation. Simplify and extract code common to both
1944 CONST_INT and CONST_DOUBLE. Add gcc_unreachable for unhandled
1946 * config/rs6000/rs6000-protos.h (num_insns_const_wide): Delete.
1948 2018-11-30 Richard Biener <rguenther@suse.de>
1950 PR tree-optimization/88274
1951 * tree-vrp.c (ranges_from_anti_range): Fix handling of
1954 2018-11-29 Andi Kleen <ak@linux.intel.com>
1956 * config/i386/i386.c (current_fentry_section): Add.
1957 (x86_function_profiler): Handle fentry section.
1958 (ix86_attribute_table): Add fentry section.
1959 * config/i386/i386.opt: Add -mfentry-section.
1960 * doc/extend.texi: Document fentry_section attribute.
1961 * doc/invoke.texi: Document -mfentry-section.
1963 2018-11-29 Andi Kleen <ak@linux.intel.com>
1965 * config/i386/i386.c (x86_print_call_or_nop): Handle nop name.
1966 (current_fentry_name): Add.
1967 (x86_function_profiler): Handle fentry_name attribute.
1968 (ix86_handle_fentry_name): Add.
1969 (ix86_attribute_table): Add fentry_name.
1970 * config/i386/i386.opt: Add -mfentry-name
1971 * doc/extend.texi: Document fentry_name.
1972 * doc/invoke.texi: Document minstrument-return.
1974 2018-11-29 Andi Kleen <ak@linux.intel.com>
1976 * config/i386/i386-opts.h (enum instrument_return): Add.
1977 * config/i386/i386.c (output_return_instrumentation): Add.
1978 (ix86_output_function_return): Call output_return_instrumentation.
1979 (ix86_output_call_insn): Call output_return_instrumentation.
1980 * config/i386/i386.opt: Add -minstrument-return=.
1981 * doc/invoke.texi (-minstrument-return): Document.
1983 2018-11-29 Eric Botcazou <ebotcazou@adacore.com>
1986 * config/sparc/sparc-modes.def: Minor tweak.
1987 * config/sparc/sparc.c: Minor reordering.
1988 (sparc_pass_by_reference): Move around.
1989 (traverse_record_type): Change offset from HOST_WIDE_INT to int.
1990 (classify_registers): Likewise for bitpos.
1991 (function_arg_slotno): Remove dead test and tweak comments.
1992 <MODE_RANDOM>: Remove useless assertion and test whether the
1993 parameter is named in order to pass it in FP registers. Return
1994 the regno for floating-point vector types.
1995 (compute_int_layout): Change bitpos from HOST_WIDE_INT to int.
1996 (compute_fp_layout): Likewise.
1997 (count_registers): Likewise.
1998 (assign_int_registers): Likewise.
1999 (assign_fp_registers): Likewise.
2000 (assign_registers): Likewise.
2001 (function_arg_record_value): Change size from HOST_WIDE_INT to int
2002 and use CEIL_NWORDS to compute the number of registers.
2003 (function_arg_union_value): Minor tweaks.
2004 (function_arg_vector_value): Add slotno and named parameters, use
2005 CEIL_NWORDS to compute the number of registers.
2006 (sparc_function_arg_1): Rework handling of vector types. Change
2007 size from HOST_WIDE_INT to int.
2008 (sparc_arg_partial_bytes): Rework handling of 32-bit ABI and deal
2009 with vector types for the 64-bt ABI.
2010 (sparc_function_arg_advance): Likewise.
2011 (sparc_return_in_memory): Add reference to -fpcc-struct-return.
2012 (sparc_struct_value_rtx): Return NULL_RTX instead of 0.
2013 (sparc_function_value_1): Rework handling of vector types. Change
2014 size from HOST_WIDE_INT to int.
2016 2018-11-29 Jakub Jelinek <jakub@redhat.com>
2019 * config/i386/sse.md (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_lt,
2020 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt,
2021 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_shift,
2022 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift,
2023 *<sse2_avx2>_pmovmskb_lt, *<sse2_avx2>_pmovmskb_zext_lt): New
2024 define_insn_and_split patterns.
2027 * config/i386/sse.md
2028 (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
2029 *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
2030 *<sse4_1_avx2>_pblendvb_lt): Change define_insn into
2031 define_insn_and_split.
2033 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
2036 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
2037 -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
2039 2018-11-29 Uros Bizjak <ubizjak@gmail.com>
2041 * config/i386/i386.c (inline_memory_move_cost):
2042 Check "in" for 2 in MMX_CLASS_P case.
2043 * config/i386/mmx.md (*mov<mode>_internal): Correct
2044 TARGET_INTER_UNIT_MOVES_FROM_VEC and TARGET_INTER_UNIT_MOVES_TO_VEC
2045 alternatives in preferred_for_speed attribute calculation.
2047 2018-11-29 Martin Sebor <msebor@redhat.com>
2051 * doc/extend.texi (attribute constructor): Clarify.
2053 2018-11-29 Martin Liska <mliska@suse.cz>
2056 * builtins.c (expand_movstr): Fix thinko introduced
2057 when switching to the new enum.
2059 2018-11-29 qing zhao <qing.zhao@oracle.com>
2061 * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code.
2062 * common.opt: Add -flive-patching flag.
2063 * doc/invoke.texi: Document -flive-patching.
2064 * flag-types.h (enum live_patching_level): New enum.
2065 * ipa-inline.c (can_inline_edge_p): Disable external functions from
2066 inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC.
2067 * opts.c (control_options_for_live_patching): New function.
2068 (finish_options): Make flag_live_patching incompatible with flag_lto.
2069 Control IPA optimizations based on different levels of
2072 2018-11-29 Giuliano Belinassi <giuliano.belinassi@usp.br>
2074 * match.pd (sinh (atanh (x))): New simplification rules.
2075 (cosh (atanh (x))): Likewise.
2077 2018-11-29 Jakub Jelinek <jakub@redhat.com>
2080 * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries.
2081 (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs.
2082 (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
2083 *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
2084 *<sse4_1_avx2>_pblendvb_lt): New define_insns.
2087 * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare.
2088 * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions.
2089 * match.pd (define_predicates): Add uniform_integer_cst_p.
2090 (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust
2091 so that it works also for vector comparisons with uniform constants
2092 with INTEGER_CST element.
2095 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
2096 vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
2097 in unsigned_type_for instead of vector integral type where overflow
2100 2018-11-29 Michael Ploujnikov <michael.ploujnikov@oracle.com>
2102 There can be at most one .resolver clone per function
2103 * config/rs6000/rs6000.c (make_resolver_func): Generate
2104 resolver symbol with clone_function_name instead of
2105 clone_function_name_numbered.
2107 2018-11-29 Richard Biener <rguenther@suse.de>
2109 PR tree-optimization/88243
2110 * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
2111 type of all pattern-sequence stmts to vect_internal_def.
2113 2018-11-29 Andre Vieira <andre.simoesdiasvieira@arm.com>
2116 * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
2117 (cortex-r7, cortex-r8): Update default and add new configuration.
2118 * doc/invoke.texi (armv7-r): Add two new vfp options.
2119 (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
2120 support this option.
2122 2018-11-29 Alan Modra <amodra@gmail.com>
2124 * config/rs6000/rs6000.c (rs6000_emit_move): Disable long
2125 double split for targets other than Darwin.
2127 2018-11-29 Alan Modra <amodra@gmail.com>
2129 * config/rs6000/rs6000.md (movdi_internal64): Correct lengths.
2131 2018-11-29 Alan Modra <amodra@gmail.com>
2133 * config/rs6000/predicates.md (easy_fp_constant): Remove code
2134 dealing with integers in const_double. Assert on unexpected
2135 modes. Delete superfluous ABI_V4 test.
2136 * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
2137 Don't call easy_fp_constant for Pmode.
2138 (rs6000_hash_constant): Remove code dealing with integers in
2140 (rs6000_legitimate_constant_p): Likewise.
2141 (output_toc): Formatting, use CONST_DOUBLE_P.
2143 2018-11-29 Alan Modra <amodra@gmail.com>
2145 * config/rs6000/constraints.md (G, H): Comment on purpose of
2146 constraint. Correct mode comments and "G" mode test.
2147 * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
2148 alternative handling "F". Add length attr. Formatting.
2149 (mov<mode>_softfloat <FMOVE128>): Delete "GH" from
2150 alternative, and move "F" to separate alternative. Correct
2153 2018-11-29 Alan Modra <amodra@gmail.com>
2155 * config/rs6000/predicates.md (easy_vector_constant): Don't call
2158 2018-11-29 Alan Modra <amodra@gmail.com>
2160 * config.in (HAVE_AS_PLTSEQ): Add.
2161 * config/rs6000/predicates.md (indirect_call_operand): New.
2162 * config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
2163 (rs6000_sibcall_sysv): Declare.
2164 * config/rs6000/rs6000.c (init_cumulative_args): Set cookie
2165 CALL_LONG for -fno-plt.
2166 (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
2167 (rs6000_indirect_call_template_1): Emit .reloc directives for
2168 UNSPEC_PLTSEQ calls.
2169 (rs6000_pltseq_template): New function.
2170 (rs6000_longcall_ref): Add arg parameter. Use PLT16 insns if
2171 relocs supported by assembler. Move SYMBOL_REF test to callers.
2172 (rs6000_call_aix): Adjust rs6000_longcall_ref call. Package
2173 insns in UNSPEC_PLTSEQ, preserving original func_desc.
2174 (rs6000_call_sysv): Likewise.
2175 (rs6000_sibcall_sysv): New function.
2176 * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
2177 * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
2178 UNSPEC_PLT16_LO): New.
2179 (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
2180 (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
2181 cookie in constraints. Test explicitly for flags in length attr.
2182 Handle unspec operand 1.
2183 (call_value_indirect_nonlocal_sysv): Likewise.
2184 (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
2185 (call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
2186 (sibcall, sibcall_value): Use rs6000_sibcall_sysv.
2187 (sibcall_indirect_nonlocal_sysv): New pattern.
2188 (sibcall_value_indirect_nonlocal_sysv): Likewise.
2189 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
2191 * configure.ac: Check for gas plt sequence marker support.
2192 * configure: Regenerate.
2194 2018-11-29 Alan Modra <amodra@gmail.com>
2196 * config/rs6000/predicates.md (unspec_tls): New.
2197 * config/rs6000/rs6000-protos.h (rs6000_call_template),
2198 (rs6000_sibcall_template): Update prototype.
2199 (rs6000_longcall_ref): Delete.
2200 (rs6000_call_sysv): Declare.
2201 * config/rs6000/rs6000.c (edit_tls_call_insn): New function.
2202 (global_tlsarg): New variable.
2203 (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
2205 (print_operand): Extract UNSPEC_TLSGD address operand.
2206 (rs6000_call_template, rs6000_sibcall_template): Remove arg
2207 parameter, extract from second call operand instead.
2208 (rs6000_longcall_ref): Make static, localize vars.
2209 (rs6000_call_aix): Rename parameter to reflect new usage. Take
2210 tlsarg from global_tlsarg. Don't create unused rtl or nop insns.
2211 (rs6000_sibcall_aix): Rename parameter to reflect new usage. Take
2212 tlsarg from global_tlsarg.
2213 (rs6000_call_sysv): New function.
2214 * config/rs6000/rs6000.md: Adjust rs6000_call_template and
2215 rs6000_sibcall_template throughout.
2216 (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
2217 (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
2218 (tls_gdld_nomark): New insn.
2219 (tls_gd): Swap operand order. Simplify mode selection.
2220 (tls_gd_high, tls_gd_low): Swap operand order.
2221 (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
2222 Simplify mode selection.
2223 (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
2224 (call, call_value): Don't assert for second call operand.
2225 Use rs6000_call_sysv.
2227 2018-11-29 Alan Modra <amodra@gmail.com>
2229 * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64),
2230 (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64),
2231 (call_value_nonlocal_darwin64): Remove constraints from second call
2232 arg, the rounded_stack_size_rtx arg.
2233 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix),
2234 (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix),
2235 (tls_ld_call_sysv, call_local32, call_local64, call_value_local32),
2236 (call_value_local64, call_indirect_nonlocal_sysv),
2237 (call_nonlocal_sysv, call_nonlocal_sysv_secure),
2238 (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv),
2239 (call_value_nonlocal_sysv_secure, call_local_aix),
2240 (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix),
2241 (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2),
2242 (call_value_indirect_elfv2, sibcall_local32, sibcall_local64),
2243 (sibcall_value_local32, sibcall_value_local64, sibcall_aix),
2244 (sibcall_value_aix): Likewise.
2246 2018-11-29 Alan Modra <amodra@gmail.com>
2248 * config/rs6000/rs6000.md (TLSmode): Delete mode iterator. Replace
2249 with P throughout except for call mems which should use SI.
2250 (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
2251 attributes. Replace with bits, mode and ptrload respectively.
2253 2018-11-29 Alan Modra <amodra@gmail.com>
2255 * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
2256 (rs6000_indirect_sibcall_template): Declare.
2257 * config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
2258 (rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
2260 * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
2261 (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
2262 (call_indirect_aix, call_value_indirect_aix): Use
2263 rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
2264 call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
2265 handle both speculation and non-speculation cases.
2266 (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
2267 (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
2269 2018-11-29 Alan Modra <amodra@gmail.com>
2271 * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
2272 (rs6000_sibcall_template): Declare.
2273 (macho_call_template): Rename from output_call.
2274 * config/rs6000/rs6000.c (rs6000_call_template_1): New function.
2275 (rs6000_call_template, rs6000_sibcall_template): Likewise.
2276 (macho_call_template): Rename from output_call.
2277 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
2278 (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
2279 (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
2280 (call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
2281 (call_value_nonlocal_sysv_secure, call_nonlocal_aix),
2282 (call_value_nonlocal_aix): Use rs6000_call_template and update
2283 occurrences of output_call to macho_call_template.
2284 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
2285 (sibcall_value_aix): Use rs6000_sibcall_template.
2287 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
2289 * config/rs6000/rs6000-string.c (expand_block_clear): Change how
2290 we determine if unaligned vsx is ok.
2292 2018-11-28 Jan Hubicka <jh@suse.cz>
2294 * predict.c (determine_unlikely_bbs): Forward declare; also determine
2295 edges that are always known to be taken because it is only likely
2296 edge out of given BB.
2297 (tree_estimate_probability): Call before profile guessing to get bit
2300 2018-11-28 Jan Hubicka <jh@suse.cz>
2302 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle
2303 profile_probability::always better.
2305 2018-11-28 Jan Hubicka <jh@suse.cz>
2307 * profile-count.h (profile_count::split): Give better result when
2308 splitting profile_probability::always.
2310 2018-11-28 Vladimir Makarov <vmakarov@redhat.com>
2313 * ira-costs.c (scan_one_insn): Process subregs when updating costs
2314 for pseudos and allocnos from insn.
2316 2018-11-28 David Edelsohn <dje.gcc@gmail.com>
2318 * config/rs6000/aix72.h: Update to match aix71.h changes.
2320 2018-11-28 Jeff Law <law@redhat.com>
2322 * config/h8300/constraints.md: Add "C" constraint for call insns.
2323 * config/h8300/h8300.md (call, call_value): Turn into a define_expand
2324 and define_insn pair. Move invalid call targets into a register in
2325 the expander and fix constraints in the matching pattern.
2326 * config/h8300/predicates.md (call_expander_operand): Renamed from
2327 call_insn_operand. Reject things we shouldn't be trying to handle.
2328 (call_insn_operand): New predicate for use by the call/call_value
2330 (small_call_insn_operand): Update appropriately.
2332 2018-11-28 Sam Tebbs <sam.tebbs@arm.com>
2334 * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
2335 calls to strtok with strtok_r.
2337 2018-11-28 Richard Biener <rguenther@suse.de>
2339 PR tree-optimization/88223
2340 * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
2341 over a stored-same value may-alias store make sure to consider
2342 partial overlaps which are valid when TBAA reasonings do not
2343 apply and byte-granular overlaps are possible at all.
2345 2018-11-28 Richard Biener <rguenther@suse.de>
2347 PR tree-optimization/88217
2348 * vr-values.c (vr_values::extract_range_from_phi_node): Make
2349 sure to handle results > +INF and < -INF correctly when
2350 trying to drop down to +INF - 1 or -INF + 1.
2352 2018-11-28 Alan Modra <amodra@gmail.com>
2354 * xcoffout.c (do_block): Signed/unsigned warning fix.
2356 2018-11-28 Richard Biener <rguenther@suse.de>
2358 PR tree-optimization/79351
2359 * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
2360 empty CONSTRUCTORs ensure the store is at a constant position.
2362 2018-11-28 Richard Biener <rguenther@suse.de>
2364 PR tree-optimization/88229
2365 * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
2366 is an INTEGER_CST before accessing it so.
2368 2018-11-28 Sam Tebbs <sam.tebbs@arm.com>
2370 * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
2372 2018-11-28 Jakub Jelinek <jakub@redhat.com>
2375 * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
2376 SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1. Formatting fixes.
2377 * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
2379 2018-11-27 Martin Liska <mliska@suse.cz>
2381 * gcov.c (generate_results): Append current_working_directory
2384 2018-11-27 Mike Gulick <mgulick@mathworks.com>
2386 PR preprocessor/83173
2387 * input.c (dump_location_info): Dump reason and included_from
2388 fields from line_map_ordinary struct. Fix indentation when
2389 location > 5 digits.
2390 * diagnostic-show-locus.c (num_digits, num_digits): Move to
2391 diagnostic.c to allow it to be utilized by input.c.
2392 * diagnostic.c (num_digits, selftest::test_num_digits): Moved
2394 (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
2395 * diagnostic.h (num_digits): Add extern definition.
2397 2018-11-27 Fredrik Noring <noring@nocrew.org>
2399 * config/mips/mips.c (mips_reorg_process_insns)
2400 (mips_option_override): Handle `-mfix-r5900'.
2401 * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
2403 * config/mips/mips.opt (mfix-r5900): New option.
2404 * doc/invoke.texi: Document the `r5900' processor name, and
2405 `-mfix-r5900' and `-mno-fix-r5900' options.
2407 2018-11-27 Jakub Jelinek <jakub@redhat.com>
2410 * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
2411 output_operand_lossage instead of gcc_assert.
2412 <case 't'>: Likewise.
2413 <case 'z'>: Likewise.
2414 <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
2416 2018-11-27 Jeff Law <law@redhat.com>
2418 * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
2419 in call to move_by_pieces.
2421 * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
2424 * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
2425 call to move_by_pieces.
2427 * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
2428 in call to move_by_pieces.
2430 * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
2431 RETURN_BEGIN in call to move_by_pieces.
2432 (microblaze_expand_block_move): Likewise.
2434 2018-11-27 Tamar Christina <tamar.christina@arm.com>
2436 * doc/md.texi (xorsign): Document it.
2438 2018-11-27 Martin Liska <mliska@suse.cz>
2440 * config/gnu-user.h (TARGET_F951_OPTIONS): New.
2441 * gcc.c (find_fortran_preinclude_file): New function
2442 to handle Fortran pre-include.
2444 2018-11-27 Martin Liska <mliska@suse.cz>
2446 * asan.c (asan_emit_stack_protection): Use new enum values
2447 instead of int constants.
2448 * builtins.c (expand_builtin_memory_copy_args): Replace int
2449 type with memop_ret enum type.
2450 (expand_builtin_mempcpy_args): Likewise.
2451 (expand_builtin_memcpy): Use new enum values
2452 instead of int constants. Likewise.
2453 (expand_builtin_mempcpy): Likewise.
2454 (expand_movstr): Likewise.
2455 (expand_builtin_strcpy_args): Likewise.
2456 (expand_builtin_stpcpy_1): Likewise.
2457 (expand_builtin_strncpy): Likewise.
2458 (expand_builtin_memset_args): Likewise.
2459 * expr.c (move_by_pieces_d::finish_endp): Rename to ...
2460 (move_by_pieces_d::finish_retmode): ... this.
2461 (move_by_pieces): Change last argument type to memop_ret.
2462 (store_by_pieces): Use new enum values
2463 instead of int constants.
2464 (emit_block_move_hints): Likewise.
2465 (emit_push_insn): Likewise.
2466 (store_expr): Likewise.
2467 * expr.h (store_by_pieces): Change int to newly added enum
2469 * rtl.h (enum memop_ret): Define.
2470 (move_by_pieces): Use the enum type.
2472 2018-11-27 Alan Modra <amodra@gmail.com>
2474 * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
2476 (ASM_CPU_SPEC): Do so here. Rewrite using if .. else if .. specs
2477 form. Error on missing -mcpu case.
2478 * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
2479 (asm_names <!_AIX>): Add missing cpus. Update NULL case. Apply
2480 PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
2481 * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
2482 else if .. specs form. Error on missing -mcpu case. Don't output
2483 duplicate -maltivec. Apply PR63177 fix for -mcpu=powerpc64le.
2485 2018-11-26 David Malcolm <dmalcolm@redhat.com>
2487 * dump-context.h (dump_context::dump_loc): Convert 1st param from
2488 dump_flags_t to const dump_metadata_t &. Convert 2nd param from
2489 const dump_location_t & to const dump_user_location_t &.
2490 (dump_context::dump_loc_immediate): Convert 2nd param from
2491 const dump_location_t & to const dump_user_location_t &.
2492 (dump_context::dump_gimple_stmt): Convert 1st param from
2493 dump_flags_t to const dump_metadata_t &.
2494 (dump_context::void dump_gimple_stmt_loc): Likewise; convert
2495 2nd param from const dump_location_t & to
2496 const dump_user_location_t &.
2497 (dump_context::dump_gimple_expr): Convert 1st param from
2498 dump_flags_t to const dump_metadata_t &.
2499 (dump_context::dump_gimple_expr_loc): Likewise; convert
2500 2nd param from const dump_location_t & to
2501 const dump_user_location_t &.
2502 (dump_context::dump_generic_expr): Convert 1st param from
2503 dump_flags_t to const dump_metadata_t &.
2504 (dump_context::dump_generic_expr_loc): Likewise; convert
2505 2nd param from const dump_location_t & to
2506 const dump_user_location_t &.
2507 (dump_context::dump_printf_va): Convert 1st param from
2508 dump_flags_t to const dump_metadata_t &.
2509 (dump_context::dump_printf_loc_va): Likewise; convert
2510 2nd param from const dump_location_t & to
2511 const dump_user_location_t &.
2512 (dump_context::dump_dec): Convert 1st param from
2513 dump_flags_t to const dump_metadata_t &.
2514 (dump_context::dump_symtab_node): Likewise.
2515 (dump_context::begin_scope): Split out 2nd param into
2516 user and impl locations.
2517 (dump_context::ensure_pending_optinfo): Add metadata param.
2518 (dump_context::begin_next_optinfo): Replace dump_location_t param
2519 with metadata and user location.
2520 * dumpfile.c (dump_context::dump_loc): Convert 1st param from
2521 dump_flags_t to const dump_metadata_t &. Convert 2nd param from
2522 const dump_location_t & to const dump_user_location_t &.
2523 (dump_context::dump_loc_immediate): Convert 2nd param from
2524 const dump_location_t & to const dump_user_location_t &.
2525 (dump_context::dump_gimple_stmt): Convert 1st param from
2526 dump_flags_t to const dump_metadata_t &.
2527 (dump_context::void dump_gimple_stmt_loc): Likewise; convert
2528 2nd param from const dump_location_t & to
2529 const dump_user_location_t &.
2530 (dump_context::dump_gimple_expr): Convert 1st param from
2531 dump_flags_t to const dump_metadata_t &.
2532 (dump_context::dump_gimple_expr_loc): Likewise; convert
2533 2nd param from const dump_location_t & to
2534 const dump_user_location_t &.
2535 (dump_context::dump_generic_expr): Convert 1st param from
2536 dump_flags_t to const dump_metadata_t &.
2537 (dump_context::dump_generic_expr_loc): Likewise; convert
2538 2nd param from const dump_location_t & to
2539 const dump_user_location_t &.
2540 (dump_context::dump_printf_va): Convert 1st param from
2541 dump_flags_t to const dump_metadata_t &.
2542 (dump_context::dump_printf_loc_va): Likewise; convert
2543 2nd param from const dump_location_t & to
2544 const dump_user_location_t &.
2545 (dump_context::dump_dec): Convert 1st param from
2546 dump_flags_t to const dump_metadata_t &.
2547 (dump_context::dump_symtab_node): Likewise.
2548 (dump_context::begin_scope): Split out 2nd param into
2549 user and impl locations.
2550 (dump_context::ensure_pending_optinfo): Add metadata param.
2551 (dump_context::begin_next_optinfo): Replace dump_location_t param
2552 with metadata and user location.
2553 (dump_gimple_stmt): Convert 1st param from dump_flags_t to
2554 const dump_metadata_t &.
2555 (dump_gimple_stmt_loc): Likewise; convert 2nd param from
2556 const dump_location_t & to const dump_user_location_t &.
2557 (dump_gimple_expr): Convert 1st param from dump_flags_t to
2558 const dump_metadata_t &.
2559 (dump_gimple_expr_loc): Likewise; convert 2nd param from
2560 const dump_location_t & to const dump_user_location_t &.
2561 (dump_generic_expr): Convert 1st param from dump_flags_t to
2562 const dump_metadata_t &.
2563 (dump_generic_expr_loc): Likewise; convert 2nd param from
2564 const dump_location_t & to const dump_user_location_t &.
2565 (dump_printf): Convert 1st param from dump_flags_t to
2566 const dump_metadata_t &.
2567 (dump_printf_loc): Likewise; convert 2nd param from
2568 const dump_location_t & to const dump_user_location_t &.
2569 (dump_dec): Convert 1st param from dump_flags_t to
2570 const dump_metadata_t &.
2571 (dump_symtab_node): Likewise.
2572 (dump_begin_scope): Split out 2nd param into user and impl
2574 (selftest::assert_impl_location_eq): New function.
2575 (ASSERT_IMPL_LOCATION_EQ): New macro.
2576 (selftest::test_impl_location): Update to use
2577 ASSERT_IMPL_LOCATION_EQ.
2578 (selftest::test_capture_of_dump_calls): Convert "loc" to
2579 dump_user_location_t. Add ASSERT_IMPL_LOCATION_EQ throughout,
2580 verifying line numbers of dump emissions.
2581 * dumpfile.h (class dump_metadata_t): New class.
2582 (dump_printf): Convert 1st param from dump_flags_t to
2583 const dump_metadata_t &.
2584 (dump_printf_loc): Likewise; convert 2nd param from
2585 const dump_location_t & to const dump_user_location_t &.
2586 (dump_generic_expr_loc): Likewise.
2587 (dump_generic_expr): Convert 1st param from dump_flags_t to
2588 const dump_metadata_t &.
2589 (dump_gimple_stmt_loc): Likewise; convert 2nd param from
2590 const dump_location_t & to const dump_user_location_t &.
2591 (dump_gimple_stmt): Convert 1st param from dump_flags_t to
2592 const dump_metadata_t &.
2593 (dump_gimple_expr_loc): Likewise; convert 2nd param from
2594 const dump_location_t & to const dump_user_location_t &.
2595 (dump_gimple_expr): Convert 1st param from dump_flags_t to
2596 const dump_metadata_t &.
2597 (dump_symtab_node): Likewise.
2598 (dump_dec): Likewise.
2599 (dump_begin_scope): Split out 2nd param into user and impl
2601 (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
2602 location and impl_location, and capture the impl_location.
2603 (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
2604 * loop-unroll.c (report_unroll): Update for changes to
2605 dump_printf_loc and dump_printf.
2606 * opt-problem.cc (opt_problem::opt_problem): Update for change to
2608 * optinfo-emit-json.cc
2609 (selftest::test_building_json_from_dump_calls): Convert "loc" from
2610 dump_location_t to dump_user_location_t.
2611 * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
2613 * profile.c (compute_branch_probabilities): Update for change to
2615 * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
2616 * tree-vect-slp.c (vect_print_slp_tree): Update for change to
2619 2018-11-27 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2621 * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
2622 and check_effective_target_ptr_eq_long.
2624 2018-11-27 Alan Modra <amodra@gmail.com>
2626 * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
2627 (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
2628 (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
2629 (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
2630 (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
2631 * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
2632 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2633 (ASM_APP_ON, ASM_APP_OFF): Don't define.
2634 * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2635 * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2636 * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2637 * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
2638 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2639 * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2640 * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
2641 (GNU_USER_TARGET_CRTN): Redefine.
2642 (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
2643 (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
2644 (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
2645 (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
2646 (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
2647 (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
2648 (LINK_EH_SPEC): Don't define.
2650 2018-11-26 Martin Sebor <msebor@redhat.com>
2653 * expr.c (string_constant): Handle top-level decls of all character
2654 types and subobjects of narrow character type.
2656 2018-11-27 Alan Modra <amodra@gmail.com>
2658 * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
2659 * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
2660 GNU_USER_TARGET_OS_CPP_BUILTINS.
2661 (RS6000_ABI_NAME): Define.
2662 * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
2663 GNU_USER_TARGET_OS_CPP_BUILTINS.
2664 (MUSL_DYNAMIC_LINKER32): Undef before defining.
2665 (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
2666 (CHOOSE_DYNAMIC_LINKER): Don't define.
2667 (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
2668 * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
2669 (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
2670 not already defined.
2671 (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
2672 TARGET_OS_CPP_BUILTINS.
2674 2018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2676 * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
2678 2018-11-26 Jakub Jelinek <jakub@redhat.com>
2681 * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
2682 and !TARGET_64BIT, return NULL_TREE.
2685 * dwarf2out.c (secname_for_decl): For functions with
2686 DECL_SECTION_NAME if in_cold_section_p, try to return
2687 current_function_section's name if it is a named section.
2689 2018-11-26 Maya Rashish <coypu@sdf.org>
2692 * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
2693 * config/x-netbsd: New file.
2694 * config/host-netbsd: New file.
2696 2018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2698 * doc/sourcebuild.texi: Document check_effective_target_size20plus.
2699 Clarify documentation for check_effective_target_size32plus.
2701 2018-11-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2703 * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
2704 and modify expand_strncmp_vec_sequence.
2705 (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
2706 (generate_6432_conversion): New function.
2707 (expand_block_compare): Add support for vsx.
2708 (expand_block_compare_gpr): New function.
2709 * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
2710 default limit to 63 because of more compact vsx code.
2712 2018-11-26 Uros Bizjak <ubizjak@gmail.com>
2715 * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
2716 for registers for which no debug information can be generated.
2717 Use INVALID_REGNUM for non-existent registers.
2718 (dbx64_register_map): Ditto.
2719 (svr4_dbx_register_map): Ditto.
2721 2018-11-26 Sandra Loosemore <sandra@codesourcery.com>
2724 * doc/invoke.texi (Optimize Options): Document that
2725 -freorder-functions also uses "hot" and "cold" attributes.
2727 2018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2729 * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
2731 2018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2733 * config/arm/types.md (mul64): Rename to...
2734 (widen_mul64): ... This.
2735 * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
2736 * config/arm/cortex-a15.md: Likewise.
2737 * config/arm/cortex-a5.md: Likewise.
2738 * config/arm/cortex-a53.md: Likewise.
2739 * config/arm/cortex-a57.md: Likewise.
2740 * config/arm/cortex-a7.md: Likewise.
2741 * config/arm/cortex-m4.md: Likewise.
2742 * config/arm/exynos-m1.md: Likewise.
2743 * config/arm/marvell-pj4.md: Likewise.
2744 * config/arm/xgene1.md: Likewise.
2746 2018-11-26 Richard Biener <rguenther@suse.de>
2748 PR tree-optimization/88182
2749 * tree-vect-loop.c (vectorizable_reduction): Pick up single
2750 correct reduc_def_info.
2751 * tree-vect-slp.c (vect_analyze_slp_instance): Set
2752 STMT_VINFO_REDUC_DEF of the first stmt.
2754 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2756 * doc/invoke.texi: Document z14/arch12 -march option.
2758 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
2761 * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
2762 (Warning Options): Likewise.
2764 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
2767 * doc/extend.texi (Common Function Attributes): Clarify that
2768 functions with "const" attribute can read const global variables.
2770 2018-11-26 Alan Modra <amodra@gmail.com>
2772 * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
2774 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
2777 * doc/extend.texi (Common Variable Attributes): Use preferred
2778 placement of type attributes in examples, plus whitespace fixes.
2779 (Type Attributes): Clarify why placement of attributes
2780 immediately after struct/union/enum keyword is preferred.
2781 (Common Type Attributes): Use preferred placement of type
2782 attributes in examples, plus more whitespace fixes.
2784 2018-11-25 Paul Koning <ni1d@arrl.net>
2786 * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
2788 2018-11-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2790 * doc/sourcebuild.texi (Effective-Target Keywords, Environment
2791 attributes): Document d_runtime.
2793 2018-11-25 Vladimir Makarov <vmakarov@redhat.com>
2796 * ira-costs.c (record_operand_costs): Initiate register move cost
2799 2018-11-23 Jeff Law <law@redhat.com>
2801 PR rtl-optimization/87468
2802 * tree-ssa-threadupdate.c (create_block_for_threading): Clear
2803 EDGE_IGNORE on all outgoing edges of the duplicate block.
2805 2018-11-23 Vladimir Makarov <vmakarov@redhat.com>
2808 * ira-costs.c (record_operand_costs): Use bigger hard reg class if
2809 its mode does not fit to the original class.
2811 2018-11-23 Martin Sebor <msebor@redhat.com>
2813 PR tree-optimization/87756
2814 * expr.c (string_constant): Synthesize a string literal from
2815 the address of a constant character.
2816 * tree.c (build_string_literal): Add an argument.
2817 * tree.h (build_string_literal): Same.
2819 2018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com>
2821 * config/aarch64/aarch64-cores.def: Define emag.
2822 * config/aarch64/aarch64-tune.md: Regenerated with emag.
2823 * config/aarch64/aarch64.c (emag_tunings): New struct.
2824 * doc/invoke.texi: Document mtune value.
2826 2018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com>
2827 Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
2829 * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
2831 * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
2833 * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
2835 * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
2838 2018-11-23 Richard Biener <rguenther@suse.de>
2840 PR tree-optimization/88149
2841 * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
2842 the case where there are two different def types for the
2843 same operand at different operand position in the same stmt.
2845 2018-23-11 Mihail Ionescu <mihail.ionescu@arm.com>
2847 * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
2848 gen_atomic_compare_swap_1.
2849 (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
2850 (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
2851 (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
2852 (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
2853 (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
2855 * config/arm/arm.md (neon_vext<mode>),
2856 (neon_vrev64<mode>, neon_vrev32<mode>),
2857 (neon_vrev16<mode>, neon_vtrn<mode>_internal),
2858 (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
2859 before the pattern name.
2860 * config/arm/sync.md:
2861 (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
2862 (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
2864 2018-11-23 Jakub Jelinek <jakub@redhat.com>
2866 PR tree-optimization/86614
2867 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
2868 if TREE_NO_WARNING is set on ref.ref.
2870 2018-11-22 Sandra Loosemore <sandra@codesourcery.com>
2871 Alan Coopersmith <alan.coopersmith@oracle.com>
2874 * doc/extend.texi (Designated Inits): Clarify handling of multiple
2875 initializers for unions.
2877 2018-11-22 Jan Hubicka <jh@suse.cz>
2880 * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
2881 parameters; do not warn here.
2882 (odr_subtypes_equivalent_p): Likewise.
2883 (warn_odr): Fix typo.
2884 (warn_types_mismatch): Do not output confused warnings on integer types.
2885 (odr_types_equivalent_p): Update.
2887 2018-11-22 Jakub Jelinek <jakub@redhat.com>
2889 * config/i386/i386.c (ix86_option_override_internal,
2890 ix86_can_inline_p, classify_argument, construct_container,
2891 ix86_expand_prologue, ix86_expand_split_stack_prologue,
2892 ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
2893 expand_setmem_epilogue_via_loop, promote_duplicated_reg,
2894 ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
2897 * config/i386/i386.c (ix86_option_override_internal): For
2898 stack_protector_guard related options, use opts_set->x_ instead
2899 of global_options_set. and prefix options with opts->x_ . Move
2900 defaults for offset and reg into else block.
2902 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
2904 * config/i386/i386.c (ix86_check_avx_upper_register):
2905 Return true for all SSE registers with mode bitsize > 128.
2907 2018-11-22 Vladimir Makarov <vmakarov@redhat.com>
2909 PR rtl-optimization/87718
2910 * ira-costs.c: Remove trailing white-spaces.
2911 (record_operand_costs): Add a special treatment for moves
2912 involving a hard register.
2914 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
2916 * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
2917 (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
2919 2018-11-22 Martin Liska <mliska@suse.cz>
2921 * common/config/i386/i386-common.c (processor_names): Add
2922 static assert and add missing "znver2".
2923 (ix86_get_valid_option_values): Add checking assert for null
2924 values and add "native" value if feasible.
2925 * config/i386/i386.h: Do not declare size of processor_names.
2926 * common/config/i386/i386-common.c:
2927 * config/i386/i386.c: Add static assert for size
2928 of processor_cost_table.
2930 2018-11-22 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2932 * target-insns.def (stack_protect_combined_set): Define new standard
2934 (stack_protect_combined_test): Likewise.
2935 * cfgexpand.c (stack_protect_prologue): Try new
2936 stack_protect_combined_set pattern first.
2937 * function.c (stack_protect_epilogue): Try new
2938 stack_protect_combined_test pattern first.
2939 * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
2940 parameters to control which register to use as PIC register and force
2941 reloading PIC register respectively. Insert in the stream of insns if
2943 (legitimize_pic_address): Expose above new parameters in prototype and
2944 adapt recursive calls accordingly. Use pic_reg if non null instead of
2946 (arm_load_pic_register): Add pic_reg parameter and use it if non null.
2947 (arm_legitimize_address): Adapt to new legitimize_pic_address
2949 (thumb_legitimize_address): Likewise.
2950 (arm_emit_call_insn): Adapt to require_pic_register prototype change.
2951 (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
2952 (thumb1_expand_prologue): Likewise.
2953 * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
2955 (arm_load_pic_register): Likewise.
2956 * config/arm/predicated.md (guard_addr_operand): New predicate.
2957 (guard_operand): New predicate.
2958 * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
2960 (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
2962 (stack_protect_combined_set): New expander..
2963 (stack_protect_combined_set_insn): New insn_and_split pattern.
2964 (stack_protect_set_insn): New insn pattern.
2965 (stack_protect_combined_test): New expander.
2966 (stack_protect_combined_test_insn): New insn_and_split pattern.
2967 (arm_stack_protect_test_insn): New insn pattern.
2968 * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
2969 * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
2970 (UNSPEC_SP_TEST): Likewise.
2971 * doc/md.texi (stack_protect_combined_set): Document new standard
2973 (stack_protect_set): Clarify that the operand for guard's address is
2975 (stack_protect_combined_test): Document new standard pattern name.
2976 (stack_protect_test): Clarify that the operand for guard's address is
2979 2018-11-22 Richard Biener <rguenther@suse.de>
2981 PR tree-optimization/88148
2982 * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
2983 valueization callback parameter and handle it.
2984 * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
2985 * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
2987 2018-11-22 Richard Biener <rguenther@suse.de>
2991 * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
2992 which can appear in size expressions.
2993 * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
2994 CALL_EXPR_BY_DESCRIPTOR.
2995 (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
2996 * tree-streamer-out.c (pack_ts_base_value_fields): Stream
2997 CALL_EXPR_BY_DESCRIPTOR.
2998 (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
3002 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
3005 2018-11-22 Richard Biener <rguenther@suse.de>
3007 PR tree-optimization/88069
3008 * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
3009 apply to default defs.
3011 2018-11-22 Jakub Jelinek <jakub@redhat.com>
3015 * config/i386/i386.c (ix86_option_override_internal): Default
3016 ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
3018 * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
3019 stack_protect_test, stack_protect_test_<mode>): Use empty condition
3020 instead of TARGET_SSP_TLS_GUARD.
3022 2018-11-22 Martin Liska <mliska@suse.cz>
3025 * config/i386/i386.c (ix86_option_override_internal): Error
3026 about usage -mabi=ms and -fsanitize=thread.
3028 2018-11-21 Uros Bizjak <ubizjak@gmail.com>
3031 2013-10-26 Vladimir Makarov <vmakarov@redhat.com>
3034 2013-10-25 Vladimir Makarov <vmakarov@redhat.com>
3036 * lra-spills.c (lra_final_code_change): Remove useless move insns.
3038 2018-11-21 Uros Bizjak <ubizjak@gmail.com>
3041 * function.c (expand_function_end): Do not emit extra blockage insn.
3043 2018-11-21 Lokesh Janghel <lokeshjanghel91@gmail.com>
3046 * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
3047 of FIRST_SSE_REG for 4 or 8 byte modes.
3049 2018-11-21 Jan Hubicka <jh@suse.cz>
3052 * tree.c (fld_decl_context): Break out from ...
3053 (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
3054 DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
3055 (fld_incomplete_type_of): Build copy of TYP_DECL.
3056 * ipa-devirt.c (free_enum_values): Rename to ...
3057 (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
3058 and TREE_TYPEs of TYPE_DECLs.
3059 (get_odr_type): Initialize odr_vtable_hash if needed.
3061 2018-11-21 Alexandre Oliva <oliva@adacore.com>
3063 * final.c (compute_discriminator): Declare. Renamed from...
3064 (maybe_set_discriminator): ... this. Set and return a local.
3065 (override_discriminator): New.
3066 (final_scan_insn_1): Set it.
3067 (notice_source_line): Adjust. Always set discriminator.
3069 2018-11-21 Jakub Jelinek <jakub@redhat.com>
3072 * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
3073 rIJ constraint for aarch64_plus_operand rather than rn.
3075 2018-11-21 Renlin Li <renlin.li@arm.com>
3078 * explow.h (get_dynamic_stack_size): Declare it as external.
3079 * explow.c (record_new_stack_level): Remove function static attribute.
3080 * function.c (assign_stack_local_1): Dynamically align the stack slot
3081 addr for parameter copy on the stack.
3083 2018-11-21 Richard Biener <rguenther@suse.de>
3086 * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
3087 * Makefile.in (bitmap.o-warn): Remove again.
3089 2018-11-20 Jeff Law <law@redhat.com>
3091 PR tree-optimization/88069
3092 * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
3093 degenerate virtual PHIs.
3095 2018-11-21 H.J. Lu <hongjiu.lu@intel.com>
3098 * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
3099 nonimmediate_operand with register_operand.
3100 (avx2_<code>v8qiv8si2<mask_name>): Likewise.
3101 (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
3102 (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
3103 (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
3104 (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
3105 (avx2_<code>v4qiv4di2<mask_name>): Likewise.
3106 (avx2_<code>v4hiv4di2<mask_name>): Likewise.
3107 (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
3108 (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
3109 (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
3110 (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
3111 (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
3112 (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
3113 (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
3114 (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
3115 (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
3116 (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
3117 (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
3118 (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
3119 (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
3120 (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
3121 (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
3122 (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
3123 (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
3124 (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
3125 (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
3126 (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
3128 2018-11-21 H.J. Lu <hongjiu.lu@intel.com>
3130 * read-rtl.c (apply_subst_iterator): Handle define_split and
3131 define_insn_and_split.
3133 2018-11-21 Jakub Jelinek <jakub@redhat.com>
3135 PR rtl-optimization/87817
3136 * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
3137 *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
3138 in the pattern to avoid triggering UB when operands[2] is zero.
3139 (tbm_bextri_<mode>): New expander. Renamed the old define_insn to ...
3140 (*tbm_bextri_<mode>): ... this.
3142 2018-11-21 Tom de Vries <tdevries@suse.de>
3145 * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
3148 2018-11-21 Andreas Krebbel <krebbel@linux.ibm.com>
3150 * configure.ac: Add check for Binutils to determine whether vector
3151 load/store alignments hints are being supported.
3152 * config.in: Regenerate.
3153 * configure: Regenerate.
3154 * config/s390/s390.c (print_operand): Support new output
3156 * config/s390/s390.md ("movti"): Append alignment hint output
3157 using the new output modifier 'A'.
3158 * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
3159 ("*vec_ti_to_v1ti"): Likewise.
3161 2018-11-20 Martin Sebor <msebor@redhat.com>
3163 * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
3165 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3168 * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
3169 compare ENUM values.
3171 2018-11-20 Jakub Jelinek <jakub@redhat.com>
3173 PR tree-optimization/87895
3174 * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
3175 or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
3176 (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
3177 redirects edges to EXIT to edges to incr_bb, iterate while EXIT
3178 has any preds and always use EDGE_PRED (, 0).
3180 2018-11-20 Uros Bizjak <ubizjak@gmail.com>
3183 * mode-switching.c (create_pre_exit): After reload, always split the
3184 fallthrough edge to the exit block.
3186 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3188 * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
3189 * tree.c (build_array_type_1): Forward declare.
3190 (fld_type_variant_equal_p): Add INNER_TYPE parameter.
3191 (fld_type_variant): Likewise.
3192 (fld_simplified_types): New hash.
3193 (fld_process_array_type): New function.
3194 (fld_incomplete_type_of): Handle array and enumeration types.
3195 (fld_simplified_type): Handle simplification of arrays.
3196 (free_lang_data): Allocate and free simplified types hash.
3198 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3201 * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
3202 when sybtype already violates ODR.
3203 (get_odr_type): Do not ICE when insert is false and type duplicate
3204 is not registered yet.
3205 (register_odr_type): Be sure to register subtypes first.
3207 2018-11-20 Andreas Krebbel <krebbel@linux.ibm.com>
3209 * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
3210 DImode parts of the target operand.
3212 2018-11-20 Nathan Sidwell <nathan@acm.org>
3215 * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
3217 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3220 * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
3222 (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
3224 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3227 * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
3228 * ipa.c (possible_inline_candidate_p): Break out from ..
3229 (process_references): ... here ; drop before_inlining_p;
3230 cleanup handling of alises.
3231 (walk_polymorphic_call_targets): Likewise.
3232 (symbol_table::remove_unreachable_nodes): Likewise.
3233 * passes.c (pass_data_ipa_remove_symbols): New structure.
3234 (pass_ipa_remove_symbols): New pass.
3235 (make_pass_ipa_remove_symbols): New function.
3236 * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
3237 * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
3239 2018-11-20 Richard Biener <rguenther@suse.de>
3241 * tree-vect-stmts.c (vectorizable_condition): Do not get
3242 at else_clause vect def for EXTRACT_LAST_REDUCTION. Remove
3243 pointless vect_is_simple_use calls.
3245 2018-11-20 Richard Biener <rguenther@suse.de>
3247 PR tree-optimization/88074
3248 * tree-ssa-dom.c (pass_dominator::execute): Do not walk
3251 2018-11-20 Richard Biener <rguenther@suse.de>
3253 PR tree-optimization/88069
3254 * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
3255 virtual PHI arguments.
3257 2018-11-20 Ilya Leoshkevich <iii@linux.ibm.com>
3260 * config/s390/s390.md: Skip LT(G) peephole when literal pool is
3262 * rtl.h (contains_constant_pool_address_p): New function.
3263 * rtlanal.c (contains_constant_pool_address_p): Likewise.
3265 2018-11-20 Richard Biener <rguenther@suse.de>
3268 * alias.c (component_uses_parent_alias_set_from): Remove
3269 alias-set zero and TYPE_TYPELESS_STORAGE case both already
3270 handled in other ways.
3272 2018-11-20 Richard Biener <rguenther@suse.de>
3274 PR tree-optimization/88087
3275 * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
3277 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
3280 2018-11-20 Richard Biener <rguenther@suse.de>
3283 * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
3285 2018-11-19 Jan Hubicka <hubicka@ucw.cz>
3288 * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
3290 2018-11-19 Sandra Loosemore <sandra@codesourcery.com>
3293 * doc/invoke.texi (Link Options): Mention shared libraries
3294 in documentation for the -l option. Simplify discussion and
3295 point to the system linker documentation for details.
3297 2018-11-19 Peter Bergner <bergner@linux.ibm.com>
3299 PR rtl-optimization/88033
3300 * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
3301 to itself. Use HARD_REGISTER_NUM_P.
3303 2018-11-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3305 * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
3306 of the loop if vf is non-constant.
3308 2018-11-19 David Malcolm <dmalcolm@redhat.com>
3310 PR tree-optimization/87025
3311 * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
3312 immediately after creating the scope optinfo.
3313 (selftest::test_pr87025): New function.
3314 (selftest::dumpfile_c_tests): Call it.
3315 * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
3316 that we're not popping the top-level records array.
3317 * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
3318 not changing the kind of a "scope" optinfo.
3320 2018-11-19 David Malcolm <dmalcolm@redhat.com>
3322 PR tree-optimization/87025
3323 * dump-context.h: Include "optinfo.h".
3324 (class optrecord_json_writer): New forward decl.
3325 (dump_context::forcibly_enable_optinfo_p): Delete.
3326 (dump_context::optinfo_enabled_p): New member function.
3327 (dump_context::optimization_records_enabled_p): New member
3329 (dump_context::set_json_writer): New member function.
3330 (dump_context::emit_optinfo): New member function.
3331 (dump_context::m_forcibly_enable_optinfo): Delete.
3332 (dump_context::m_json_writer): New member data.
3333 * dumpfile.c (dump_context::set_json_writer): New member function.
3334 (dump_context::finish_any_json_writer): New member function.
3335 (dump_context::end_scope): Replace call to
3336 optimization_records_maybe_pop_dump_scope with call to
3337 m_json_writer->pop_scope.
3338 (dump_context::optinfo_enabled_p): New member function.
3339 (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
3340 to dump_context::emit_optinfo.
3341 (dump_context::emit_optinfo): New member function.
3342 (temp_dump_context::temp_dump_context): Replace
3343 m_forcibly_enable_optinfo with call to set_json_writer.
3344 (temp_dump_context::~temp_dump_context): Clean up any json writer.
3345 * optinfo-emit-json.cc (class optrecord_json_writer): Move to
3347 (the_json_writer): Delete.
3348 (optimization_records_start): Delete.
3349 (optimization_records_finish): Delete.
3350 (optimization_records_enabled_p): Delete, in favor of
3351 dump_context::optimization_records_enabled_p.
3352 (optimization_records_maybe_record_optinfo): Delete.
3353 (optimization_records_maybe_pop_dump_scope): Delete.
3354 * optinfo-emit-json.h: Include "json.h". Delete forward
3356 (optimization_records_start): Delete.
3357 (optimization_records_finish): Delete.
3358 (optimization_records_enabled_p): Delete.
3359 (optimization_records_maybe_record_optinfo): Delete.
3360 (optimization_records_maybe_pop_dump_scope): Delete.
3361 (class optrecord_json_writer): Move here from
3362 optinfo-emit-json.cc.
3363 * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
3364 to optinfo::emit with call to dump_context::emit_optinfo.
3365 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
3366 (optinfo_enabled_p): Delete, in favor of
3367 dump_context::optinfo_enabled_p.
3368 (optinfo_wants_inlining_info_p): Update for conversion o
3369 optimization_records_enabled_p to a member function of
3371 * optinfo.h (optinfo_enabled_p): Delete, in favor of
3372 dump_context::optinfo_enabled_p.
3373 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
3374 * toplev.c: Include "dump-context.h".
3375 (compile_file): Replace call to optimization_records_finish with
3376 dump_context::finish_any_json_writer.
3377 (do_compile): Replace call to optimization_records_start with
3378 conditionally creating a optrecord_json_writer for the
3381 2018-11-19 Jakub Jelinek <jakub@redhat.com>
3383 PR tree-optimization/88071
3384 * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
3385 of false as last argument to gsi_remove.
3386 * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
3387 false as last argument to gsi_replace.
3390 * omp-expand.c: Don't include debug.h.
3391 (adjust_context_and_scope): Add REGION argument. Find DECL_CONTEXT
3392 from innermost outer parallel, task, teams or target that has a
3393 child_fn set, or, if there is no such outer region, use
3394 current_function_decl. Do the DECL_CONTEXT adjustment regardless of
3395 whether a suitable BLOCK is found or not.
3396 (expand_parallel_call, expand_teams_call): Don't call
3397 adjust_context_and_scope here.
3398 (grid_expand_target_grid_body): Revert 2017-01-25 changes.
3399 (expand_omp_taskreg, expand_omp_target): Likewise. Call
3400 adjust_context_and_scope.
3401 * dwarf2out.c (dwarf2out_early_global_decl): For
3402 decl_function_context recurse instead of calling dwarf2out_decl.
3404 2018-11-20 Eric Botcazou <ebotcazou@adacore.com>
3406 PR rtl-optimization/85925
3407 * rtl.h (word_register_operation_p): New predicate.
3408 * combine.c (record_dead_and_set_regs_1): Only apply specific handling
3409 for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
3410 * rtlanal.c (nonzero_bits1): Likewise. Adjust couple of comments.
3411 (num_sign_bit_copies1): Likewise.
3413 2018-11-19 Richard Biener <rguenther@suse.de>
3416 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
3419 2018-11-19 Eric Botcazou <ebotcazou@adacore.com>
3421 * compare-elim.c (struct comparison): Add not_in_a field.
3422 (is_not): New static function.
3423 (strip_not): Likewise.
3424 (conforming_compare): Handle a NOT in the first operand.
3425 (can_eliminate_compare): Likewise.
3426 (find_comparison_dom_walker::before_dom_children): Likewise.
3427 (try_eliminate_compare): Likewise.
3428 * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
3429 (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
3431 2018-11-19 Jonathan Wakely <jwakely@redhat.com>
3433 * doc/extend.texi (Common Type Attributes): Fix typos.
3435 2018-11-19 Martin Liska <mliska@suse.cz>
3437 PR gcov-profile/88045
3438 * coverage.c (coverage_begin_function): Add assert.
3440 2018-11-18 Sandra Loosemore <sandra@codesourcery.com>
3443 * doc/extend.texi (Common Function Attributes): Document that
3444 no_instrument_function applies to -p and -pg, too.
3445 * doc/invoke.texi (Instrumentation Options): Add cross-references
3446 to docs for -p, -pg, and -finstrument-functions.
3448 2018-11-18 Alan Modra <amodra@gmail.com>
3450 * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
3451 (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
3452 (GNU_USER_TARGET_CRTN): Define.
3453 (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
3454 (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
3456 2018-11-18 Alan Modra <amodra@gmail.com>
3458 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
3459 !HAVE_LD_PIE variant.
3460 (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
3462 2018-11-17 Nick Clifton <nickc@redhat.com>
3463 Sandra Loosemore <sandra@codesourcery.com>
3466 * doc/invoke.texi (Overall Options): Document that --help and
3467 --help=<value> options cannot be combined.
3469 2018-11-17 Jim Wilson <jimw@sifive.com>
3471 * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
3472 (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
3473 epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
3474 (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset. Set it
3475 to step2 if frame_pointer_needed and step1 is 0.
3477 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
3480 * doc/cpp.texi (System Headers): Add note about implicit
3481 extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
3483 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
3486 * doc/invoke.texi (C++ Dialect Options): Clarify usage of
3487 -fno-implicit-templates.
3489 2018-11-17 Jakub Jelinek <jakub@redhat.com>
3491 PR tree-optimization/87546
3492 * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
3493 min_precision variable, initially set it to orig_precision, only does
3494 something if op_type's precision is <= min_precision and update
3495 min_precision whenever calling set_op.
3497 2018-11-16 Jan Hubicka <hubicka@ucw.cz>
3500 * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
3502 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
3504 * doc/invoke.texi (Option Summary): Fix whitespace and line
3505 breaks in @gccoptlist environments.
3506 (Warning Options): Likewise.
3507 (Optimize Options): Likewise.
3508 (PowerPC SPE Options): Likewise.
3509 (RS/6000 and PowerPC Options): Likewise.
3511 2018-11-16 Jeff Law <law@redhat.com>
3513 * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
3516 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
3519 * doc/invoke.texi (Optimize Options): Update options enabled by
3520 fprofile-generate, -fprofile-use, and -fauto-profile.
3522 2018-11-16 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3525 * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
3526 Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
3527 * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
3529 * target.h: Add psi, pdi, pti to struct asm_int_op definition.
3530 * targhooks.c (default_print_patchable_function_entry): Assert
3531 asm_int_op does not return a NULL string.
3532 * varasm.c (integer_asm_op): Return the op for a partial int type
3533 when the requested size does not correspond to an integer type.
3534 * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
3535 * doc/tm.texi: Regenerate.
3537 2018-11-12 Jason Merrill <jason@redhat.com>
3539 * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
3541 2018-11-16 Michael Meissner <meissner@linux.ibm.com>
3543 * config/rs6000/constraints.md (wF constraint): Remove power9
3544 fusion documentation. Just document wF's use for power8 fusion.
3545 * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
3546 (fusion_gpr_addis): Delete power9 fusion support. Change power8
3547 fusion support to require the upper 12 bits to be all 0's or all
3549 (fusion_gpr_mem_load): Add comment.
3550 (fusion_addis_mem_combo_load): Remove power9 fusion support.
3551 (fusion_addis_mem_combo_store): Delete.
3552 (fusion_offsettable_mem_operand): Delete.
3553 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
3555 (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
3556 (POWERPC_MASKS): Delete power9 fusion.
3557 * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
3558 (fusion_p9_p): Delete.
3559 (expand_fusion_p9_load): Delete.
3560 (expand_fusion_p9_store): Delete.
3561 (emit_fusion_p9_load): Delete.
3562 (emit_fusion_p9_store): Delete.
3563 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
3565 (rs6000_option_override_internal): Set power8 fusion based on
3566 whether we are tuning for power8. Delete power9 fusion support.
3567 (rs6000_opt_masks): Delete -mpower9-fusion switch.
3568 (emit_fusion_load): Rename emit_fusion_load_store to
3569 emit_fusion_load, and drop fusion store support. Update callers.
3570 (emit_fusion_load_store): Likewise.
3571 (emit_fusion_gpr_load): Likewise.
3572 (fusion_p9_p): Delete.
3573 (expand_fusion_p9_load): Delete.
3574 (expand_fusion_p9_store): Delete.
3575 (emit_fusion_p9_load): Delete.
3576 (emit_fusion_p9_store): Delete.
3577 * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
3578 (GPR_FUSION): Delete.
3579 (FPR_FUSION): Delete.
3580 (power9 fusion peephole2s): Delete.
3581 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
3582 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
3583 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
3584 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
3585 (fusion_p9_<mode>_constant): Delete.
3586 * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
3587 power9 fusion switch.
3588 * doc/md.texi (PowerPC constraints): Update wF constraint
3589 documentation for power8 fusion only.
3591 2018-11-16 Jakub Jelinek <jakub@redhat.com>
3593 PR rtl-optimization/87475
3594 * cfgrtl.c (patch_jump_insn): Allow redirection failure for
3595 CROSSING_JUMP_P insns.
3596 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
3598 2018-11-16 Uros Bizjak <ubizjak@gmail.com>
3601 * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
3602 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
3603 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
3605 2018-11-16 Jakub Jelinek <jakub@redhat.com>
3608 * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
3609 is VOIDmode, in which case use int_mode. Similarly for op1_mode.
3611 2018-11-16 Eric Botcazou <ebotcazou@adacore.com>
3613 * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
3614 location_t parameter.
3615 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3616 (switch_decision_tree::do_jump_if_equal): Likewise.
3617 * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
3618 of switch statement to emit_case_nodes.
3619 (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
3620 set it on the newly built GIMPLE comparison statement.
3621 (switch_decision_tree::do_jump_if_equal): Likewise.
3622 (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
3623 in calls to do_jump_if_equal as well as recursive calls.
3625 2018-11-16 Andi Kleen <ak@linux.intel.com>
3627 * config/i386/i386.md: Allow memory operands to ptwrite.
3629 2018-11-16 Michael Matz <matz@suse.de>
3631 * system.h (PRsa): New macro.
3632 (SIZE_AMOUNT): Cast number to uint64_t.
3633 * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
3634 (pool_usage::dump_footer): Likewise and also use PRIu64.
3635 * bitmap.h (bitmap_usage::dump): Likewise.
3636 * ggc-common.c (ggc_usage::dump): Likewise.
3637 * ggc-page.c (ggc_print_statistics): Likewise.
3638 * input.c (dump_line_table_statistics): Likewise.
3639 * mem-stats.h (mem_usage::dump): Likewise.
3640 (mem_usage::dump_footer): Likewise.
3641 * rtl.c (dump_rtx_statistics): Likewise.
3642 * tree-cfg.c (dump_cfg_stats): Likewise.
3643 * tree-dfa.c (dump_dfa_stats): Likewise.
3644 * tree-phinodes.c (phinodes_print_statistics): Likewise.
3645 * tree-ssanames (ssanames_print_statistics): Likewise.
3646 * vec.c (vec_usage::dump): Likewise.
3647 (vec_usage::dump_footer): Likewise.
3649 2018-11-16 Richard Biener <rguenther@suse.de>
3651 PR tree-optimization/88011
3652 * tree-vrp.c (extract_range_from_binary_expr): Fix error in
3653 replacing set_value_range_to_undefined and
3654 set_value_range_to_varying with method calls.
3656 2018-11-16 Ilya Leoshkevich <iii@linux.ibm.com>
3658 * config/s390/s390.md
3659 (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
3661 (*r<noxa>sbg_<mode>_sll): Likewise.
3662 (*r<noxa>sbg_<mode>_srl): Likewise.
3663 (*r<noxa>sbg_sidi_srl): New pattern.
3665 2018-11-16 Jerome Lambourg <lambourg@adacore.com>
3667 * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
3668 Default to TARGET_VXWORKS_RTP.
3669 (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3670 of TARGET_VXWORKS_RTP.
3671 * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3672 of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
3674 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
3677 * doc/extend.texi (Common Type Attributes): Make it explicit
3678 that attribute "packed" can apply to C++ classes.
3680 2018-11-15 Martin Sebor <msebor@redhat.com>
3684 * tree.c (type_argument_type): New function.
3685 * tree.h (type_argument_type): Declare it.
3686 * gcc/doc/extend.texi (alloc_align): Update and clarify.
3687 (alloc_size, nonnull, sentinel): Same.
3689 2018-11-15 Andrew Stubbs <ams@codesourcery.com>
3690 Kwok Cheung Yeung <kcy@codesourcery.com>
3692 * tree-vect-stmts.c (vectorizable_store): Don't ICE when
3693 int_mode_for_size fails.
3694 (vectorizable_load): Likewise.
3696 2018-11-15 David Malcolm <dmalcolm@redhat.com>
3698 * doc/ux.texi (Group logically-related diagnostics): Move
3699 discussion of auto_diagnostic_group into this new subsection.
3700 Give an example of where this grouping is used.
3702 2018-11-15 David Malcolm <dmalcolm@redhat.com>
3705 * Makefile.in (OBJS): Move json.o to...
3706 (OBJS-libcommon): ...here and add diagnostic-format-json.o.
3707 * common.opt (fdiagnostics-format=): New option.
3708 (diagnostics_output_format): New enum.
3709 * diagnostic-format-json.cc: New file.
3710 * diagnostic.c (default_diagnostic_final_cb): New function, taken
3711 from start of diagnostic_finish.
3712 (diagnostic_initialize): Initialize final_cb to
3713 default_diagnostic_final_cb.
3714 (diagnostic_finish): Move "being treated as errors" messages to
3715 default_diagnostic_final_cb. Call any final_cb.
3716 (default_diagnostic_finalizer): Add diagnostic_t param.
3717 (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
3718 diagnostic_finalizer callback.
3719 * diagnostic.h (enum diagnostics_output_format): New enum.
3720 (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
3721 (struct diagnostic_context): Add "final_cb".
3722 (default_diagnostic_finalizer): Add diagnostic_t param.
3723 (diagnostic_output_format_init): New decl.
3724 * doc/invoke.texi (-fdiagnostics-format): New option.
3725 * dwarf2out.c (gen_producer_string): Ignore
3726 OPT_fdiagnostics_format_.
3727 * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
3728 * lto-wrapper.c (append_diag_options): Ignore it.
3729 * opts.c (common_handle_option): Handle it.
3731 2018-11-15 David Malcolm <dmalcolm@redhat.com>
3733 PR tree-optimization/88015
3734 * graphite-isl-ast-to-gimple.c
3735 (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
3737 * graphite-sese-to-poly.c (build_poly_scop): Likewise.
3739 2018-11-15 Richard Biener <rguenther@suse.de>
3742 * gimple.c (gimple_call_flags): Union flags from decl, type
3744 * trans-mem.c (is_tm_pure_call): Simplify.
3746 2018-11-15 Richard Biener <rguenther@suse.de>
3748 PR tree-optimization/88031
3749 * tree-vect-loop.c (vectorizable_reduction): Move check
3750 for multiple types earlier so we get the expected dump.
3751 Simplify calls to vectorizable_condition.
3752 * tree-vect-stmts.h (vectorizable_condition): Update prototype.
3753 * tree-vect-stmts.c (vectorizable_condition): Instead of
3754 reduc_def and reduc_index take just a flag. Simplify
3755 code-generation now that we can rely on the defs being set up.
3756 (vectorizable_comparison): Remove unused argument.
3758 2018-11-15 Richard Biener <rguenther@suse.de>
3760 PR tree-optimization/88030
3761 * tree-complex.c (need_eh_cleanup): New global.
3762 (update_complex_assignment): Mark blocks that need EH update.
3763 (expand_complex_comparison): Likewise.
3764 (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
3765 perform EH cleanup and schedule CFG cleanup if that did anything.
3767 2018-11-15 Jakub Jelinek <jakub@redhat.com>
3769 PR rtl-optimization/88018
3770 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
3771 on the presence of fallthru edge, rather than if it is a USE or not.
3773 2018-11-15 Richard Biener <rguenther@suse.de>
3776 * tree-data-ref.c (analyze_miv_subscript): Guard calls to
3777 analyze_subscript_affine_affine properly.
3779 2018-11-15 Xianmiao Qu <xianmiao_qu@c-sky.com>
3781 * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
3782 (GLIBC_DYNAMIC_LINKER): Define.
3783 (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
3785 2018-11-15 Bin Cheng <bin.cheng@linux.alibaba.com>
3787 PR tree-optimization/84648
3788 * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
3789 (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
3790 by calling adjust_cond_for_loop_until_wrap.
3792 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
3795 * doc/extend.texi (Common Function Attributes): Clarify linker
3796 restrictions on "aligned" attribute.
3797 (Common Variable Attributes): Likewise. Mention that linker
3798 restrictions don't apply to stack-allocated variables.
3800 2018-11-15 Joern Rennecke <joern.rennecke@riscy-ip.com>
3802 * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
3805 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
3809 * doc/invoke.texi (Optimize Options): Remove bad example about
3810 interaction between -flto and -O. Replace it with a note that
3811 you need to compile with -O and not just link. Copy-edit -flto
3812 discussion to reduce verbiage and improve flow.
3814 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
3817 * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
3818 also suppress many optimizations. Alphabetize option lists for
3819 -O1, -O2, and -Os. Add list of options disabled with -Og, and
3820 correct documentation for those options to say that.
3821 * opts.c (default_options_table): Sort table by level and option
3822 name, to make it easier to correlate to the manual.
3824 2018-11-14 Uros Bizjak <ubizjak@gmail.com>
3826 * config/i386/i386.c (ix86_print_operand_address_as): Simplify
3827 printing of the names of segment registers.
3829 2018-11-14 Nathan Sidwell <nathan@acm.org>
3833 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
3836 2018-11-14 David Malcolm <dmalcolm@redhat.com>
3838 * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
3840 2018-11-14 Jakub Jelinek <jakub@redhat.com>
3842 P1236R1 - Signed integers are two's complement
3843 * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
3846 * hash-map.h (hash_map::iterator::reference_pair): New class.
3847 (hash_map::iterator::operator*): Return it rather than std::pair.
3849 2018-11-14 Jeff Law <law@redhat.com>
3851 * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
3852 iff the operand is a constant.
3854 2018-11-14 Aldy Hernandez <aldyh@redhat.com>
3856 * gimple-ssa-evrp-analyze.c
3857 (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
3858 ignore_equivs_equal_p to equal_p.
3859 * ipa-cp.c (meet_with_1): Use equal_p instead of
3860 ignore_equivs_equal_p.
3861 * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
3862 * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
3863 (value_range::operator==): Remove.
3864 (value_range::operator!=): Remove.
3865 (vrp_prop::visit_stmt): Use equal_p.
3866 * tree-vrp.h (value_range): Remove operator==, operator!=,
3867 ignore_equivs_equal_p.
3868 * vr-values.c (update_value_range): Use equal_p.
3870 2018-11-14 Michael Matz <matz@suse.de>
3873 * gimplify.c (collect_fallthrough_labels): Add new argument,
3874 return location via that, don't modify statements.
3875 (warn_implicit_fallthrough_r): Adjust call, don't use
3876 statement location directly.
3878 2018-11-14 Richard Biener <rguenther@suse.de>
3881 * tree-data-ref.c (split_constant_offset): Add wrapper
3882 allocating a cache hash-map.
3883 (split_constant_offset_1): Cache results of expanding
3884 expressions from SSA def stmts.
3886 2018-11-14 Richard Biener <rguenther@suse.de>
3889 * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
3891 (lambda_vector_mult_const): Likewise.
3892 (lambda_matrix_right_hermite): Use lambda_int temporaries.
3894 2018-11-14 Wilco Dijkstra <wdijkstr@arm.com>
3895 Jackson Woodruff <jackson.woodruff@arm.com>
3897 PR 71026/tree-optimization
3898 * match.pd: Simplify floating point comparisons.
3900 2018-11-14 Jakub Jelinek <jakub@redhat.com>
3902 PR rtl-optimization/87817
3903 * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
3904 last argument has low 8 bits clear, fold to 0.
3906 2018-11-14 Iain Sandoe <iain@sandoe.co.uk>
3908 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
3909 MAX_OFILE_ALIGNMENT as needed.
3911 2018-11-14 Richard Biener <rguenther@suse.de>
3913 PR tree-optimization/88019
3914 * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
3915 COND_EXPR arguments to non-trapping overflow.
3917 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
3919 * config/arc/arc.h (reg_class): Reorder registers classes, remove
3920 unused register classes.
3921 (REG_CLASS_NAMES): Likewise.
3922 (REG_CLASS_CONTENTS): Likewise.
3923 (FIXED_REGISTERS): Make lp_count fixed.
3924 (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
3925 (PROGRAM_COUNTER_REGNO): Remove.
3926 * config/arc/arc.c (arc_conditional_register_usage): Remove unused
3927 register classes, use constants for register numbers, remove
3928 reg_class_contents references.
3929 (arc_process_double_reg_moves): Add asserts.
3930 (arc_secondary_reload): Remove LPCOUNT_REG reference, use
3931 lra_in_progress predicate.
3932 (arc_init_reg_tables): Remove unused register classes.
3933 (arc_register_move_cost): Likewise.
3934 (arc_preferred_reload_class): Likewise.
3935 (hwloop_optimize): Update rtx patterns involving lp_count register.
3936 (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
3937 * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
3938 CHEAP_CORE_REGS. Former one will be used for LRA.
3939 ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS. Former
3940 one will be used for LRA.
3941 ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS. Former
3942 one will be used for LRA.
3943 ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
3944 Former one will be used for LRA.
3945 ("f"): Delete constraint.
3948 ("l"): Change it from register constraint to constraint.
3949 * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
3950 (movhi_insn): Likewise.
3951 (movsi_insn): Update pattern.
3954 (stack_tie): Remove 'b' constraint letter.
3956 (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
3957 (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
3958 (R57_REG, R59_REG, PCL_REG): Likewise.
3959 (ILINK1_REGNUM): Renamed to ILINK1_REG.
3960 (ILINK2_REGNUM): Renamed to ILINK2_REG.
3962 (SP_REGS): Likewise.
3963 (Rcw): Remove unused reg classes.
3964 * config/arc/predicates.md (dest_reg_operand): Just default on
3965 register_operand predicate.
3966 (mpy_dest_reg_operand): Likewise.
3967 (move_dest_operand): Use macros instead of constants.
3968 * config/arc/arc.opt (mlra): Switch to lra as default.
3970 2018-11-14 Richard Biener <rguenther@suse.de>
3972 PR tree-optimization/87974
3973 * tree-vect-loop.c (vectorizable_reduction): When computing
3974 the vectorized reduction PHI vector type ignore constant
3977 2018-11-14 Jakub Jelinek <jakub@redhat.com>
3979 PR tree-optimization/87977
3980 * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
3981 stmt, build a new one and replace the old one with it. Formatting fix.
3982 Call release_ssa_name (x) if !has_other_use and !delete_div.
3983 (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
3984 verify lhs of stmt is still def.
3986 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
3988 PR rtl-optimization/87507
3989 * lower-subreg.c (operand_for_swap_move_operator): New function.
3990 (simple_move): Strip simple operators.
3991 (find_pseudo_copy): Likewise.
3992 (resolve_operand_for_swap_move_operator): New function.
3993 (resolve_simple_move): Strip simple operators and swap operands.
3995 2018-11-13 Jakub Jelinek <jakub@redhat.com>
3997 PR tree-optimization/87898
3998 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
3999 (ipa_simd_modify_function_body): Remove debug stmts where the first
4000 argument was changed into a non-decl.
4002 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
4004 PR rtl-optimization/87899
4005 * lra-lives.c (start_living): Update white space in comment.
4006 (enum point_type): New.
4007 (sparseset_contains_pseudos_p): New function.
4008 (update_pseudo_point): Likewise.
4009 (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
4010 (make_hard_regno_dead): Likewise. Remove ignore_reg_for_conflicts
4011 handling. Move early exit after adding conflicts.
4012 (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro. Add early exit
4013 if regno is already live. Remove all handling of program points.
4014 (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro. Add early exit
4015 after adding conflicts. Remove all handling of program points and
4016 ignore_reg_for_conflicts.
4017 (mark_regno_live): Use HARD_REGISTER_NUM_P macro. Remove return value
4018 and do not guard call to mark_pseudo_live.
4019 (mark_regno_dead): Use HARD_REGISTER_NUM_P macro. Remove return value
4020 and do not guard call to mark_pseudo_dead.
4021 (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
4022 (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
4023 Use new function update_pseudo_point. Handle register copies by
4024 removing the source register from the live set. Handle INOUT operands.
4025 Update to the next program point using the unused_set, dead_set and
4027 (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
4029 2018-11-13 David Malcolm <dmalcolm@redhat.com>
4031 * builtins.c: Replace "source_location" with "location_t".
4032 * diagnostic-show-locus.c: Likewise.
4033 * diagnostic.c: Likewise.
4034 * dumpfile.c: Likewise.
4035 * gcc-rich-location.h: Likewise.
4036 * genmatch.c: Likewise.
4037 * gimple.h: Likewise.
4038 * gimplify.c: Likewise.
4039 * input.c: Likewise.
4040 * input.h: Likewise. Eliminate the typedef.
4041 * omp-expand.c: Likewise.
4042 * selftest.h: Likewise.
4043 * substring-locations.h (get_source_location_for_substring):
4045 (get_location_within_string): ...this.
4046 * tree-cfg.c: Replace "source_location" with "location_t".
4047 * tree-cfgcleanup.c: Likewise.
4048 * tree-diagnostic.c: Likewise.
4049 * tree-into-ssa.c: Likewise.
4050 * tree-outof-ssa.c: Likewise.
4051 * tree-parloops.c: Likewise.
4052 * tree-phinodes.c: Likewise.
4053 * tree-phinodes.h: Likewise.
4054 * tree-ssa-loop-ivopts.c: Likewise.
4055 * tree-ssa-loop-manip.c: Likewise.
4056 * tree-ssa-phiopt.c: Likewise.
4057 * tree-ssa-phiprop.c: Likewise.
4058 * tree-ssa-threadupdate.c: Likewise.
4059 * tree-ssa.c: Likewise.
4060 * tree-ssa.h: Likewise.
4061 * tree-vect-loop-manip.c: Likewise.
4063 2018-11-13 Michael Ploujnikov <michael.ploujnikov@oracle.com>
4065 * doc/extend.texi: Fix typo in the weakref description.
4067 2018-11-13 Richard Biener <rguenther@suse.de>
4069 PR tree-optimization/86991
4070 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
4071 group building until we have successfully detected the SLP
4073 (vect_is_simple_reduction): Remove fixup code here.
4075 2018-11-13 David Malcolm <dmalcolm@redhat.com>
4077 * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
4078 (dump_gimple_stmt): Use it.
4079 (dump_gimple_stmt_loc): Likewise.
4080 (dump_gimple_expr): Likewise.
4081 (dump_gimple_expr_loc): Likewise.
4082 (dump_generic_expr): Likewise.
4083 (dump_generic_expr_loc): Likewise.
4084 (dump_printf): Likewise.
4085 (dump_printf_loc): Likewise.
4086 (dump_dec): Likewise.
4087 (dump_dec): Likewise.
4088 (dump_hex): Likewise.
4089 (dump_symtab_node): Likewise.
4090 * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4091 Guard dump call with dump_enabled_p.
4092 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
4093 * graphite-optimize-isl.c (optimize_isl): Likewise.
4094 * graphite.c (graphite_transform_loops): Likewise.
4095 * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
4096 * tree-parloops.c (parallelize_loops): Likewise.
4097 * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
4098 * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
4099 (vect_prune_runtime_alias_test_list): Likewise.
4100 * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
4101 (vect_estimate_min_profitable_iters): Likewise.
4102 * tree-vect-slp.c (vect_record_max_nunits): Likewise.
4103 (vect_build_slp_tree_2): Likewise.
4104 (vect_supported_load_permutation_p): Likewise.
4105 (vect_slp_analyze_operations): Likewise.
4106 (vect_slp_analyze_bb_1): Likewise.
4107 (vect_slp_bb): Likewise.
4108 * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
4109 * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
4110 (pass_slp_vectorize::execute): Likewise.
4111 (increase_alignment): Likewise.
4113 2018-11-13 David Malcolm <dmalcolm@redhat.com>
4116 * ipa-inline.c (report_inline_failed_reason): Guard calls to
4117 cl_target_option_print_diff and cl_optimization_print_diff with
4120 2018-11-13 David Malcolm <dmalcolm@redhat.com>
4122 * doc/invoke.texi (-fsave-optimization-record): Note that the
4123 output is compressed.
4124 * optinfo-emit-json.cc: Include <zlib.h>.
4125 (optrecord_json_writer::write): Compress the output.
4127 2018-11-13 Aldy Hernandez <aldyh@redhat.com>
4129 * tree-vrp.c (value_range_base::dump): Dump type.
4130 Do not use INF nomenclature for 1-bit types.
4131 (dump_value_range): Group all variants to common dumping code.
4132 (debug): New overloaded functions for value_ranges.
4133 (value_range_base::dump): Remove no argument version.
4134 (value_range::dump): Same.
4136 2018-11-13 Richard Biener <rguenther@suse.de>
4138 PR tree-optimization/87931
4139 * tree-vect-loop.c (vect_is_simple_reduction): Restrict
4140 nested cycles we support to latch computations vectorizable_reduction
4143 2018-11-13 Martin Liska <mliska@suse.cz>
4145 PR tree-optimization/87885
4146 * cfghooks.c (account_profile_record): Rename
4148 (profile_record_check_consistency): ... this.
4149 Calculate missing num_mismatched_freq_in.
4150 (profile_record_account_profile): New function
4151 that calculates time and size of a function.
4152 * cfghooks.h (struct profile_record): Remove
4154 (struct cfg_hooks): Remove after_pass flag.
4155 (account_profile_record): Rename to ...
4156 (profile_record_check_consistency): ... this.
4157 (profile_record_account_profile): New.
4158 * cfgrtl.c (rtl_account_profile_record): Remove
4160 * passes.c (check_profile_consistency): Do only
4162 (account_profile): Calculate size and time of
4164 (pass_manager::dump_profile_report): Reformat
4166 (execute_one_ipa_transform_pass): Call
4167 consistency check before clean upand call account_profile
4168 after a clean up is done.
4169 (execute_one_pass): Call check_profile_consistency and
4170 account_profile instead of using after_pass flag..
4171 * tree-cfg.c (gimple_account_profile_record): Likewise.
4173 2018-11-13 Martin Liska <mliska@suse.cz>
4176 * config/i386/i386.c (ix86_option_override_internal): Error
4177 about usage -mabi=ms and -fsanitize={,kernel-}address.
4179 2018-11-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
4181 * config/aarch64/aarch64.c
4182 (aarch64_vectorize_preferred_vector_alignment): Change return type to
4184 (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
4185 alignment being a poly int.
4186 * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
4187 return type to poly_uint64.
4188 * target.def (default_preferred_vector_alignment): Likewise.
4189 * targhooks.c (default_preferred_vector_alignment): Likewise.
4190 * targhooks.h (default_preferred_vector_alignment): Likewise.
4191 * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
4192 (vect_compute_data_ref_alignment): Adapt to vector alignment being a
4194 (vect_update_misalignment_for_peel): Likewise.
4195 (vect_enhance_data_refs_alignment): Likewise.
4196 (vect_find_same_alignment_drs): Likewise.
4197 (vect_duplicate_ssa_name_ptr_info): Likewise.
4198 (vect_setup_realignment): Likewise.
4199 (vect_can_force_dr_alignment_p): Change alignment parameter type to
4201 * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
4202 mask with a compile time variable vector alignment.
4203 (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
4205 (vect_do_peeling): Exit early if vector alignment is not constant.
4206 * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
4208 (vectorizable_store): Likewise.
4209 (vectorizable_load): Likweise.
4210 * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
4212 (vect_known_alignment_in_bytes): Adapt to vector alignment being a
4214 (vect_can_force_dr_alignment_p): Change alignment parameter type to
4217 2018-11-13 Richard Biener <rguenther@suse.de>
4219 PR tree-optimization/87962
4220 * tree-vect-loop.c (vect_is_simple_reduction): More reliably
4221 detect outer reduction for disqualifying in-loop uses.
4223 2018-11-13 Richard Biener <rguenther@suse.de>
4225 PR tree-optimization/87967
4226 * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
4227 for constants for the scalar loop.
4229 2018-11-13 Alan Modra <amodra@gmail.com>
4231 * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
4232 integer constants when -mcmodel=medium.
4234 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4236 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
4237 * config/arc/elf.h (PROFILE_HOOK): Define.
4238 * config/arc/linux.h (PROFILE_HOOK): Likewise.
4240 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4242 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
4243 jump table data in the text section.
4245 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4247 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
4248 to fit the eh_return pattern.
4249 * config/arc/arc.md (eh_return): Define.
4250 (VUNSPEC_ARC_EH_RETURN): Likewise.
4251 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
4253 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
4256 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4258 * common/config/arc/arc-common.c (arc_option_optimization_table):
4259 Millicode optimization is default on for size optimizations.
4260 * config/arc/arc-protos.h (arc_check_multi): New function.
4261 * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
4262 (ENTER_LEAVE_START_REG): Define.
4263 (ENTER_LEAVE_END_REG): Likewise.
4264 (arc_override_options): Disable millicode when long calls option
4266 (arc_frame_info): Change it from int to bool.
4267 (arc_compute_frame_size): Clean up.
4268 (arc_save_restore): Remove.
4269 (frame_save_reg): New function.
4270 (frame_restore_reg): Likewise.
4271 (arc_enter_leave_p): Likewise.
4272 (arc_save_callee_saves): Likewise.
4273 (arc_restore_callee_saves): Likewise.
4274 (arc_save_callee_enter): Likewise.
4275 (arc_restore_callee_leave): Likewise.
4276 (arc_save_callee_milli): Likewise.
4277 (arc_restore_callee_milli): Likewise.
4278 (arc_expand_prologue): Reimplement to emit enter/leave
4280 (arc_expand_epilogue): Likewise.
4281 (arc_check_multi): New function.
4282 * config/arc/arc.md (push_multi_fp): New pattern.
4283 (push_multi_fp_blink): Likewise.
4284 (pop_multi_fp): Likewise.
4285 (pop_multi_fp_blink): Likewise.
4286 (pop_multi_fp_ret): Likewise.
4287 (pop_multi_fp_blink_ret): Likewise.
4288 * config/arc/arc.opt (mmillicode): Update option.
4289 (mcode-density-frame): New option.
4290 * config/arc/predicates.md (push_multi_operand): New predicate.
4291 (pop_multi_operand): Likewise.
4292 * doc/invoke.texi (ARC): Update ARC options information.
4295 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4297 * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
4298 * config/arc/arc.c (operands_ok_ldd_std): New function.
4299 (mem_ok_for_ldd_std): Likewise.
4300 (gen_operands_ldd_std): Likewise.
4301 * config/arc/arc.md: Add peephole2 rules for std/ldd.
4303 2018-11-13 Eric Botcazou <ebotcazou@adacore.com>
4305 * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
4306 test on stack_usage_file.
4307 (lang_dependent_init): Do not open the .su file if generating LTO.
4309 2018-11-13 Jakub Jelinek <jakub@redhat.com>
4311 PR rtl-optimization/87918
4312 * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
4313 simplify_gen_relational rather than simplify_gen_binary.
4315 2018-11-13 Richard Biener <rguenther@suse.de>
4317 * tree-ssanames.h (set_range_info): Use value_range_base.
4318 (get_range_info): Likewise.
4319 * tree-ssanames.c (set_range_info): Likewise.
4320 (get_range_info): Likewise.
4321 * tree-vrp.c (value_range_base::union_helper): Split
4322 out common parts of value_range[_base]::union_.
4323 (value_range_base::union_): Update.
4324 (value_range::union_): Likewise.
4325 (determine_value_range_1): Use value_range_base.
4326 (determine_value_range): Likewise.
4327 * tree-vrp.h (value_range_base::union_helper): Move ...
4328 (value_range::union_helper): ... from here.
4330 2018-11-13 Alan Modra <amodra@gmail.com>
4332 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
4334 (rs6000_secondary_reload_gpr): Don't call find_replacement.
4336 2018-11-13 Sandra Loosemore <sandra@codesourcery.com>
4339 * doc/invoke.texi (Optimize Options): Clarify that the
4340 l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
4341 apply to data cache size.
4343 2018-11-13 Alan Modra <amodra@gmail.com>
4345 * config/rs6000/predicates.md (logical_const_operand),
4346 (logical_operand): Correct comment.
4347 * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
4349 2018-11-13 Alan Modra <amodra@gmail.com>
4351 * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
4353 2018-11-13 Alan Modra <amodra@gmail.com>
4355 * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
4357 * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
4358 covered by alternative.
4359 (movcc_internal1): Ignore h for register preference.
4360 (mov<mode>_hardfloat64): Likewise.
4361 (mov<mode>_softfloat): Ignore c, l, h for register preference.
4363 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
4365 PR preprocessor/47823
4366 * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
4369 * doc/extend.texi (Pragmas): Note additional pragmas documented
4372 2018-11-13 Kugan Vivekanandarajah <kuganv@linaro.org>
4376 * tree-scalar-evolution.c (expression_expensive_p): Make
4377 BUILTIN POPCOUNT as expensive when backend does not define it.
4379 2018-11-12 Fredrik Noring <noring@nocrew.org>
4381 * config.gcc: Update with-llsc defaults for MIPS r5900.
4383 2018-11-12 Martin Liska <mliska@suse.cz>
4385 PR gcov-profile/87442
4386 * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
4388 * doc/invoke.texi: Document them.
4389 * tree-profile.c (parse_profile_filter): New.
4390 (parse_profile_file_filtering): Likewise.
4391 (release_profile_file_filtering): Likewise.
4392 (include_source_file_for_profile): Likewise.
4393 (tree_profiling): Filter source files based on the
4394 newly added options.
4396 2018-11-12 Martin Liska <mliska@suse.cz>
4399 * doc/extend.texi: Add missing values for __builtin_cpu_is and
4400 __builtin_cpu_supports for x86 target.
4402 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
4405 * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
4407 (Arithmetic): Likewise for "lo_sum".
4409 2018-11-12 Renlin Li <renlin.li@arm.com>
4412 * dse.c (get_stored_val): Add check for compile-time constantness
4415 2018-11-12 Sudakshina Das <sudi.das@arm.com>
4417 * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
4418 (ARMv8_5a): New fgroup.
4419 (armv8.5-a): New arch.
4420 (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
4422 * config/arm/arm-tables.opt: Regenerate.
4423 * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
4424 * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
4425 * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
4426 Add matching rules for -march=armv8.5-a and extensions.
4427 * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
4428 Add sb and predres to all armv8-a except armv8.5-a.
4430 2018-11-12 Richard Biener <rguenther@suse.de>
4432 * tree-vrp.h (value_range[_base]::set): Make public. Provide
4433 overload for single value.
4434 (value_range[_base]::set_nonnull): New.
4435 (value_range[_base]::set_null): Likewise.
4436 (value_range): Document bitmap copying behavior, mark
4437 copy constructor and assignment operator deleted.
4438 (value_range::move): New.
4439 (value_range::set_and_canonicalize): Default bitmap to zero.
4440 (set_value_range_to_nonnull): Remove.
4441 (set_value_range_to_null): Likewise.
4442 (set_value_range): Likewise.
4443 (set_value_range_to_value): Likewise.
4444 (extract_range_from_unary_expr): Work on value_range_base.
4445 (extract_range_from_binary_expr_1): Likewise. Rename to...
4446 (extract_range_from_binary_expr): ... this.
4447 * tree-vrp.c (value_range::update): Clear equiv bitmap
4449 (value_range::move): New, move equiv bitmap.
4450 (value_range_base::set_undefined): Avoid assignment.
4451 (value_range::set_undefined): Likewise.
4452 (value_range_base::set_varying): Likewise.
4453 (value_range::set_varying): Likewise.
4454 (set_value_range): Remove.
4455 (value_range_base::set): New overload for value.
4456 (value_range::set): Likewise.
4457 (set_value_range_to_nonnull): Remove.
4458 (value_range_base::set_nonnull): New.
4459 (value_range::set_nonnull): Likewise.
4460 (set_value_range_to_null): Remove.
4461 (value_range_base::set_null): New.
4462 (value_range::set_null): Likewise.
4463 (range_is_null): Work on value_range_base.
4464 (range_is_nonnull): Likewise.
4465 (ranges_from_anti_range): Likewise.
4466 (extract_range_into_wide_ints): Likewise.
4467 (extract_range_from_multiplicative_op): Likewise.
4468 (extract_range_from_binary_expr): Likewise. Update for API changes.
4469 (extract_range_from_unary_expr): Likewise. Remove OBJ_TYPE_REF
4471 (value_range::intersect_helper): Avoid copy and assignment.
4472 (value_range::union_helper): Likewise.
4473 (determine_value_range_1): Adjust.
4474 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
4475 Avoid assignment by using move.
4476 (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
4477 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4479 * tree-ssanames.c (get_range_info): Likewise.
4480 * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
4481 * vr-values.c (vr_values::get_value_range): Adjust.
4482 (vr_values::update_value_range): Likewise.
4483 (symbolic_range_based_on_p): Work on value_range_base.
4484 (vr_values::extract_range_from_binary_expr): Use value_range_base.
4485 (vr_values::extract_range_from_unary_expr): Likewise.
4486 (vr_values::extract_range_from_cond_expr): Avoid assignment.
4487 (vr_values::extract_range_from_comparison): Adjust.
4488 (vr_values::check_for_binary_op_overflow): Use value_range_base.
4489 (vr_values::extract_range_basic): Adjust.
4490 (vr_values::adjust_range_with_scev): Likewise.
4491 (vr_values::vrp_visit_assignment_or_call): Likewise.
4492 (vr_values::get_vr_for_comparison): Change API to avoid
4493 assignment and copy construction.
4494 (vr_values::compare_name_with_value): Adjust accordingly.
4495 (vr_values::compare_names): Likewise.
4496 (vr_values::extract_range_from_phi_node): Avoid assignment and
4497 bogus in-place modify of equiv bitmap.
4498 (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
4499 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
4500 for extract_range_from_unary_expr API change.
4501 * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
4503 2018-11-12 Eric Botcazou <ebotcazou@adacore.com>
4505 * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
4507 2018-11-12 Richard Biener <rguenther@suse.de>
4509 * tree-vrp.h (value_range_base::symbolic_p,
4510 value_range_base::constant_p, value_range_base::zero_p,
4511 value_range_base::singleton_p): Move from value_range.
4512 (value_range::dump): Add.
4513 * gimple-ssa-evrp-analyze.c
4514 (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
4515 * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
4516 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4518 * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
4519 (value_range::constant_p): Likewise.
4520 (value_range::singleton_p): Likewise.
4521 (value_range_base::dump): Add.
4522 (set_value_range_to_undefined): Remove.
4523 (set_value_range_to_varying): Likewise.
4524 (range_int_cst_p): Take value_range_base argument.
4525 (range_int_cst_singleton_p): Likewise.
4526 (value_range_constant_singleton): Likewise.
4527 (vrp_set_zero_nonzero_bits): Likewise.
4528 (extract_range_from_multiplicative_op): Use set_varying.
4529 (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
4530 (extract_range_from_unary_expr): Likewise.
4531 (dump_value_range_base): Change to overload of dump_value_range.
4532 (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
4533 (vrp_prop::visit_stmt): Likewise.
4534 (value_range::intersect_helper): Likewise.
4535 (value_range::union_helper): Likewise.
4536 (determine_value_range_1): Likewise.
4538 2018-11-12 Richard Biener <rguenther@suse.de>
4540 * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
4541 (set_value_range_to_null): Likewise.
4542 * vr-values.c (vr_values::extract_range_from_comparison):
4543 Clear equiv for constant singleton ranges.
4545 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
4547 * config/i386/sse.md: Combine VFIXUPIMM* patterns
4548 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4549 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4550 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
4551 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4552 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4553 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
4555 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
4558 * doc/extend.texi (Common Type Attributes): For the align type
4559 attribute, copy language about decreasing alignment from the
4560 corresponding variable attribute.
4562 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
4564 * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
4567 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
4570 * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
4571 -frtti and -fno-rtti code.
4573 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
4576 * doc/extend.texi (Other Builtins): Document probability associated
4577 with __builtin_expect.
4579 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
4582 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
4583 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
4584 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
4585 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
4587 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
4589 * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
4591 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
4593 * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
4595 2018-11-11 Richard Biener <rguenther@suse.de>
4597 * tree-vrp.h (class value_range_base): New base class for
4598 value_range containing all but the m_equiv member.
4599 (dump_value_range_base): Add.
4600 (range_includes_zero_p): Work on value_range_base.
4601 * tree-vrp.c (value_range_base::set): Split out base handling
4603 (value_range::set): this.
4604 (value_range::set_equiv): New.
4605 (value_range_base::value_range_base): New constructors.
4606 (value_range_base::check): Split out base handling from...
4607 (value_range::check): this.
4608 (value_range::equal_p): Refactor in terms of
4609 ignore_equivs_equal_p which is now member of the base.
4610 (value_range_base::set_undefined): New.
4611 (value_range_base::set_varying): Likewise.
4612 (value_range_base::dump):Split out base handling from...
4613 (value_range::dump): this.
4614 (value_range_base::set_and_canonicalize): Split out base handling
4616 (value_range::set_and_canonicalize): this.
4617 (value_range_base::union_): New.
4618 * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
4620 * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
4621 instead of value_range everywhere.
4622 (ipcp_vr_lattice::print): Use dump_value_range_base.
4623 (ipcp_vr_lattice::meet_with): Adjust.
4624 (ipcp_vr_lattice::meet_with_1): Likewise.
4625 (ipa_vr_operation_and_type_effects): Likewise.
4626 (propagate_vr_across_jump_function): Likewise.
4627 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
4628 (ipa_get_value_range): Likewise.
4629 (ipa_set_jfunc_vr): Likewise.
4630 (ipa_compute_jump_functions_for_edge): Likewise.
4632 2018-11-10 Sandra Loosemore <sandra@codesourcery.com>
4635 * doc/invoke.texi (Optimize Options): Add @opindex entries
4636 for the positive forms of -fno-xxx and -mno-xxx options
4637 that were lacking them.
4639 2018-11-10 Segher Boessenkool <segher@kernel.crashing.org>
4641 * combine.c (make_more_copies): Only make an intermediate copy if the
4642 dest of a move is a pseudo.
4644 2018-11-09 Maya Rashish <coypu@sdf.org>
4647 * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
4648 (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
4650 2018-11-09 Sandra Loosemore <sandra@codesourcery.com>
4654 * doc/invoke.texi (Optimize Options): Clarify default behavior
4655 for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
4657 2018-11-09 Bernd Edlinger <bernd.edlinger@hotmail.de>
4659 PR tree-optimization/87940
4660 * expr.c (string_constant): Don't strip NOPS in subexpressions.
4661 Fold PLUS_EXPR correctly.
4663 2018-11-09 Ilya Leoshkevich <iii@linux.ibm.com>
4666 * config/s390/s390.c (s390_safe_relative_long_p): New function.
4667 (annotate_constant_pool_refs): Skip insns which support
4668 relative addressing.
4669 (annotate_constant_pool_refs_1): New helper function.
4670 (find_constant_pool_ref): Skip insns which support relative
4672 (find_constant_pool_ref_1): New helper function.
4673 (replace_constant_pool_ref): Skip insns which support
4674 relative addressing.
4675 (replace_constant_pool_ref_1): New helper function.
4676 (s390_mainpool_start): Adapt to the new signature.
4677 (s390_mainpool_finish): Likewise.
4678 (s390_chunkify_start): Likewise.
4679 (s390_chunkify_finish): Likewise.
4680 (pass_s390_early_mach::execute): Likewise.
4681 (s390_prologue_plus_offset): Likewise.
4682 (s390_emit_prologue): Likewise.
4683 (s390_emit_epilogue): Likewise.
4685 2018-11-09 Jakub Jelinek <jakub@redhat.com>
4687 * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
4688 but unsupported lastprivate with conditional modifier.
4690 2018-11-09 Jeff Law <law@redhat.com>
4692 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
4693 unused argument better. Add gcc_unreachable to silence warning.
4695 2018-11-09 Martin Sebor <msebor@redhat.com>
4698 * attribs.c (has_attribute): New helper function.
4699 (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
4700 * attribs.h (decls_mismatched_attributes): Declare.
4701 * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
4702 (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
4703 * common.opt (-Wattribute-alias): Take an argument.
4704 (-Wno-attribute-alias): New option.
4705 * doc/extend.texi (Common Function Attributes): Document copy.
4706 (Common Variable Attributes): Same.
4707 * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
4708 (-Wattribute-alias): Document new option argument.
4710 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
4712 * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
4713 matching criteria. Remove unused array initializer.
4715 2018-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
4716 Jinsong Ji <jji@us.ibm.com>
4718 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
4719 constraints by introducing a new temporary.
4720 (_mm_cvtss_si64): Likewise.
4722 2018-11-09 Martin Liska <mliska@suse.cz>
4724 * common.opt: Add -fipa-stack-alignment flag.
4725 * doc/invoke.texi: Document it.
4726 * final.c (rest_of_clean_state): Guard stack
4727 shrinking with flag.
4729 2018-11-09 Martin Liska <mliska@suse.cz>
4731 * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
4733 (ipa_discover_variable_flags): ... this.
4734 * common.opt: Come up with new flag -fipa-reference-addressable.
4735 * doc/invoke.texi: Document it.
4736 * ipa-reference.c (propagate): Call the renamed fn.
4737 * ipa-visibility.c (whole_program_function_and_variable_visibility):
4739 * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
4741 (ipa_discover_variable_flags): ... this. Discover
4742 non-addressable variables only with the newly added flag.
4743 * opts.c: Enable the newly added flag with -O1 and higher
4746 2018-11-09 David Malcolm <dmalcolm@redhat.com>
4748 * json.cc (selftest::test_writing_literals): Fix comment.
4750 2018-11-09 Martin Liska <mliska@suse.cz>
4752 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
4753 string to a stack buffer.
4754 (aarch64_parse_cpu): Likewise.
4755 (aarch64_parse_tune): Likewise.
4757 2018-11-09 Richard Biener <rguenther@suse.de>
4759 PR tree-optimization/87953
4760 * tree-vect-loop.c (vectorizable_reduction): For analysis
4761 always pass ops[0] to vectorizable_condition.
4763 2018-11-09 Stafford Horne <shorne@gmail.com>
4764 Richard Henderson <rth@twiddle.net>
4765 Joel Sherrill <joel@rtems.org>
4767 * common/config/or1k/or1k-common.c: New file.
4768 * config/or1k/*: New.
4769 * config.gcc (or1k*-*-*): New.
4770 * configure.ac (or1k*-*-*): New test for openrisc tls.
4771 * configure: Regenerated.
4772 * doc/install.texi: Document OpenRISC triplets.
4773 * doc/invoke.texi: Document OpenRISC arguments.
4774 * doc/md.texi: Document OpenRISC.
4776 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
4778 * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
4779 (arm7tdmi-s): Delete CPU.
4780 (arm710t): Add aliases for arm720t and arm740t.
4781 (arm720t, arm740t): Delete CPUs.
4782 (arm920t): Add aliases for arm920, arm922t and arm940t.
4783 (arm920, arm922t, arm940t): Delete CPUs.
4784 (arm10tdmi): Add alias for arm1020t.
4785 (arm1020t): Delete CPU.
4786 (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
4787 (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
4788 (arm10e): Add aliases for arm1020e and arm1022e.
4789 (arm1020e, arm1022e): Delete CPU.
4790 * config/arm/arm.md (generic_sched): Remove entries that are now
4792 (generic_vfp): Likewise.
4793 * config/arm/arm1020e.md: Simplify tuning selection based on alias
4795 * config/arm/arm-tune.md: Regenerated.
4796 * config/arm/arm-tables.opt: Regenerated.
4798 2018-11-09 Richard Biener <rguenther@suse.de>
4800 PR tree-optimization/87621
4801 * tree-vect-loop.c (vectorizable_reduction): Handle reduction
4802 op with only phi inputs.
4803 * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
4804 (ch_base::copy_headers): Run CSE on copied loop headers.
4805 (pass_ch_vect::process_loop_p): Simplify.
4807 2018-11-09 Alexandre Oliva <oliva@adacore.com>
4809 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
4810 for non-w64 x86_64 biarch.
4812 2018-11-09 Alexandre Oliva <aoliva@redhat.com>
4814 PR rtl-optimization/86438
4815 * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
4816 of in_b for the compare if in_b is SET_DEST.
4819 * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
4820 non-toplevel UNSPEC.
4822 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4824 * tree-vrp.c (value_range::check): Do not access internals
4826 (value_range::singleton_p): Same.
4827 (value_range::type): Same.
4828 (vrp_finalize): Use value_range API.
4830 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4832 * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
4834 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4836 * vr-values.c (vr_values::get_value_range): Use value_range API
4837 instead of piecing together ranges.
4838 (vr_values::update_value_range): Same.
4840 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4842 * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
4843 of performing ad-hoc calculations.
4844 * tree-ssanames.c (set_range_info): New overloaded function
4845 accepting value_range &.
4846 (get_range_info): Same.
4847 * tree-ssanames.h (set_range_info_raw): Remove.
4848 (set_range_info): New prototype.
4849 (get_range_info): Same.
4850 * tree-vrp.h (value_range::null_p): Rename to zero_p.
4851 * tree-vrp.c (value_range::null_p): Same.
4853 2018-11-09 Jan Hubicka <jh@suse.cz>
4855 * tree.c (fld_type_variant_equal_p): Test user align flag.
4856 (flt_type_variant): Copy user align flag.
4857 (fld_incomplete_type_of): Clear it.
4859 2018-11-09 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4861 * config/arm/neon.md (div<mode>3): New pattern.
4863 2018-11-08 Andi Kleen <ak@linux.intel.com>
4865 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
4866 (OPTION_MASK_ISA_PTWRITE_UNSET): New.
4867 (ix86_handle_option): Handle OPT_mptwrite.
4868 * config/i386/cpuid.h (bit_PTWRITE): Add.
4869 * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
4870 * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
4871 * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
4872 * config/i386/i386.c (ix86_target_string): Handle ptwrite.
4873 (ix86_option_override_internal): Handle PTA_PTWRITE.
4874 (ix86_valid_target_attribute_inner_p): Define ptwrite.
4875 (def_builtin2): Force UINT64 to be 64bit only.
4876 * config/i386/i386.h (TARGET_PTWRITE): Add.
4877 (TARGET_PTWRITE_P): Add.
4879 * config/i386/i386.md: Define ptwrite.
4880 * config/i386/i386.opt: Add -mptwrite.
4881 * config/i386/immintrin.h (_ptwrite64): Add.
4883 * doc/extend.texi: Document __builtin_ia32_ptwrite*.
4884 * doc/invoke.texi: Document -mptwrite.
4886 2018-11-08 Peter Bergner <bergner@linux.ibm.com>
4888 PR rtl-optimization/87600
4889 * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
4890 * lra-constraints.c (process_alt_operands): Skip illegal hard
4891 register usage. Prefer reloading non hard register operands.
4893 2018-11-08 Sandra Loosemore <sandra@codesourcery.com>
4896 * doc/invoke.texi (Optimize Options): Clarify default behavior
4897 for -fno-sched-interblock and -fno-sched-spec.
4899 2018-11-08 Roman Geissler <roman.geissler@amadeus.com>
4901 * collect2.c (linker_select): Add USE_LLD_LD.
4902 (ld_suffixes): Add ld.lld.
4903 (main): Handle -fuse-ld=lld.
4904 * common.opt (-fuse-ld=lld): New option.
4905 * doc/invoke.texi (-fuse-ld=lld): Document.
4906 * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
4908 2018-11-08 Paul Koning <ni1d@arrl.net>
4910 * config/pdp11/constraints.md: Add "Z" series constraints for use
4911 with pre-dec and post-inc addressing.
4912 * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
4913 (pdp11_expand_operands): Add int argument (word count).
4914 (pdp11_sp_frame_offset): Delete.
4915 (pdp11_cmp_length): New function.
4916 (pushpop_regeq): New function.
4917 * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
4919 (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
4921 (pdp11_initial_elimination_offset): Ditto.
4922 (pdp11_expand_operands): Add word count argument. Bugfixes.
4923 (output_move_multiple): Change how pointer adjustment is done.
4924 (pdp11_gen_int_label): Correct format.
4925 (output_ascii): Ditto.
4926 (pdp11_asm_output_var): Add code for DEC assembler case.
4927 (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
4929 (legitimate_const_double_p): Ditto.
4930 (pdp11_register_move_cost): Adjust for new register classes.
4931 (pdp11_regno_reg_class): Ditto.
4932 (expand_block_move): Delete.
4933 (pushpop_regeq): New function.
4934 (pdp11_legitimate_address_p): Bugfix in check for constant
4936 (pdp11_sp_frame_offset): Delete.
4937 (pdp11_reg_save_size): New helper function for new frame layout.
4938 (output_addr_const_pdp11): Remove CONST_DOUBLE case.
4939 (pdp11_expand_shift): Bugfix in check for constant shift count.
4940 (pdp11_shift_length): Ditto.
4941 (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
4942 (pdp11_cmp_length): New function.
4943 * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
4944 some compile options.
4945 (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
4946 (CALL_USED_REGISTERS): Ditto.
4947 (ELIMINABLE_REGS): Ditto.
4948 (REGISTER_NAMES): Ditto.
4949 (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
4951 (REG_CLASS_NAMES): Ditto.
4952 (REG_CLASS_CONTENTS): Ditto. Also remove
4953 HARD_FRAME_POINTER_REGNUM.
4954 (CPU_REG_CLASS): New macro.
4955 (CLASS_MAX_NREGS): Adjust for new register classes.
4956 (FUNCTION_PROFILER): Make no-op.
4957 (may_call_alloca): Remove unused declaration.
4958 (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
4959 (ASM_OUTPUT_SKIP): Fix format.
4960 * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
4961 (HARD_FRAME_POINTER_REGNUM): Remove.
4963 (*rts): Rename. Remove epilogue related checks.
4964 (cmpsi, cmpdi): New insn.
4965 (cbranch<mode>4): Change to apply to SI and DI modes as well.
4966 (mov<mode>): Change constraints to enforce that push/pop
4967 destination cannot use the same register as source.
4968 (*mov<mode><cc_cc>): Ditto.
4969 (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
4970 at assembly output rather than as RTL expander.
4971 (zero_extendqihi2): Bugfix in check for same registers.
4972 (adddi3_nocc): Bugfix in check for constant operand.
4973 (addsi3_nocc): Ditto.
4974 (subdi3_nocc): Ditto.
4975 (subsi3_nocc): Ditto.
4976 (negdi2_nocc): Copy input to pdp11_expand_operands.
4977 (negsi2_nocc): Ditto.
4978 (bswap2_nocc): Ditto.
4979 * config/pdp11/pdp11.opt (mlra): Fix documentation.
4980 * config/pdp11/t-pdp11: Use -Os.
4982 2018-11-08 Richard Earnshaw <rearnsha@arm.com>
4984 * config/arm/parsecpu.awk (/alias/): New parsing rule.
4985 (/begin cpu/): Check that the cpu name hasn't been previously defined.
4986 (gen_comm_data): Print out CPU alias tables.
4987 (check_cpu): Match aliases when checking the CPU name.
4988 * config/arm/arm-protos.h (cpu_alias): New structure.
4989 (cpu_option): Add entry for aliases.
4990 * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
4991 strongarm1100 and strongarm1110.
4992 (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
4993 (config/arm/arm-generic.md): Remove redundant references to
4994 strongarm110, strongarm1100 and strongarm1110.
4995 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4996 Scan aliases for additional hints.
4997 (arm_parse_cpu_option_name): Also match a cpu name against the list
4999 * config/arm/arm-tables.opt: Regenerated.
5000 * config/arm/arm-tune.md: Regenerated.
5002 2018-11-08 Jakub Jelinek <jakub@redhat.com>
5004 * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
5005 BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
5006 BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
5007 BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
5008 BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
5009 BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
5010 * gengtype.c (open_base_files): Add omp-general.h.
5011 * gimple.c (gimple_build_omp_critical):
5012 (gimple_build_omp_taskgroup): Add CLAUSES argument. Call
5013 gimple_omp_taskgroup_set_clauses.
5014 (gimple_build_omp_atomic_load): Add mo argument, call
5015 gimple_omp_atomic_set_memory_order.
5016 (gimple_build_omp_atomic_store): Likewise.
5017 (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
5018 * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
5020 (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
5021 of GSS_OMP_SINGLE_LAYOUT, adjust comments.
5022 * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
5023 and GF_OMP_ATOMIC_MEMORY_ORDER. Remove GF_OMP_ATOMIC_SEQ_CST, use
5024 different value for GF_OMP_ATOMIC_NEED_VALUE.
5025 (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
5027 (struct gimple_statement_omp_single_layout): And remove here.
5028 (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
5029 than gimple_statement_omp_single_layout.
5030 (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
5032 (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
5033 (gimple_omp_subcode): Formatting fix.
5034 (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
5035 gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
5036 gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
5037 gimple_omp_teams_host, gimple_omp_teams_set_host,
5038 gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
5039 gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
5040 gimple_omp_taskgroup_set_clauses): New inline functions.
5041 (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
5042 (gimple_build_omp_atomic_store): Likewise.
5043 (gimple_omp_atomic_seq_cst_p): Remove.
5044 (gimple_omp_atomic_memory_order): New function.
5045 (gimple_omp_atomic_set_seq_cst): Remove.
5046 (gimple_omp_atomic_set_memory_order): New function.
5047 (gimple_build_omp_taskgroup): Add clauses argument.
5048 * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
5049 (dump_gimple_omp_task): Print taskwait with depend clauses.
5050 (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
5051 dump_omp_atomic_memory_order.
5052 (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
5053 * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
5054 GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
5055 (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
5056 renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
5057 ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
5058 ORT_UNTIED_TASKLOOP enumerators.
5059 (enum gimplify_defaultmap_kind): New.
5060 (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
5061 target_map_pointers_as_0len_arrays members, add defaultmap.
5062 (new_omp_context): Initialize defaultmap member.
5063 (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
5064 (maybe_fold_stmt): Don't fold even in host teams regions.
5065 (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
5066 ORT_WORKSHARE. Test ctx->defaultmap[GDMK_SCALAR] instead of
5067 ctx->omp_firstprivatize_variable.
5068 (omp_add_variable): Don't add private/firstprivate for VLAs in
5070 (omp_default_clause): Print "taskloop" rather than "task" if
5072 (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
5073 Handle new defaultmap clause kinds.
5074 (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. Allow simd
5075 iterator to be lastprivate or private. Fix up diagnostics if linear
5076 is used on collapse>1 simd iterator.
5077 (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
5078 (gimplify_omp_depend): New function.
5079 (gimplify_scan_omp_clauses): Add shared clause on parallel for
5080 combined parallel master taskloop{, simd} if taskloop has
5081 firstprivate, lastprivate or reduction clause. Handle
5082 OMP_CLAUSE_REDUCTION_TASK diagnostics. Adjust tests for
5083 ORT_COMBINED_TEAMS. Gimplify depend clauses with iterators. Handle
5084 cancel and simd OMP_CLAUSE_IF_MODIFIERs. Handle
5085 OMP_CLAUSE_NONTEMPORAL. Handle new defaultmap clause kinds. Handle
5086 OMP_CLAUSE_{TASK,IN}_REDUCTION. Diagnose invalid conditional
5088 (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL. Handle
5089 GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
5090 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL. Handle
5091 OMP_CLAUSE_{TASK,IN}_REDUCTION.
5092 (gimplify_omp_task): Handle taskwait with depend clauses.
5093 (gimplify_omp_for): Add shared clause on parallel for combined
5094 parallel master taskloop{, simd} if taskloop has firstprivate,
5095 lastprivate or reduction clause. Use ORT_TASKLOOP or
5096 ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK. Adjust
5097 tests for ORT_COMBINED_TEAMS. Handle C++ range for loops with
5098 NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS. Firstprivatize
5099 __for_end and __for_range temporaries on OMP_PARALLEL for
5100 distribute parallel for{, simd}. Move OMP_CLAUSE_REDUCTION
5101 and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
5102 sandwiched in between two taskloops.
5103 (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
5104 instead of ctx->omp_firstprivatize_variable.
5105 (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
5106 ORT_COMBINED_HOST_TEAMS if not inside of target construct. If
5107 host teams, use gimplify_and_return_first etc. for body like
5108 for target or target data constructs, and at the end call
5109 gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
5110 (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
5111 of OMP_ATOMIC_SEQ_CST, pass it as new argument to
5112 gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
5113 gimple_omp_atomic_set_seq_cst calls.
5114 (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
5115 case, handle taskgroup clauses.
5116 * lto-streamer-out.c (hash_tree): Handle
5117 OMP_CLAUSE_{TASK,IN}_REDUCTION.
5118 * Makefile.in (GTFILES): Add omp-general.h.
5119 * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
5120 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
5121 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
5122 BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
5123 BUILT_IN_GOMP_LOOP_DOACROSS_START,
5124 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
5125 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
5126 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
5127 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
5128 BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
5129 BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
5130 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
5131 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
5132 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
5133 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
5134 BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
5135 BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
5136 BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
5137 BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
5138 BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
5139 * omp-expand.c (workshare_safe_to_combine_p): Return false for
5140 non-worksharing loops.
5141 (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
5142 (determine_parallel_type): Don't combine parallel with worksharing
5143 which has _reductemp_ clause.
5144 (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
5145 GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
5146 if there is nonmonotonic modifier or if there is no modifier and no
5147 ordered clause. For dynamic and guided schedule without monotonic
5148 and nonmonotonic modifier, default to nonmonotonic.
5149 (expand_omp_for): Likewise. Adjust expand_omp_for_generic caller, use
5150 GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
5152 (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
5153 there are any reduction clauses.
5154 (expand_taskwait_call): New function.
5155 (expand_teams_call): New function.
5156 (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
5157 expand_teams_call for it. Formatting fix. Handle taskwait with
5159 (expand_omp_for_generic): Add SCHED_ARG argument. Handle expansion
5160 of worksharing loops with task reductions.
5161 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
5162 expansion of worksharing loops with task reductions.
5163 (expand_omp_sections): Handle expansion of sections with task
5165 (expand_omp_synch): For host teams call expand_omp_taskreg.
5166 (omp_memory_order_to_memmodel): New function.
5167 (expand_omp_atomic_load, expand_omp_atomic_store,
5168 expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
5169 instead of gimple_omp_atomic_seq_cst_p.
5170 (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
5171 depend clauses as a standalone directive.
5172 * omp-general.c (enum omp_requires): New variable.
5173 (omp_extract_for_data): Initialize have_reductemp member. Allow
5174 NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
5175 GT_EXPR loops depending on incr sign. Formatting fixes.
5176 * omp-general.h (struct omp_for_data): Add have_reductemp member.
5177 (enum omp_requires): New enum.
5178 (omp_requires_mask): Declare.
5179 * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
5181 * omp-low.c (struct omp_context): Add task_reductions and
5182 task_reduction_map fields.
5183 (is_host_teams_ctx): New function.
5184 (is_taskreg_ctx): Return true also if is_host_teams_ctx.
5185 (use_pointer_for_field): Use is_global_var instead of
5186 TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
5188 (build_outer_var_ref): Ignore taskgroup outer contexts.
5189 (delete_omp_context): Release task_reductions and task_reduction_map.
5190 (scan_sharing_clauses): Don't add any fields for reduction clause on
5191 taskloop. Handle OMP_CLAUSE__REDUCTEMP_. Handle
5192 OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
5193 modifier. Don't ignore shared clauses in is_host_teams_ctx contexts.
5194 Handle OMP_CLAUSE_NONTEMPORAL.
5195 (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
5197 (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
5198 clauses with task modifier.
5199 (scan_omp_task): Handle taskwait with depend clauses.
5200 (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
5201 first. Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
5202 Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
5203 (scan_omp_for): Fix comment formatting.
5204 (scan_omp_teams): Handle host teams constructs.
5205 (check_omp_nesting_restrictions): Allow teams with no outer
5206 OpenMP context. Adjust diagnostics for teams strictly nested into
5207 some explicit OpenMP construct other than target. Allow OpenMP atomics
5208 inside of simd regions.
5209 (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
5210 (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
5211 taskreg_nesting_level while scanning host teams construct.
5212 (task_reduction_read): New function.
5213 (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
5214 construct. Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
5215 clauses. Handle OMP_CLAUSE_REDUCTION with task modifier. Remove
5216 second argument create_tmp_var if it is NULL. Don't ignore shared
5217 clauses in is_host_teams_ctx contexts. Handle
5218 OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
5220 (lower_reduction_clauses): Ignore reduction clauses with task
5221 modifier. Remove second argument create_tmp_var if it is NULL.
5222 Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
5223 (lower_send_clauses): Ignore reduction clauses with task modifier.
5224 Handle OMP_CLAUSE__REDUCTEMP_. Don't send anything for
5225 OMP_CLAUSE_REDUCTION on taskloop. Handle OMP_CLAUSE_IN_REDUCTION.
5226 (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
5227 rely that it is the last stmt in body so far. Ignore outer taskgroup
5229 (omp_task_reductions_find_first, omp_task_reduction_iterate,
5230 lower_omp_task_reductions): New functions.
5231 (lower_omp_sections): Handle reduction clauses with taskgroup
5232 modifiers. Adjust maybe_add_implicit_barrier_cancel caller.
5233 (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
5234 (lower_omp_for): Likewise. Handle reduction clauses with taskgroup
5236 (lower_omp_taskgroup): Handle taskgroup reductions.
5237 (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
5238 Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
5239 (lower_depend_clauses): If there are any
5240 OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
5241 depend clauses, use a new array format. If OMP_CLAUSE_DEPEND_LAST is
5242 seen, assume lowering is done already and return early. Set kind
5243 on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
5244 (lower_omp_taskreg): Handle reduction clauses with task modifier on
5245 parallel construct. Handle reduction clause on taskloop construct.
5246 Handle taskwait with depend clauses.
5247 (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
5248 for host teams constructs.
5249 * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
5250 nontemporal and _reductemp_ clause entries.
5251 (omp_clause_code_name): Likewise.
5252 (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
5253 OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
5254 * tree-core.h (enum omp_clause_code): Add
5255 OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
5256 (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
5257 (struct tree_base): Add omp_atomic_memory_order field into union.
5258 Remove OMP_ATOMIC_SEQ_CST comment.
5259 (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
5260 and OMP_CLAUSE_DEPEND_DEPOBJ.
5261 (struct tree_omp_clause): Add subcode.defaultmap_kind.
5262 * tree.def (OMP_TASKGROUP): Add another operand, move next to other
5263 OpenMP constructs with body and clauses operands.
5264 * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
5265 (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
5266 (OMP_TASKGROUP_CLAUSES): Define.
5267 (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
5268 OMP_CLAUSE__LOOPTEMP_.
5269 (OMP_ATOMIC_SEQ_CST): Remove.
5270 (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
5271 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
5272 (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
5273 OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
5274 OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
5275 OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
5276 OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
5277 (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
5278 OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
5279 OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
5281 * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
5282 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5283 OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
5284 (convert_local_omp_clauses): Likewise. Remove useless test.
5285 * tree-parloops.c (create_call_for_reduction_1): Pass
5286 OMP_MEMORY_ORDER_RELAXED as new argument to
5287 dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
5288 * tree-pretty-print.c (dump_omp_iterators): New function.
5289 (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
5290 OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION. Print
5291 reduction modifiers. Handle OMP_CLAUSE_DEPEND_DEPOBJ and
5292 OMP_CLAUSE_DEPEND_MUTEXINOUTSET. Print iterators in depend clauses.
5293 Print __internal__ for OMP_CLAUSE_DEPEND_LAST. Handle cancel and
5294 simd OMP_CLAUSE_IF_MODIFIERs. Handle new kinds of
5295 OMP_CLAUSE_DEFAULTMAP. Print conditional: for
5296 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
5297 (dump_omp_atomic_memory_order): New function.
5298 (dump_generic_node): Use it. Print taskgroup clauses. Print
5299 taskwait with depend clauses.
5300 * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
5301 * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
5302 Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
5303 * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
5304 write_ts_omp_clause_tree_pointers): Likewise.
5306 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5309 * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
5310 "-missed", "-note", and "-all" sub-options.
5311 * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
5313 (can_early_inline_edge_p): Likewise.
5314 (want_early_inline_function_p): Likewise.
5315 (want_inline_self_recursive_call_p): Likewise.
5316 (recursive_inlining): Likewise.
5317 (inline_small_functions): Likewise.
5318 (flatten_function): Likewise.
5319 (ipa_inline): Likewise.
5320 (inline_always_inline_functions): Likewise.
5321 (early_inline_small_functions): Likewise.
5322 (early_inliner): Likewise.
5323 * tree-inline.c (expand_call_inline): Likewise.
5325 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5327 * pretty-print.c (pp_format): Handle %f.
5328 (selftest::test_pp_format): Add test of %f.
5329 * pretty-print.h (pp_double): New macro.
5331 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5333 * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
5334 * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
5335 (dump_pretty_printer::decode_format): Implement "%C" for
5337 (selftest::test_capture_of_dump_calls): Rename "where" to
5338 "stmt_loc". Convert test_decl to a function decl and set its
5339 location. Add a symbol_table_test RAII instance and a
5340 cgraph_node, using it to test "%C" and dump_symtab_node.
5342 2018-11-08 Eric Botcazou <ebotcazou@adacore.com>
5345 * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
5347 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5349 * cgraph.c: Include "selftest.h".
5350 (saved_symtab): New variable.
5351 (selftest::symbol_table_test::symbol_table_test): New ctor.
5352 (selftest::symbol_table_test::~symbol_table_test): New dtor.
5353 (selftest::test_symbol_table_test): New test.
5354 (selftest::cgraph_c_tests): New.
5355 * cgraph.h (saved_symtab): New decl.
5356 (selftest::symbol_table_test): New class.
5357 * selftest-run-tests.c (selftest::run_tests): Call
5358 selftest::cgraph_c_tests.
5359 * selftest.h (selftest::cgraph_c_tests): New decl.
5361 2018-11-08 Richard Biener <rguenther@suse.de>
5363 * tree-data-ref.h (lambda_int): New typedef.
5364 (lambda_vector_gcd): Adjust.
5365 (lambda_vector_new): Likewise.
5366 (lambda_matrix_new): Likewise.
5367 * tree-data-ref.c (print_lambda_vector): Adjust.
5369 2018-11-08 Richard Biener <rguenther@suse.de>
5371 PR tree-optimization/87929
5372 * tree-complex.c (expand_complex_comparison): Clean EH.
5374 2018-11-08 Martin Liska <mliska@suse.cz>
5376 * doc/extend.texi: Reword.
5377 * predict.c (expr_expected_value_1): Likewise.
5379 2018-11-08 Richard Biener <rguenther@suse.de>
5381 PR tree-optimization/87913
5382 * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
5383 of extreme values to ordered comparisons.
5385 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
5388 * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
5390 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
5392 * doc/invoke.texi: Remove leading dash from @opindex entries
5393 throughout the file.
5395 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
5398 * doc/invoke.texi (Option Summary): Add -e and --entry.
5399 (Link Options): Likewise.
5401 2018-11-07 Nathan Sidwell <nathan@acm.org>
5404 * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
5405 --disable-checking bootstrap.
5407 2018-11-07 Hafiz Abid Qadeer <abidh@codesourcery.com>
5409 * configure: Regenerated.
5411 2018-11-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5414 * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
5415 to the mode of the widest field iff the widest field has mode class
5416 MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
5419 2018-11-07 Nikolai Merinov <n.merinov@inango-systems.com>
5421 * common.opt: Add -Wattribute-warning.
5422 * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
5423 * expr.c (expand_expr_real_1): Add new attribute to warning_at
5424 call to allow user configure behavior of "warning" attribute.
5426 2018-11-07 Segher Boessenkool <segher@kernel.crashing.org>
5428 * target.def: Put @: after every vs., e.g., and i.e. where it is
5429 followed by whitespace.
5430 * doc/extend.texi: Ditto.
5431 * doc/fragments.texi: Ditto.
5432 * doc/gimple.texi: Ditto.
5433 * doc/implement-c.texi: Ditto.
5434 * doc/install.texi: Ditto.
5435 * doc/invoke.texi: Ditto.
5436 * doc/md.texi: Ditto.
5437 * doc/plugins.texi: Ditto.
5438 * doc/rtl.texi: Ditto.
5439 * doc/sourcebuild.texi: Ditto.
5440 * doc/tm.texi.in: Ditto.
5441 * doc/ux.texi: Ditto.
5442 * doc/tm.texi: Regenerate.
5444 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5446 * config/arm/arm-cpus.in (ares): New entry.
5447 * config/arm/arm-tables.opt: Regenerate.
5448 * config/arm/arm-tune.md: Likewise.
5449 * doc/invoke.texi (ARM Options): Document ares.
5451 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5453 * config/aarch64/aarch64-cores.def (ares): Define.
5454 * config/aarch64/aarch64-tune.md: Regenerate.
5455 * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
5457 2018-11-07 Jan Hubicka <jh@suse.cz>
5459 * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
5460 (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
5461 functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
5464 2018-11-07 Richard Biener <rguenther@suse.de>
5466 PR tree-optimization/87914
5467 * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
5469 (vectorizable_reduction): Handle shifts and rotates by dispatching
5470 to vectorizable_shift.
5471 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
5472 in-loop uses of vect_nested_cycle defs. Merge cycle and internal
5474 (vectorizable_shift): Export and handle being called as
5476 (vect_analyze_stmt): Call vectorizable_shift after
5477 vectorizable_reduction.
5478 * tree-vectorizer.h (vectorizable_shift): Declare.
5480 2018-11-07 Jan Hubicka <jh@suse.cz>
5482 * ipa-devirt.c (odr_types_equivalent_p): Expect constants
5483 than const decls in TREE_VALUE of enum.
5484 (dump_type_inheritance_graph): Improve duplicate dumping.
5485 (free_enum_values): New.
5486 (build_type_inheritance_graph): Use it.
5487 * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
5488 which are not main variants or not ODR types.
5489 (verify_type_variant): Expect variants to have no TYPE_VALUES.
5491 2018-11-07 Richard Biener <rguenther@suse.de>
5493 * ipa-inline.c (want_inline_small_function_p): Compute
5494 big_speedup_p lazily and last.
5496 2018-11-07 Jan Hubicka <jh@suse.cz>
5498 * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
5499 building incomplete variant of complete type.
5500 (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
5501 variant of complete type.
5503 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5505 * config/mips/mips.c: Fix typo in documentation of
5506 mips_loongson_ext2_prefetch_cookie.
5507 (mips_option_override): fix brain twister logical.
5508 * config/mips/mips.h: Fix typo in documentation of
5509 ISA_HAS_CTZ_CTO and define pattern.
5510 * config/mips/mips.md (prefetch): Hoist EXT2 above
5512 (prefetch_indexed): Hoist EXT2 above the EXT block.
5514 2018-11-07 Jan Hubicka <jh@suse.cz>
5516 * tree.c (free_lang_data_in_type): Add fld parameter; simplify
5517 return and parameter types of function and method types.
5518 (free_lang_data_in_cgraph): Update.
5520 2018-11-07 Martin Liska <mliska@suse.cz>
5522 PR rtl-optimization/87868
5523 * postreload-gcse.c (eliminate_partially_redundant_load): Set
5524 threshold to max_count if we would overflow.
5525 * profile-count.h: Make max_count a public constant.
5527 2018-11-07 Martin Liska <mliska@suse.cz>
5529 * mem-stats.h: Fix GNU coding style.
5531 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5533 * config/mips/gs264e.md: New.
5534 * config/mips/mips-cpus.def: Define gs264e.
5535 * config/mips/mips-tables.opt: Regenerate.
5536 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
5538 (mips_issue_rate): Add support for gs264e.
5539 (mips_multipass_dfa_lookahead): Likewise.
5540 * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
5541 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
5542 (MIPS_ASE_MSA_SPEC): New.
5543 (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
5544 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
5545 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
5546 * config/mips/mips.md: Include gs264e.md.
5547 (processor): Add gs264e.
5548 * config/mips/mips.opt (MSA): Use Mask instead of Var.
5549 * doc/invoke.texi: Add gs264e to supported architectures.
5551 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5553 * config/mips/gs464e.md: New.
5554 * config/mips/mips-cpus.def: Define gs464e.
5555 * config/mips/mips-tables.opt: Regenerate.
5556 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
5558 (mips_issue_rate): Add support for gs464e.
5559 (mips_multipass_dfa_lookahead): Likewise.
5560 (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
5561 * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
5562 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
5563 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
5564 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
5565 * config/mips/mips.md: Include gs464e.md.
5566 (processor): Add gs464e.
5567 * doc/invoke.texi: Add gs464e to supported architectures.
5569 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5571 * config/mips/loongson3a.md: Rename to ...
5572 * config/mips/gs464.md: ... here.
5573 * config/mips/mips-cpus.def: Define gs464; Add loongson3a
5574 as an alias of gs464 processor.
5575 * config/mips/mips-tables.opt: Regenerate.
5576 * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
5577 instead of PROCESSOR_LOONGSON_3A.
5578 (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
5580 (mips_option_override): Enable MMI and EXT for gs464.
5581 * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
5582 Rename TUNE_LOONGSON_3A to TUNE_GS464.
5583 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
5584 (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
5585 ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
5587 * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
5588 (processor): Add gs464;
5589 * doc/invoke.texi: Add gs464 to supported architectures.
5591 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5593 * config/mips/mips-protos.h
5594 (mips_loongson_ext2_prefetch_cookie): New prototype.
5595 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
5596 (mips_option_override): Enable TARGET_LOONGSON_EXT when
5597 TARGET_LOONGSON_EXT2 is true.
5598 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
5599 __mips_loongson_ext2, __mips_loongson_ext_rev=2.
5600 (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
5601 (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
5602 TARGET_LOONGSON_EXT2.
5603 (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
5604 (define_insn "ctz<mode>2"): New insn pattern.
5605 (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
5606 (define_insn "prefetch_indexed_<mode>"): Include
5607 TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
5608 * config/mips/mips.opt (-mloongson-ext2): Add option.
5609 * gcc/doc/invoke.texi (-mloongson-ext2): Document.
5611 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5613 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
5614 __mips_loongson_ext.
5615 (MIPS_ASE_LOONGSON_EXT_SPEC): New.
5616 (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
5618 (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
5619 * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
5620 <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
5621 instead of TARGET_LOONGSON_3A.
5622 * config/mips/mips.opt (-mloongson-ext): Add option.
5623 * gcc/doc/invoke.texi (-mloongson-ext): Document.
5625 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5627 * config.gcc (extra_headers): Add loongson-mmiintrin.h.
5628 * config/mips/loongson.md: Move to ...
5629 * config/mips/loongson-mmi.md: here; Adjustment.
5630 * config/mips/loongson.h: Move to ...
5631 State as deprecated. Include loongson-mmiintrin.h for back
5632 compatibility and warning.
5633 * config/mips/loongson-mmiintrin.h: ... here.
5634 * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
5635 mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
5636 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5637 (mips_option_override): Make sure MMI use hard float;
5638 (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
5639 mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
5640 mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
5641 TARGET_LOONGSON_VECTORS.
5642 * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
5643 (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
5644 (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
5645 (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
5647 (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
5648 TARGET_LOONGSON_VECTORS.
5649 * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
5650 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5651 (Loongson MMI patterns): Include loongson-mmi.md instead of
5653 * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
5654 * gcc/doc/invoke.texi (-mloongson-mmi): Document.
5656 2018-11-07 Richard Biener <rguenther@suse.de>
5659 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
5660 BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
5662 2018-11-07 Alexandre Oliva <aoliva@redhat.com>
5664 PR rtl-optimization/87874
5665 * lra.c (lra_substitute_pseudo): Do not create a subreg for
5668 2018-11-06 Aaron Sawdey <acsawdey@linux.ibm.com>
5670 * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
5671 if not in indexed or indirect form.
5672 (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
5673 (bswap<mode>2_store): Ditto.
5675 2018-11-06 Richard Earnshaw <rearnsha@arm.com>
5677 * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
5680 2018-11-06 Richard Biener <rguenther@suse.de>
5682 PR tree-optimization/86850
5683 * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
5684 instead of src.m_vec.
5686 2018-11-06 Jan Hubicka <jh@suse.cz>
5688 * tree.c (fld_simplified_type_name): Break out form ...
5689 (free_lang_data_in_type): ... here.
5690 (fld_type_variant_equal_p): Use it.
5692 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5694 * config/default-d.c: Include memmodel.h.
5696 * config/sol2-d.c: New file.
5697 * config/t-sol2 (sol2-d.o): New rule.
5698 * config.gcc <*-*-solaris2*>: Set d_target_objs,
5699 target_has_targetdm.
5701 2018-11-06 Jan Hubicka <jh@suse.cz>
5703 * tree.c (fld_type_variant): Also copy alignment; be sure that
5704 new variant is equal.
5706 2018-11-06 Ilya Leoshkevich <iii@linux.ibm.com>
5709 * config/s390/s390.md: Add relative_long attribute.
5711 2018-11-06 Jan Hubicka <jh@suse.cz>
5713 * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
5714 * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
5715 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
5716 stream TYPE_NEEDS_CONSTRUCTING.
5717 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
5718 * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
5720 2018-11-06 Richard Biener <rguenther@suse.de>
5722 * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
5724 (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
5726 2018-11-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5728 * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
5729 REG_CLASS_CONTENTS[GEN_REGS].
5730 (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
5732 2018-11-06 Jan Hubicka <jh@suse.cz>
5734 * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
5736 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5739 * config/sol2.h (ASAN_CC1_SPEC): Define.
5740 (LD_WHOLE_ARCHIVE_OPTION): Define.
5741 (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
5742 (ASAN_REJECT_SPEC): Provide default.
5743 (LIBASAN_EARLY_SPEC): Define.
5744 (LIBTSAN_EARLY_SPEC): Define.
5745 (LIBLSAN_EARLY_SPEC): Define.
5746 * config/i386/sol2.h (CC1_SPEC): Redefine.
5747 (ASAN_REJECT_SPEC): Define.
5749 * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
5750 (TARGET_ASAN_SHADOW_OFFSET): Define.
5751 (sparc_asan_shadow_offset): New function.
5752 * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
5753 (ASAN_REJECT_SPEC): Define.
5755 2018-11-06 Jan Hubicka <jh@suse.cz>
5757 * tree.c (fld_type_variant): Copy canonical type.
5758 (fld_incomplete_type_of): Check that canonical types looks sane;
5759 copy canonical type.
5760 (verify_type): Accept when incomplete type has complete canonical type.
5762 2018-11-06 Jan Hubicka <jh@suse.cz>
5764 * tree.c (free_lang_data): Reset overwite_assembler_name,
5765 print_xnode, print_decl, print_type and print_identifier of
5768 2018-11-06 Richard Biener <rguenther@suse.de>
5770 PR tree-optimization/87889
5771 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
5772 Do nothing if old and new arg are the same
5774 2018-11-06 Andreas Krebbel <krebbel@linux.ibm.com>
5777 * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
5778 attributes for operands 3 and 4.
5780 2018-11-06 Richard Biener <rguenther@suse.de>
5783 * simplify-rtx.c (simplify_binary_operation_1): Add pattern
5784 matching bitfield insertion.
5786 2018-11-06 Alexandre Oliva <aoliva@redhat.com>
5788 * auto-inc-dec.c: Include valtrack.h. Improve comments.
5789 (reg_next_debug_use): New.
5790 (attempt_change): Propagate adjusted expression into affected
5792 (merge_in_block): Track uses in debug insns.
5793 (pass_inc_dec::execute): Allocate and release
5796 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
5798 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
5799 (_mm512_fixupimm_round_pd): Update parameters and builtin.
5800 (_mm512_maskz_fixupimm_round_pd): Ditto.
5801 (_mm512_fixupimm_round_ps): Ditto.
5802 (_mm512_maskz_fixupimm_round_ps): Ditto.
5803 (_mm_fixupimm_round_sd): Ditto.
5804 (_mm_maskz_fixupimm_round_sd): Ditto.
5805 (_mm_fixupimm_round_ss): Ditto.
5806 (_mm_maskz_fixupimm_round_ss): Ditto.
5807 (_mm512_fixupimm_pd): Ditto.
5808 (_mm512_maskz_fixupimm_pd): Ditto.
5809 (_mm512_fixupimm_ps): Ditto.
5810 (_mm512_maskz_fixupimm_ps): Ditto.
5811 (_mm_fixupimm_sd): Ditto.
5812 (_mm_maskz_fixupimm_sd): Ditto.
5813 (_mm_fixupimm_ss): Ditto.
5814 (_mm_maskz_fixupimm_ss): Ditto.
5815 (_mm512_mask_fixupimm_round_pd): Update builtin.
5816 (_mm512_mask_fixupimm_round_ps): Ditto.
5817 (_mm_mask_fixupimm_round_sd): Ditto.
5818 (_mm_mask_fixupimm_round_ss): Ditto.
5819 (_mm512_mask_fixupimm_pd): Ditto.
5820 (_mm512_mask_fixupimm_ps): Ditto.
5821 (_mm_mask_fixupimm_sd): Ditto.
5822 (_mm_mask_fixupimm_ss): Ditto.
5823 * config/i386/avx512vlintrin.h:
5824 (_mm256_fixupimm_pd): Update parameters and builtin.
5825 (_mm256_maskz_fixupimm_pd): Ditto.
5826 (_mm256_fixupimm_ps): Ditto.
5827 (_mm256_maskz_fixupimm_ps): Ditto.
5828 (_mm_fixupimm_pd): Ditto.
5829 (_mm_maskz_fixupimm_pd): Ditto.
5830 (_mm_fixupimm_ps): Ditto.
5831 (_mm_maskz_fixupimm_ps): Ditto.
5832 (_mm256_mask_fixupimm_pd): Update builtin.
5833 (_mm256_mask_fixupimm_ps): Ditto.
5834 (_mm_mask_fixupimm_pd): Ditto.
5835 (_mm_mask_fixupimm_ps): Ditto.
5836 * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
5837 * config/i386/i386-builtin.def: Update builtin definitions.
5838 * config/i386/i386.c: Handle new builtin types and remove useless ones.
5839 * config/i386/sse.md: Update VFIXUPIMM* patterns.
5840 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
5841 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
5842 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
5843 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
5844 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
5845 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
5846 * config/i386/subst.md:
5847 (round_saeonly_sd_mask_operand4): Add new subst_attr.
5848 (round_saeonly_sd_mask_op4): Ditto.
5849 (round_saeonly_expand_operand5): Ditto.
5850 (round_saeonly_expand): Update.
5852 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
5854 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
5856 2018-11-05 Segher Boessenkool <segher@kernel.crashing.org>
5858 PR rtl-optimization/87871
5859 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
5861 2018-11-05 Paul Koning <ni1d@arrl.net>
5863 * doc/sourcebuild.texi (target attributes): Document new "inf"
5864 effective target keyword.
5866 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
5868 * config/s390/s390.c (s390_register_move_cost): Increase costs for
5869 moves involving the CC reg.
5871 2018-11-05 Richard Biener <rguenther@suse.de>
5873 PR tree-optimization/87873
5874 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
5876 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
5877 * tree-vect-loop.c (vect_transform_loop): When splitting the
5878 loop exit also create forwarder PHIs for constants.
5879 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
5880 Handle constant to_arg, add extra checking we match up the correct
5883 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
5885 * config/s390/s390.md: QImode and HImode for load on condition.
5887 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
5889 * config/s390/predicates.md: Fix typo.
5890 * config/s390/s390.md: Allow immediates for load on condition.
5892 2018-11-05 Martin Liska <mliska@suse.cz>
5894 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
5895 * bitmap.h (struct bitmap_usage): Likewise.
5896 * ggc-common.c (SCALE): Remove.
5898 (struct ggc_usage): Use SIZE_AMOUNT. And update
5900 * ggc-page.c (SCALE): Remove.
5901 (STAT_LABEL): Remove.
5902 (ggc_print_statistics): Use SIZE_AMOUNT.
5903 * gimple.h (SCALE): Remove.
5905 * input.c (ONE_K): Remove.
5908 (STAT_LABEL): Likewise.
5909 (FORMAT_AMOUNT): Likewise.
5910 (dump_line_table_statistics): Use SIZE_AMOUNT.
5911 * mem-stats.h (struct mem_usage): Likewise.
5912 * rtl.c (dump_rtx_statistics): Likewise.
5913 (rtx_alloc_counts): Change type to size_t.
5914 (rtx_alloc_sizes): Likewise.
5915 (rtx_count_cmp): New.
5916 (dump_rtx_statistics): Sort first based on counts.
5917 * tree.c (tree_nodes_cmp): New.
5918 (tree_codes_cmp): New.
5919 (dump_tree_statistics): Sort first based on counts.
5920 * system.h (ONE_K): New.
5922 (SIZE_SCALE): Likewise.
5923 (SIZE_LABEL): Likewise.
5924 (SIZE_AMOUNT): Likewise.
5925 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
5926 * tree-dfa.c (dump_dfa_stats): Likewise.
5927 * tree-phinodes.c (phinodes_print_statistics): Likewise.
5928 * tree-ssanames.c (ssanames_print_statistics): Likewise.
5929 * tree.c (dump_tree_statistics): Likewise.
5930 * vec.c (struct vec_usage): Likewise.
5931 * trans-mem.c (tm_mangle): Enlarge buffer in order to not
5932 trigger a -Werror=format-overflow with
5933 --enable-gather-detailed-stats.
5935 2018-11-05 Martin Liska <mliska@suse.cz>
5937 * mem-stats.h (mem_alloc_description::release_instance_overhead):
5939 * vec.c (struct vec_usage): Register m_element_size.
5940 (vec_prefix::register_overhead): New arguments: elements and
5942 (vec_prefix::release_overhead): Subtract elements.
5943 * vec.h (struct vec_prefix): Change signature.
5944 (va_heap::reserve): Pass proper arguments.
5945 (va_heap::release): Likewise.
5947 2018-11-05 Martin Liska <mliska@suse.cz>
5949 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
5953 2018-11-05 Richard Biener <rguenther@suse.de>
5955 * tree-scalar-evolution.h (final_value_replacement_loop): Update
5957 * tree-scalar-evolution.c (final_value_replacement_loop): Return
5958 whether anything was done.
5959 (scev_const_prop): Remove constant propagation part, fold
5961 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
5962 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
5965 2018-11-05 Jakub Jelinek <jakub@redhat.com>
5967 PR tree-optimization/87859
5968 * gimple-ssa-store-merging.c (struct merged_store_group): Add
5969 only_constants and first_nonmergeable_order members.
5970 (merged_store_group::merged_store_group): Initialize them.
5971 (merged_store_group::do_merge): Clear only_constants member if
5972 adding something other than INTEGER_CST store.
5973 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
5974 stores with order >= first_nonmergeable_order. Use
5975 merged_store->only_constants instead of always recomputing it.
5976 Set merged_store->first_nonmergeable_order if we've skipped any
5977 stores. Attempt to merge overlapping INTEGER_CST stores that
5978 we would otherwise skip.
5981 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
5983 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
5986 * config/i386/emmintrin.h (__v16qs): New to cope with option
5988 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
5989 (_mm_cmplt_epi8): Likewise.
5990 (_mm_cmpgt_epi8): Likewise.
5992 2018-11-05 Richard Biener <rguenther@suse.de>
5994 PR rtl-optimization/87852
5995 * fwprop.c (use_killed_between): Only consider single-defs of the
5996 use whose definition statement dominates the use.
5998 2018-11-05 Martin Liska <mliska@suse.cz>
6001 * doc/invoke.texi: Remove options that are
6002 not disabled with -Os.
6004 2018-11-05 Martin Liska <mliska@suse.cz>
6007 * doc/extend.texi: Update constrain about the last argument
6008 of __builtin_expect_with_probability.
6010 2018-11-05 Martin Liska <mliska@suse.cz>
6013 * predict.c (expr_expected_value_1): Verify
6014 that last argument is a real constants and emit
6017 2018-11-05 Martin Liska <mliska@suse.cz>
6019 PR gcov-profile/77698
6020 * ipa-profile.c (ipa_profile): Adjust hotness threshold
6023 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
6025 PR tree-optimization/86572
6026 * builtins.c (c_strlen): Handle negative offsets in a safe way.
6028 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
6030 PR tree-optimization/87672
6031 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
6032 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
6034 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
6037 * cfgexpand.c (pass_expand::execute): Move the call to
6038 finish_eh_generation in front of the call to expand_stack_alignment.
6040 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
6042 * common/config/i386/i386-common.c (processor_alias_table): Add
6044 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
6045 (case ${target}): Add znver2.
6046 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
6047 -march=native recognize znver2 processors.
6048 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
6049 * config/i386/i386.c (m_znver2): New definition.
6050 (m_ZNVER): New definition.
6051 (m_AMD_MULTIPLE): Includes m_znver2.
6052 (processor_cost_table): Add znver2 entry.
6053 (processor_target_table): Add znver2 entry.
6054 (get_builtin_code_for_version): Set priority for
6056 (processor_model): Add M_AMDFAM17H_ZNVER2.
6057 (arch_names_table): Ditto.
6058 (ix86_reassociation_width): Include znver2.
6059 * config/i386/i386.h (TARGET_znver2): New definition.
6060 (struct ix86_size_cost): Add TARGET_ZNVER2.
6061 (enum processor_type): Add PROCESSOR_ZNVER2.
6062 * config/i386/i386.md (define_attr "cpu"): Add znver2.
6063 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
6064 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
6065 (ix86_adjust_cost): Add znver2.
6066 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
6067 * gcc/doc/extend.texi: Add details about znver2.
6068 * gcc/doc/invoke.texi: Add details about znver2.
6070 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
6074 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
6077 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
6079 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
6080 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
6082 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
6084 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
6085 shrink-wrapping when -mtrack-speculation.
6087 2018-11-02 Richard Biener <rguenther@suse.de>
6089 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
6090 (pop_cost_one_pair): Do not free pair.
6091 (pop_best_coalesce): Likewise.
6092 (create_coalesce_list): Initialize obstack.
6093 (delete_coalesce_list): Free obstack.
6094 (find_coalesce_pair): Obstack-allocate coalesce pairs.
6095 (add_cost_one_coalesce): Likewise.
6096 (struct live_track): Remove bitmap pointer indirections.
6097 (new_live_track): Adjust.
6098 (delete_live_track): Likewise.
6099 (live_track_remove_partition): Likewise.
6100 (live_track_add_partition): Likewise.
6101 (live_track_live_p): Likewise.
6102 (live_track_process_def): Likewise.
6103 (live_track_clear_base_vars): Likewise.
6105 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6107 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
6108 * configure: Regenerate.
6109 * config.in: Regenerate.
6110 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
6111 HAVE_AS_REGISTER_PSEUDO_OP guard.
6112 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
6113 !HAVE_AS_REGISTER_PSEUDO_OP support.
6114 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
6117 2018-11-02 Richard Biener <rguenther@suse.de>
6119 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
6122 2018-11-02 Richard Biener <rguenther@suse.de>
6124 PR tree-optimization/87776
6125 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
6126 executable when iterating but running into rpo-vn-max-loop-depth
6127 and not eliding the iteration.
6129 2018-11-30 Jan Hubicka <jh@suse.cz>
6131 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
6134 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
6136 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
6138 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
6139 Combine with rs6000_address_for_fpconvert.
6140 (rs6000_address_for_fpconvert): Combine with
6141 rs6000_force_indexed_or_indirect_mem.
6142 (rs6000_expand_vector_init): Change function call from
6143 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
6144 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
6145 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
6146 (floatsi<mode>2_lfiwax_mem): Ditto.
6147 (floatunssi<mode>2_lfiwzx): Ditto.
6148 (floatunssi<mode>2_lfiwzx_mem): Ditto.
6149 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
6150 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
6151 (fix_trunc<mode>si2_stfiwx): Ditto.
6152 (fixuns_trunc<mode>si2_stfiwx): Ditto.
6153 (float_<mode>si2_hw): Ditto.
6154 (floatuns_<mode>si2_hw): Ditto.
6155 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
6156 (vsx_splat_<mode>): Ditto.
6158 2018-11-01 Joseph Myers <joseph@codesourcery.com>
6160 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
6161 at end of assembler input text.
6162 * configure: Regenerate.
6164 2018-11-01 Jakub Jelinek <jakub@redhat.com>
6166 PR tree-optimization/87826
6167 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
6168 negative or larger or equal to type's precision.
6170 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
6172 * opts.c (default_options_table): Do not enable
6173 OPT_fdelayed_branch at -Og.
6174 * doc/invoke.texi (-fdelayed-branch): Document it.
6176 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
6178 * optabs-libfuncs.c (build_libfunc_function_visibility):
6179 New, split out from...
6180 (build_libfunc_function): ... here.
6181 (init_one_libfunc_visibility): New, split out from ...
6182 (init_one_libfunc): ... here.
6184 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
6185 scratch register need not be early-clobber. Document the reason
6186 why we cannot use ST<OP>.
6188 2018-10-31 Joseph Myers <joseph@codesourcery.com>
6191 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
6192 line for second argument of AC_DEFINE_UNQUOTED.
6193 * doc/install.texi (Tools/packages necessary for modifying GCC):
6194 Update to autoconf 2.69 and automake 1.15.1.
6195 * aclocal.m4, config.in, configure: Regenerate.
6197 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
6199 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
6202 2018-10-31 Martin Liska <mliska@suse.cz>
6205 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
6206 Add new argument invalid_extension.
6207 (aarch64_get_all_extension_candidates): New function.
6208 (aarch64_rewrite_selected_cpu): Add NULL to function call.
6209 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
6211 (aarch64_get_all_extension_candidates): New function.
6212 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
6213 argument invalid_extension.
6214 (aarch64_parse_cpu): Likewise.
6215 (aarch64_print_hint_for_extensions): New function.
6216 (aarch64_validate_mcpu): Provide hint about invalid extension.
6217 (aarch64_validate_march): Likewise.
6218 (aarch64_handle_attr_arch): Pass new argument.
6219 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
6220 (aarch64_handle_attr_isa_flags): Likewise.
6222 2018-10-31 Richard Biener <rguenther@suse.de>
6226 * tree-outof-ssa.c (insert_backedge_copies): Restrict
6227 copy generation to useful cases. Place the copy before
6228 the definition of the backedge value when possible.
6230 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6232 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
6233 * config/arc/arc.c (arc_active_insn): New function.
6234 (check_store_cacheline_hazard): Likewise.
6235 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
6236 (arc_override_options): Disable delay slot scheduler for older
6238 (arc_store_addr_hazard_p): New implementation, old one renamed to
6240 (arc_store_addr_hazard_internal_p): Renamed.
6241 (arc_reorg): Don't combine into brcc instructions which are part
6242 of hardware hazard solution.
6243 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
6244 (tune_arc700): Likewise.
6245 * config/arc/arc.opt (arc7xx): New tune value.
6246 * config/arc/arc700.md: Improve A7 scheduler.
6248 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6250 * config/arc/arc.c (arc_override_options): Remove
6251 TARGET_COMPACT_CASESI.
6252 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
6253 (CASE_VECTOR_MODE): Likewise.
6254 (CASE_VECTOR_PC_RELATIVE): Likewise.
6255 (CASE_VECTOR_SHORTEN_MODE): Likewise.
6256 (CASE_VECTOR_SHORTEN_MODE1): Delete.
6257 (ADDR_VEC_ALIGN): Update.
6258 (ASM_OUTPUT_CASE_LABEL): Undefine.
6259 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
6260 (TARGET_BI_BIH): Define.
6261 (DEFAULT_BRANCH_INDEX): Likewise.
6262 * config/arc/arc.md (casesi): Rework to accept BI/BIH
6263 instructions, remove compact_casesi use case.
6264 (casesi_compact_jump): Remove.
6265 (casesi_dispatch): New pattern.
6266 * config/arc/arc.opt: Add mbranch-index option. Deprecate
6267 compact_casesi option.
6268 * doc/invoke.texi: Document mbranch-index option.
6270 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6272 * config/arc/arc.c (arc_get_tp): Remove function.
6273 (arc_emit_call_tls_get_addr): Likewise.
6274 (arc_call_tls_get_addr): New function.
6275 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
6276 * config/arc/arc.md (tls_load_tp_soft): Remove.
6277 (tls_gd_get_addr): Likewise.
6279 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6281 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
6282 (mulsi3_600_lib): Remove pattern.
6283 (umulsi3_highpart_600_lib_le): Likewise.
6284 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
6285 (umulsidi3): Remove call to umulsidi3_600_lib.
6286 (umulsidi3_600_lib): Remove pattern.
6287 (peephole2): Remove peephole using the above deprecated patterns.
6289 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
6292 * config/arm/arm.c (arm_option_check_internal): Disable the combined
6293 use of -mslow-flash-data and -mword-relocations.
6294 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
6295 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
6297 * doc/invoke.texi (-mword-relocations): Mention conflict with
6299 (-mslow-flash-data): Reciprocally.
6301 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
6303 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
6304 16-byte modes held in GP registers to use an even regno.
6306 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
6307 (aarch64_atomic_ldop_supported_p): Remove.
6308 (aarch64_gen_atomic_ldop): Remove.
6309 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
6310 Fully expand LSE operations here.
6311 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
6312 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
6313 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
6314 and use ATOMIC_LDOP instead; use register_operand for the input;
6315 drop the split and emit insns directly.
6316 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
6317 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
6318 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
6320 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
6321 (aarch64_gen_atomic_ldop): Don't call it.
6322 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
6323 Use aarch64_reg_or_zero.
6324 (aarch64_atomic_exchange<ALLI>): Likewise.
6325 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
6326 operand 0; use aarch64_reg_or_zero for input; merge ...
6327 (@aarch64_atomic_swp<ALLI>): ... this and remove.
6329 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
6330 (aarch64_split_compare_and_swap): Use it.
6331 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
6332 test oldval against the proper predicate.
6333 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
6334 Use nonmemory_operand for expected.
6335 (cas_short_expected_pred): New.
6336 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
6337 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
6338 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
6339 (aarch64_plushi_operand): New.
6341 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
6342 Force oldval into the rval register for TARGET_LSE; emit the compare
6343 during initial expansion so that it may be deleted if unused.
6344 (aarch64_gen_atomic_cas): Remove.
6345 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
6346 Change =&r to +r for operand 0; use match_dup for operand 2;
6347 remove is_weak and mod_f operands as unused. Drop the split
6349 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
6350 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
6351 (@aarch64_atomic_cas<GPI>): Similarly.
6353 2018-10-31 Richard Biener <rguenther@suse.de>
6355 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
6358 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6360 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
6361 * config/aarch64/aarch64.md: Include saphira.md
6362 * config/aarch64/saphira.md: New file for pipeline description.
6364 2018-10-30 Martin Sebor <msebor@redhat.com>
6367 * gimple-ssa-sprintf.c (format_directive): Use %G to include
6369 (sprintf_dom_walker::compute_format_length):
6370 Avoid setting POSUNDER4K here.
6371 (get_destination_size): Handle null argument values.
6372 (get_user_idx_format): New function.
6373 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
6374 functions, including user-defined with attribute format printf.
6375 Use %G to include inlining context.
6376 Set POSUNDER4K here.
6378 2018-10-30 Jan Hubicka <jh@suse.cz>
6380 * params.def (lto-partitions): Bump from 32 to 128.
6382 2018-10-30 Jan Hubicka <jh@suse.cz>
6385 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
6387 (free_lang_data_in_type): Forward declare.
6388 (fld_type_variant_equal_p): New function.
6389 (fld_type_variant): New function
6390 (fld_incomplete_types): New hash.
6391 (fld_incomplete_type_of): New function
6392 (fld_simplfied-type): New function.
6393 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
6394 (free_lang_data): Allocate and free fld_incomplete_type; update call
6395 of free_lang_data_in_decl.
6397 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
6399 * gcov.c (output_lines): Remove duplicate line.
6401 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
6403 * config/rs6000/rs6000.md (bswapdi2): Force address into register
6404 if not in indexed or indirect form.
6405 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
6406 (bswapdi2_store): Ditto.
6407 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
6409 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
6410 Prototype for helper function.
6412 2018-10-30 Martin Sebor <msebor@redhat.com>
6414 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
6415 (target, pragma GCC optimize, pragma GCC target): Ditto.
6417 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
6419 * doc/extend.texi: Fix prototype and description of
6420 __builtin_expect_with_probability.
6422 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
6424 * cgraph.h (clone_function_name_1): Replaced by new
6425 clone_function_name_numbered that takes name as string; for
6426 privatize_symbol_name_1 use only.
6427 (clone_function_name): Renamed to
6428 clone_function_name_numbered to be explicit about numbering.
6429 (clone_function_name): New two-argument function that does
6430 not number its output.
6431 (clone_function_name): New three-argument function that
6432 takes a number to append to its output.
6433 * cgraphclones.c (duplicate_thunk_for_node):
6434 (clone_function_name_1): Renamed.
6435 (clone_function_name_numbered): Two new functions.
6436 (clone_function_name): Improved documentation.
6437 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
6438 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
6439 * final.c (final_scan_insn_1): Use the new clone_function_name
6441 * multiple_target.c (create_dispatcher_calls): Ditto.
6442 (create_target_clone): Ditto.
6443 * omp-expand.c (grid_expand_target_grid_body): Ditto.
6444 * omp-low.c (create_omp_child_function_name): Ditto.
6445 * omp-simd-clone.c (simd_clone_create): Ditto.
6446 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
6447 new clone_function_name without numbering.
6449 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
6451 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
6452 Assert that the allocation size is not zero.
6454 2018-10-30 Richard Biener <rguenther@suse.de>
6456 PR tree-optimization/87800
6457 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
6458 non-induction or reduction PHIs.
6460 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org>
6462 * config/aarch64/falkor-tag-collision-avoidance.c
6463 (execute_tag_collision_avoidance): Call df_note_add_problem.
6465 2018-10-30 Martin Liska <mliska@suse.cz>
6467 * doc/extend.texi: Fix typo in documentation
6468 of __builtin_expect_with_probability.
6470 2018-10-29 David Malcolm <dmalcolm@redhat.com>
6473 * input.c (get_substring_ranges_for_loc): Detect if
6474 linemap_resolve_location gives us a NULL map, and reject
6477 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
6479 * config.gcc (xstormy16-*-elf): Set tm_d_file.
6481 2018-10-29 David Malcolm <dmalcolm@redhat.com>
6482 Martin Sebor <msebor@redhat.com>
6483 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
6485 * doc/ux.texi (Quoting): New subsection, adapted from material at
6486 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
6487 MartinSebor and ManuelLopezIbanez.
6488 (Fix-it hints): Note that fix-it hints shouldn't be marked for
6491 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
6494 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
6497 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
6499 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
6501 2018-10-29 Paul Koning <ni1d@arrl.net>
6503 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
6505 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
6507 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
6508 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
6509 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
6510 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
6511 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
6512 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
6513 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
6514 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
6515 Change 'vector' to '__vector'.
6516 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
6517 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
6518 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
6519 _mm_avg_pu16): Likewise. And, whitespace corrections.
6521 2018-10-29 Richard Biener <rguenther@suse.de>
6523 PR tree-optimization/87785
6524 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
6527 2018-10-29 Olivier Hainque <hainque@adacore.com>
6529 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
6531 2018-10-29 Olivier Hainque <hainque@adacore.com>
6533 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
6534 ports configuration macro, defaults to "gnu".
6535 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
6536 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
6538 2018-10-29 Olivier Hainque <hainque@adacore.com>
6540 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
6542 2018-10-29 Richard Biener <rguenther@suse.de>
6544 PR tree-optimization/87790
6545 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
6546 (vect_make_slp_decision): Adjust.
6547 (vect_slp_analyze_bb_1): Likewise.
6548 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
6551 2018-10-29 Richard Biener <rguenther@suse.de>
6553 PR tree-optimization/87785
6554 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
6556 (vect_build_slp_tree): Likewise.
6557 (vect_gather_slp_loads): New function.
6558 (vect_analyze_slp_instance): Gather loads separately from the
6561 2018-10-29 Martin Liska <mliska@suse.cz>
6563 * Makefile.in: Make dependency to json.o.
6564 * doc/gcov.texi: Document new JSON format, remove
6565 old intermediate format documentation.
6566 * gcov.c (struct function_info): Come up with m_name and
6568 (function_info::function_info): Initialize it.
6569 (function_info::~function_info): Release it.
6570 (main): Rename flag_intermediate_format to flag_json_format.
6571 (print_usage): Describe --json-format.
6572 (process_args): Set flag_json_format.
6573 (output_intermediate_line): Remove.
6574 (output_intermediate_json_line): Likewise.
6575 (get_gcov_intermediate_filename): Return new extension
6577 (output_intermediate_file): Implement JSON emission.
6578 (output_json_intermediate_file): Implement JSON emission.
6579 (generate_results): Use ::get_name for function name.
6580 Handle JSON output file.
6581 (read_graph_file): Use ::get_name instead of cplus_demangle.
6582 (read_count_file): Likewise.
6583 (solve_flow_graph): Likewise.
6584 (add_line_counts): Likewise.
6585 (accumulate_line_counts): Use new flag_json_format.
6586 (output_function_details): Use ::get_name instead of cplus_demangle.
6587 (output_lines): Likewise.
6588 * json.cc (test_writing_literals): Add new tests.
6589 * json.h (class literal): Add new boolean constructor.
6591 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
6593 PR rtl-optimization/87701
6594 PR rtl-optimization/87780
6595 * combine.c (make_more_copies): Rewrite.
6597 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
6599 * doc/generic.texi (ABSU_EXPR): Document.
6600 * match.pd (absu(x)*absu(x) -> x*x): Handle.
6601 (absu(absu(X)) -> absu(X)): Likewise.
6602 (absu(-X) -> absu(X)): Likewise.
6603 (absu(X) where X is nonnegative -> X): Likewise.
6605 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
6607 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
6608 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
6609 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
6610 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
6611 (s-tm-texi): Also check timestamp on d-target.def.
6612 (generated_files): Add TM_D_H and d-target-hooks-def.h.
6613 (build/genhooks.o): Also depend on D_TARGET_DEF.
6614 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
6616 * config/aarch64/aarch64-d.c: New file.
6617 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
6619 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
6621 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
6622 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
6623 * config/arm/arm-d.c: New file.
6624 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
6625 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
6626 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6627 * config/arm/t-arm (arm-d.o): New rule.
6628 * config/default-d.c: New file.
6629 * config/glibc-d.c: New file.
6630 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6631 * config/i386/i386-d.c: New file.
6632 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
6633 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
6634 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6635 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
6636 * config/i386/t-i386 (i386-d.o): New rule.
6637 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6638 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6639 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
6640 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6641 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6642 * config/mips/mips-d.c: New file.
6643 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
6644 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
6645 * config/mips/t-mips (mips-d.o): New rule.
6646 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6647 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6648 * config/powerpcspe/powerpcspe-d.c: New file.
6649 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
6651 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6652 Support GNU D by using 0 as the language type.
6653 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
6654 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
6655 * config/riscv/riscv-d.c: New file.
6656 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
6658 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
6659 * config/riscv/t-riscv (riscv-d.o): New rule.
6660 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6661 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6662 * config/rs6000/rs6000-d.c: New file.
6663 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
6665 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6666 Support GNU D by using 0 as the language type.
6667 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
6668 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
6669 * config/s390/s390-d.c: New file.
6670 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
6671 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
6672 * config/s390/t-s390 (s390-d.o): New rule.
6673 * config/sparc/sparc-d.c: New file.
6674 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
6676 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
6677 * config/sparc/t-sparc (sparc-d.o): New rule.
6678 * config/t-glibc (glibc-d.o): New rule.
6679 * configure: Regenerated.
6680 * configure.ac (tm_d_file): New variable.
6681 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
6682 * doc/contrib.texi (Contributors): Add self for the D frontend.
6683 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
6684 * doc/install.texi (Configuration): Mention libphobos as an option for
6685 --enable-shared. Mention d as an option for --enable-languages.
6686 (Testing): Mention check-d as a target.
6687 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
6688 name suffixes. Mention d as a -x option.
6689 * doc/sourcebuild.texi (Top Level): Mention libphobos.
6690 * doc/standards.texi (Standards): Add section on D language.
6691 * doc/tm.texi: Regenerated.
6692 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
6693 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
6694 * dwarf2out.c (is_dlang): New function.
6695 (gen_compile_unit_die): Use DW_LANG_D for D.
6696 (declare_in_namespace): Return module die for D, instead of adding
6697 extra declarations into the namespace.
6698 (gen_namespace_die): Generate DW_TAG_module for D.
6699 (gen_decl_die): Handle CONST_DECLSs for D.
6700 (dwarf2out_decl): Likewise.
6701 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
6702 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
6703 kinds of aggregates.
6704 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
6705 * genhooks.c: Include d/d-target.def.
6707 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
6710 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
6711 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
6712 (STACK_DYNAMIC_OFFSET): Likewise.
6714 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
6717 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
6720 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
6722 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
6723 return type and other typos.
6725 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
6727 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
6728 a shorter sequence with fewer branches.
6729 (emit_final_str_compare_gpr): Ditto.
6731 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
6733 * config/rs6000/tmmintrin.h: New file.
6734 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
6736 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
6738 * config/rs6000/mmintrin.h: Enable 32bit compilation.
6739 * config/rs6000/xmmintrin.h: Likewise.
6741 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
6743 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
6745 2018-10-26 Richard Biener <rguenther@suse.de>
6747 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
6749 (vect_mark_slp_stmts_relevant): Likewise.
6750 (vect_detect_hybrid_slp_stmts): Likewise.
6751 (vect_bb_slp_scalar_cost): Likewise.
6752 (vect_remove_slp_scalar_calls): Likewise.
6754 2018-10-26 Jan Hubicka <jh@suse.cz>
6756 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
6757 (warn_types_mismatch): Fix walk of DECL_NAME.
6758 (odr_types_equivalent_p): Fix overactive assert.
6760 2018-10-26 Richard Biener <rguenther@suse.de>
6762 PR tree-optimization/87105
6763 * tree-vectorizer.h (_slp_tree::refcnt): New member.
6764 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
6766 (vect_create_new_slp_node): Initialize refcnt to one.
6768 (scalar_stmts_set_t, bst_fail): Remove.
6769 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
6770 (vect_build_slp_tree): Add bst_map argument and lookup
6771 already created SLP nodes.
6772 (vect_print_slp_tree): Handle a SLP graph, print SLP node
6774 (vect_slp_rearrange_stmts): Handle a SLP graph.
6775 (vect_analyze_slp_instance): Adjust and free SLP nodes from
6776 the CSE map. Fix indenting.
6777 (vect_schedule_slp_instance): Add short-cut.
6779 2018-10-26 Martin Liska <mliska@suse.cz>
6782 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
6783 addr_expr and not with pointers.
6785 2018-10-26 Jan Hubicka <jh@suse.cz>
6787 * tree.c (free_lang_data_in_type): Only check main variants.
6788 * ipa-devirt.c (warn_odr): Make static.
6789 (types_same_for_odr): Drop strict variant.
6790 (types_odr_comparable): Likewise.
6791 (odr_or_derived_type_p): Look for main variants.
6792 (odr_name_hasher::equal): Cleanup comment.
6793 (odr_subtypes_equivalent): Add warn and warned arguments; check main
6795 (type_variants_equivalent_p): break out from ...
6796 (odr_types_equivalent): ... here; go for main variants where needed.
6797 (warn_odr): ... here; turn static.
6798 (warn_types_mismatch): Compare mangled names of main variants.
6799 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
6800 (type_with_linkage_p): Sanity check that we look at main variant.
6801 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
6802 * tree.h (types_same_for_odr): Drop strict argument.
6804 2018-10-26 Richard Biener <rguenther@suse.de>
6806 PR tree-optimization/87746
6807 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
6808 Simplify and fix WRT strided store groups with size not
6809 equal to step in element count.
6810 (vect_analyze_group_access_1): Dump the whole group.
6812 2018-10-25 Carl Love <cel@us.ibm.com>
6814 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
6815 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
6816 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
6817 precicion entry for each overloaded builtin.
6818 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
6819 VSCEDPUO): Rename overloaded name.
6820 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
6821 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
6822 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
6823 define_expand for xscmpexqp instruction.
6824 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
6826 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
6827 Jinsong Ji <jji@us.ibm.com>
6829 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
6830 function with vec_sl.
6831 (_mm_slli_epi32): Likewise.
6832 (_mm_slli_epi64): Likewise.
6833 (_mm_srai_epi16): Replace deprecated function with vec_sra.
6834 (_mm_srai_epi32): Likewise.
6835 (_mm_srli_epi16): Replace deprecated function with vec_sr.
6836 (_mm_srli_epi32): Likewise.
6837 (_mm_srli_epi64): Likewise.
6838 (_mm_sll_epi16): Replace deprecated function with vec_sl.
6839 (_mm_sll_epi32): Likewise.
6840 (_mm_sll_epi64): Likewise.
6841 (_mm_sra_epi16): Replace deprecated function with vec_sra.
6842 (_mm_sra_epi32): Likewise.
6843 (_mm_srl_epi16): Replace deprecated function with vec_sr.
6844 (_mm_srl_epi32): Likewise.
6845 (_mm_srl_epi64): Likewise.
6847 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
6848 Jinsong Ji <jji@us.ibm.com>
6850 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
6851 comparison operators with vec_cmp* for compatibility due to
6852 unfortunate history; clean up formatting and use types more
6854 (_mm_sll_epi32): Likewise.
6855 (_mm_sll_epi64): Likewise.
6856 (_mm_srl_epi16): Likewise.
6857 (_mm_srl_epi32): Likewise.
6858 (_mm_srl_epi64): Likewise.
6860 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
6861 Jinsong Ji <jji@us.ibm.com>
6863 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
6864 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
6865 __vector __bool int. Use vec_cmpgt in preference to deprecated
6866 function vec_vcmpgtfp.
6867 (_mm_max_ps): Likewise.
6869 2018-10-25 Jeff Law <law@redhat.com>
6871 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
6874 2018-10-25 Martin Sebor <msebor@redhat.com>
6876 * doc/extend.texi (aligned): Expand attribute description.
6877 (Alignment): Rename section. Discuss function arguments.
6879 2018-10-25 Jan Hubicka <jh@suse.cz>
6881 * ipa-devirt.c (main_odr_variant): Remove.
6882 (hash_odr_name, types_same_for_odr, types_odr_comparable,
6883 odr_name_hasher::equal, odr_subtypes_equivalent_p):
6884 Drop use of main_odr_variant.
6885 (add_type_duplicate): Silence confused warnings on integer types.
6886 (get_odr_type): Always look for main variant.
6887 (register_odr_type): Simplify.
6889 2018-10-25 Richard Biener <rguenther@suse.de>
6891 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
6892 Initialize ng to silence error with release checking bootstrap.
6894 2018-10-25 Richard Biener <rguenther@suse.de>
6896 * tree-if-conv.c: Include tree-ssa-sccvn.h.
6897 (tree_if_conversion): Run CSE on the if-converted loop body.
6899 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
6901 * config/s390/constraints.md (ZL): New constraint.
6902 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
6904 * config/s390/s390.md (movdi_larl): Remove.
6905 (movdi_64): Add the LARL alternative.
6907 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
6910 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
6911 (RTX_CODE_HWINT_P): New macro.
6912 (rtx_code_size): Use RTX_CODE_HWINT_P ().
6914 2018-10-25 Jan Hubicka <jh@suse.cz>
6916 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
6919 2018-10-25 Richard Biener <rguenther@suse.de>
6921 PR tree-optimization/87665
6922 PR tree-optimization/87745
6923 * tree-vectorizer.h (get_earlier_stmt): Remove.
6924 (get_later_stmt): Pick up UID from the original non-pattern stmt.
6926 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
6928 * options.texi (Deprecated): Move list to Var section.
6930 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
6931 Jinsong Ji <jji@us.ibm.com>
6933 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
6934 __vector long to __vector long long.
6935 (_mm_cvtpd_ps): Likewise.
6936 (_mm_cvttpd_epi32): Likewise.
6937 (_mm_cvtpi32_pd): Likewise.
6938 (_mm_unpackhi_epi64): Likewise.
6939 (_mm_unpacklo_epi64): Likewise.
6941 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
6943 PR rtl-optimization/87720
6944 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
6946 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
6948 * gimple-ssa-isolate-paths.c
6949 (find_implicit_erroneous_behavior): Do not change code if the
6950 pass is running for warnings only.
6951 (find_explicit_erroneous_behavior): Likewise.
6953 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
6955 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
6956 Define as rs6000_mangle_decl_assembler_name.
6957 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
6958 long double to IEEE long double, switch the names of the long
6959 double built-in functions to be <func>f128 instead of <func>l.
6961 2018-10-24 Martin Sebor <msebor@redhat.com>
6963 * doc/extend.texi (nonnull): List no-argument form. Reference
6964 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
6966 2018-10-24 Richard Biener <rguenther@suse.de>
6968 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
6970 2018-10-24 Martin Liska <mliska@suse.cz>
6972 PR tree-optimization/84436
6973 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
6975 (switch_conversion::contains_linear_function_p): New.
6976 (switch_conversion::build_one_array): Support linear
6977 transformation on input.
6978 * tree-switch-conversion.h (struct switch_conversion): Add
6979 contains_linear_function_p declaration.
6981 2018-10-24 Richard Biener <rguenther@suse.de>
6983 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
6984 if its argument is CONSTANT_CLASS_P.
6986 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
6988 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
6989 it is wrong for forward declarations.
6991 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
6993 * config/s390/s390.c (s390_check_qrst_address): Add the missing
6994 SYMBOL_REF_P () check.
6996 2018-10-24 Richard Biener <rguenther@suse.de>
6998 PR tree-optimization/87105
6999 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
7000 dump classification.
7001 (vect_analyze_data_ref_accesses): Handle duplicate loads and
7002 stores by splitting the affected group after the fact.
7003 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
7004 fail the SLP build because of size constraints.
7006 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7008 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
7009 * configure: Regenerate.
7010 * config.in: Regenerate.
7011 * varasm.c (mergeable_string_section): Use readonly_data_section
7012 if linker doesn't support SHF_MERGE with alignment > 8.
7013 (mergeable_constant_section): Likewise.
7015 2018-10-24 Richard Biener <rguenther@suse.de>
7017 PR tree-optimization/84013
7018 * tree-ssa-structalias.c (struct msdi_data): New struct for
7019 marshalling data to walk_stmt_load_store_ops.
7020 (maybe_set_dependence_info): Refactor as callback for
7021 walk_stmt_load_store_ops.
7022 (compute_dependence_clique): Set restrict info on all stmt kinds.
7024 2018-10-24 Martin Liska <mliska@suse.cz>
7026 * cgraph.c (cgraph_node::dump):
7027 Remove reduntant dumps and make tp_first_run dump more compact.
7029 2018-10-24 Richard Biener <rguenther@suse.de>
7031 PR tree-optimization/87665
7032 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
7035 2018-10-12 Jeff Law <law@redhat.com>
7037 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
7040 2018-10-23 Richard Biener <rguenther@suse.de>
7042 * tree-vrp.c (add_assert_info): Guard dump_printf with
7044 * gimple-ssa-evrp-analyze.c
7045 (evrp_range_analyzer::record_ranges_from_incoming_edge):
7046 Use value_range::ignore_equivs_equal_p.
7048 2018-10-23 Richard Biener <rguenther@suse.de>
7050 PR tree-optimization/87105
7051 PR tree-optimization/87608
7052 * passes.def (pass_all_early_optimizations): Add early phi-opt
7054 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
7055 addition to debug stmts.
7056 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
7057 and abs replacement early.
7058 * tree-cfg.c (gimple_empty_block_p): Likewise.
7060 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
7063 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
7064 specified to configure.
7065 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
7067 2018-10-23 Richard Biener <rguenther@suse.de>
7069 PR tree-optimization/87700
7070 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
7072 2018-10-23 Jakub Jelinek <jakub@redhat.com>
7075 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
7076 second argument from __mmask16 to __mmask8.
7077 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
7078 _mm_mask_packs_epi32): Likewise.
7079 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
7081 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
7083 2018-10-23 Richard Biener <rguenther@suse.de>
7085 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
7087 2018-10-23 Richard Biener <rguenther@suse.de>
7089 PR tree-optimization/86144
7090 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
7091 over simd attribute.
7093 2018-10-23 Richard Biener <rguenther@suse.de>
7095 PR tree-optimization/87693
7096 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
7097 the case we do not find the taken edge.
7099 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
7100 Jinsong Ji <jji@us.ibm.com>
7102 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
7103 (_mm_store_pd): Use unaligned vector type for pointer cast.
7104 (_mm_maskmoveu_si128): Likewise.
7105 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
7106 (_mm_store_ps): Use unaligned vector type for pointer cast.
7108 2018-10-22 Paul Koning <ni1d@arrl.net>
7110 * symtab.c (symtab_node::increase_alignment): Correct max
7113 2018-10-22 Yury Gribov <tetra2005@gmail.com>
7115 PR tree-optimization/87633
7116 * match.pd: Do not generate unordered integer comparisons.
7118 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
7120 PR rtl-optimization/87600
7121 * combine.c: Add include of expr.h.
7122 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
7123 register to a pseudo.
7124 (make_more_copies): New function, add a copy to a new pseudo after
7125 the moves from hard registers into pseudos.
7126 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
7127 later. Call make_more_copies.
7129 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
7131 * lra-constraints.c (process_alt_operands): New local array,
7132 matching_early_clobber. Check matching_early_clobber before
7133 decrementing reject, and set matching_early_clobber after.
7135 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
7138 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
7139 call output_addr_const and hope for the best.
7141 2018-10-22 Richard Biener <rguenther@suse.de>
7143 * gimple-ssa-evrp-analyze.c
7144 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
7145 smarter about what ranges to use.
7146 * tree-vrp.c (add_assert_info): Dump here.
7147 (register_edge_assert_for_2): Instead of here at multiple but
7150 * gcc.dg/tree-ssa/evrp12.c: New testcase.
7151 * gcc.dg/predict-6.c: Adjust.
7152 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
7153 * gcc.dg/tree-ssa/vrp02.c: Likewise.
7154 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
7156 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
7157 Richard Biener <rguenther@suse.de>
7159 * bitmap.h: Update data structure documentation, including a
7160 description of bitmap views as either linked-lists or splay trees.
7161 (struct bitmap_element_def): Update comments for splay tree bitmaps.
7162 (struct bitmap_head_def): Likewise.
7163 (bitmap_list_view, bitmap_tree_view): New prototypes.
7164 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
7166 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
7167 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
7168 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
7169 released bitmap element here.
7170 (bitmap_element_free): Remove.
7171 (bitmap_elt_clear_from): Work on splay tree bitmaps.
7172 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
7173 this function similar ones such that linked-list bitmap implementation
7174 functions are grouped.
7175 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
7176 and moved for grouping.
7177 (bitmap_list_insert_element_after): Renamed from
7178 bitmap_elt_insert_after, and moved for grouping.
7179 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
7180 (bitmap_tree_link_left, bitmap_tree_link_right,
7181 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
7182 bitmap_tree_link_element, bitmap_tree_unlink_element,
7183 bitmap_tree_find_element): New functions for splay-tree bitmap
7185 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
7186 Renamed and moved, see above entries.
7187 (bitmap_tree_listify_from): New function to convert part of a splay
7188 tree bitmap to a linked-list bitmap.
7189 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
7190 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
7191 (bitmap_find_bit): Remove.
7192 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
7193 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
7194 Handle splay tree bitmaps.
7195 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
7196 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
7197 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
7198 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
7199 bitmap_intersect_compl_p, bitmap_ior_and_compl,
7200 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
7201 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
7202 corresponding changes to use linked-list specific bitmap_element
7203 manipulation functions as applicable for efficiency.
7204 (bitmap_tree_to_vec): New function.
7205 (debug_bitmap_elt_file): New function split out from ...
7206 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
7207 (bitmap_print): Likewise.
7209 PR tree-optimization/63155
7210 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
7212 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
7215 2018-10-22 Martin Liska <mliska@suse.cz>
7217 PR tree-optimization/87686
7219 2018-08-29 Martin Liska <mliska@suse.cz>
7221 * tree-switch-conversion.c (switch_conversion::expand):
7222 Strenghten assumption about gswitch statements.
7224 2018-10-22 Martin Liska <mliska@suse.cz>
7226 * ipa-icf.c (sem_item::compare_attributes): Remove.
7227 (sem_item::compare_referenced_symbol_properties): Use
7228 attribute_list_equal instead.
7229 (sem_function::equals_wpa): Likewise.
7230 * ipa-icf.h: Remove compare_attributes.
7232 2018-10-22 Richard Biener <rguenther@suse.de>
7235 * mem-stats.h (mem_usage::operator==): Fix pasto.
7237 2018-10-22 Richard Biener <rguenther@suse.de>
7239 PR tree-optimization/87640
7240 * tree-vrp.c (set_value_range_with_overflow): Decompose
7242 (extract_range_from_binary_expr_1): Adjust.
7244 2018-10-22 Martin Jambor <mjambor@suse.cz>
7246 * tree-eh.h (stmt_could_throw_p): Add function parameter.
7247 (stmt_can_throw_external): Likewise.
7248 (stmt_can_throw_internal): Likewise.
7249 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
7250 (lower_eh_constructs_2): Likewise.
7251 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
7252 (stmt_can_throw_external): Likewise.
7253 (stmt_can_throw_internal): Likewise.
7254 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
7255 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
7256 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
7257 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
7258 (pass_lower_eh_dispatch::execute): Pass cfun to
7259 stmt_can_throw_external.
7260 (cleanup_empty_eh): Likewise.
7261 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
7262 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
7263 stmt_can_throw_external instead of pushing it to cfun.
7264 (symbol_table::create_edge): Likewise.
7265 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
7266 stmt_can_throw_internal.
7267 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
7268 to stmt_could_throw_p.
7269 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
7270 stmt_can_throw_internal.
7271 (pass_store_merging::execute): Likewise.
7272 * gimple-ssa-strength-reduction.c
7273 (find_candidates_dom_walker::before_dom_children): Pass cfun to
7275 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
7276 stmt_can_throw_internal.
7277 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
7278 to stmt_can_throw_external.
7279 (check_stmt): Pass cfun to stmt_could_throw_p.
7280 (check_stmt): Pass cfun to stmt_can_throw_external.
7281 (pass_nothrow::execute): Likewise.
7282 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
7283 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
7284 stmt_can_throw_internal.
7285 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
7286 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
7287 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
7288 * tree-complex.c (expand_complex_libcall): Pass cfun to
7289 stmt_could_throw_p and to stmt_can_throw_internal.
7290 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
7291 * tree-inline.c (copy_edges_for_bb): Likewise.
7292 (maybe_move_debug_stmts_to_successors): Likewise.
7293 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
7295 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
7296 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
7297 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
7298 stmt_can_throw_internal.
7299 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
7300 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
7302 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
7303 stmt_can_throw_internal.
7304 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
7305 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
7307 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
7308 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
7309 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7310 (convert_mult_to_fma_1): Likewise.
7311 (convert_to_divmod): Likewise.
7312 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
7313 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
7314 * tree-ssa-propagate.c
7315 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
7316 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
7317 (maybe_optimize_range_tests): Likewise.
7318 (linearize_expr_tree): Likewise.
7319 (reassociate_bb): Likewise.
7320 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7321 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
7322 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
7323 (handle_char_store): Likewise.
7324 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
7325 stmt_can_throw_internal.
7326 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
7328 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7329 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
7330 (vectorizable_simd_clone_call): Likewise.
7331 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
7332 (gimple_stringop_fixed_value): Likewise.
7334 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
7336 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
7337 literal pool references.
7338 (s390_check_qrst_address): Adapt to the new behavior of
7339 s390_loadrelative_operand_p ().
7341 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
7344 * config/i386/sse.md (*andnot<mode>3_bcst): New.
7346 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
7349 * config/i386/sse.md (*<code><mode>3_bcst): New.
7351 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
7354 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
7355 V4DI, V16SI and V8DI.
7356 (*sub<mode>3<mask_name>_bcst): New.
7357 (*add<mode>3<mask_name>_bcst): Likewise.
7359 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
7360 Jinsong Ji <jji@us.ibm.com>
7362 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
7363 __m64 with __vector unsigned long long for compatibility.
7364 (_mm_movemask_epi8): Likewise.
7365 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
7366 (_mm_cvttps_pi32): Likewise.
7367 (_mm_cvtpi32_ps): Likewise.
7368 (_mm_cvtps_pi16): Likewise.
7369 (_mm_loadh_pi): Likewise.
7370 (_mm_storeh_pi): Likewise.
7371 (_mm_movehl_ps): Likewise.
7372 (_mm_movelh_ps): Likewise.
7373 (_mm_loadl_pi): Likewise.
7374 (_mm_storel_pi): Likewise.
7375 (_mm_movemask_ps): Likewise.
7376 (_mm_shuffle_pi16): Likewise.
7378 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7381 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
7382 __builtin_ia32_vfnmsubpd512_mask.
7383 (_mm512_mask_fnmsub_round_pd): Likewise.
7384 (_mm512_fnmsub_pd): Likewise.
7385 (_mm512_mask_fnmsub_pd): Likewise.
7386 (_mm512_maskz_fnmsub_round_pd): Use
7387 __builtin_ia32_vfnmsubpd512_maskz.
7388 (_mm512_maskz_fnmsub_pd): Likewise.
7389 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
7390 (_mm512_mask_fnmsub_round_ps): Likewise.
7391 (_mm512_fnmsub_ps): Likewise.
7392 (_mm512_mask_fnmsub_ps): Likewise.
7393 (_mm512_maskz_fnmsub_round_ps): Use
7394 __builtin_ia32_vfnmsubps512_maskz.
7395 (_mm512_maskz_fnmsub_ps): Likewise.
7396 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
7397 __builtin_ia32_vfnmsubpd256_mask.
7398 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
7399 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
7400 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
7401 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
7402 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
7403 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
7404 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
7405 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
7406 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
7407 __builtin_ia32_vfnmsubpd.
7408 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
7409 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
7410 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
7411 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
7412 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
7413 * config/i386/i386-builtin.def: Add
7414 __builtin_ia32_vfnmsubpd256_mask,
7415 __builtin_ia32_vfnmsubpd256_maskz,
7416 __builtin_ia32_vfnmsubpd128_mask,
7417 __builtin_ia32_vfnmsubpd128_maskz,
7418 __builtin_ia32_vfnmsubps256_mask,
7419 __builtin_ia32_vfnmsubps256_maskz,
7420 __builtin_ia32_vfnmsubps128_mask,
7421 __builtin_ia32_vfnmsubps128_maskz,
7422 __builtin_ia32_vfnmsubpd512_mask,
7423 __builtin_ia32_vfnmsubpd512_maskz,
7424 __builtin_ia32_vfnmsubps512_mask,
7425 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
7426 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
7427 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
7428 __builtin_ia32_vfnmsubpd256.
7429 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
7430 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
7431 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
7433 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
7435 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
7437 (fmai_vmfnmsub_<mode><round_name>): Likewise.
7439 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7442 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
7443 __builtin_ia32_vfnmaddpd512_mask.
7444 (_mm512_mask_fnmadd_round_pd): Likewise.
7445 (_mm512_fnmadd_pd): Likewise.
7446 (_mm512_mask_fnmadd_pd): Likewise.
7447 (_mm512_maskz_fnmadd_round_pd): Use
7448 __builtin_ia32_vfnmaddpd512_maskz.
7449 (_mm512_maskz_fnmadd_pd): Likewise.
7450 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
7451 (_mm512_mask_fnmadd_round_ps): Likewise.
7452 (_mm512_fnmadd_ps): Likewise.
7453 (_mm512_mask_fnmadd_ps): Likewise.
7454 (_mm512_maskz_fnmadd_round_ps): Use
7455 __builtin_ia32_vfnmaddps512_maskz.
7456 (_mm512_maskz_fnmadd_ps): Likewise.
7457 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
7458 __builtin_ia32_vfnmaddpd256_mask.
7459 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
7460 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
7461 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
7462 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
7463 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
7464 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
7465 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
7466 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
7467 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
7468 __builtin_ia32_vfnmaddpd.
7469 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
7470 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
7471 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
7472 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
7473 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
7474 * config/i386/i386-builtin.def: Add
7475 __builtin_ia32_vfnmaddpd256_mask,
7476 __builtin_ia32_vfnmaddpd256_maskz,
7477 __builtin_ia32_vfnmaddpd128_mask,
7478 __builtin_ia32_vfnmaddpd128_maskz,
7479 __builtin_ia32_vfnmaddps256_mask,
7480 __builtin_ia32_vfnmaddps256_maskz,
7481 __builtin_ia32_vfnmaddps128_mask,
7482 __builtin_ia32_vfnmaddps128_maskz,
7483 __builtin_ia32_vfnmaddpd512_mask,
7484 __builtin_ia32_vfnmaddpd512_maskz,
7485 __builtin_ia32_vfnmaddps512_mask,
7486 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
7487 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
7488 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
7489 __builtin_ia32_vfnmaddpd256.
7490 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
7491 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
7492 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
7494 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
7496 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
7498 (fmai_vmfnmadd_<mode><round_name>): Likewise.
7500 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7503 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
7504 __builtin_ia32_vfmsubpd512_mask.
7505 (_mm512_mask_fmsub_round_pd): Likewise.
7506 (_mm512_fmsub_pd): Likewise.
7507 (_mm512_mask_fmsub_pd): Likewise.
7508 (_mm512_maskz_fmsub_round_pd): Use
7509 __builtin_ia32_vfmsubpd512_maskz.
7510 (_mm512_maskz_fmsub_pd): Likewise.
7511 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
7512 (_mm512_mask_fmsub_round_ps): Likewise.
7513 (_mm512_fmsub_ps): Likewise.
7514 (_mm512_mask_fmsub_ps): Likewise.
7515 (_mm512_maskz_fmsub_round_ps): Use
7516 __builtin_ia32_vfmsubps512_maskz.
7517 (_mm512_maskz_fmsub_ps): Likewise.
7518 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
7519 __builtin_ia32_vfmsubpd256_mask.
7520 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
7521 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
7522 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
7523 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
7524 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
7525 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
7526 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
7527 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
7528 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
7529 __builtin_ia32_vfmsubpd.
7530 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
7531 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
7532 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
7533 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
7534 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
7535 * config/i386/i386-builtin.def: Add
7536 __builtin_ia32_vfmsubpd256_mask,
7537 __builtin_ia32_vfmsubpd256_maskz,
7538 __builtin_ia32_vfmsubpd128_mask,
7539 __builtin_ia32_vfmsubpd128_maskz,
7540 __builtin_ia32_vfmsubps256_mask,
7541 __builtin_ia32_vfmsubps256_maskz,
7542 __builtin_ia32_vfmsubps128_mask,
7543 __builtin_ia32_vfmsubps128_maskz,
7544 __builtin_ia32_vfmsubpd512_mask,
7545 __builtin_ia32_vfmsubpd512_maskz,
7546 __builtin_ia32_vfmsubps512_mask,
7547 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
7548 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
7549 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
7550 __builtin_ia32_vfmsubpd256.
7551 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
7552 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
7553 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
7555 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
7557 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
7559 (fmai_vmfmsub_<mode><round_name>): Likewise.
7561 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7563 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
7564 Remove plus. Renamed to ...
7565 (*sub<mode>3<mask_name>_bcst): This.
7566 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
7567 (*add<mode>3<mask_name>_bcst): This.
7569 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7572 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
7574 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7577 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
7578 (_mm_or_epi32): Likewise.
7579 (_mm256_xor_epi32): Likewise.
7580 (_mm_xor_epi32): Likewise.
7581 (_mm256_or_epi64): Likewise.
7582 (_mm_or_epi64): Likewise.
7583 (_mm256_xor_epi64): Likewise.
7584 (_mm_xor_epi64): Likewise.
7586 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
7589 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
7591 2018-10-20 Jakub Jelinek <jakub@redhat.com>
7594 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
7596 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
7598 * doc/ux.texi: Move @section directly after @node.
7600 2018-10-19 Jakub Jelinek <jakub@redhat.com>
7604 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
7605 depend closely nested inside of loop with ordered clause with
7608 2018-10-19 David Malcolm <dmalcolm@redhat.com>
7610 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
7611 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
7612 * doc/ux.texi: New file.
7614 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
7616 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
7617 be the first CR field allocated.
7619 2018-10-19 Richard Biener <rguenther@suse.de>
7622 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
7623 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
7625 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
7628 * config/i386/sse.md
7629 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
7630 (*add<mode>3<mask_name>_bcst_2): Likewise.
7632 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
7634 * config/i386/sse.md
7635 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7636 Replace nonimmediate_operand with register_operand.
7637 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7639 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7642 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
7644 PR rtl-optimization/87596
7645 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
7646 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
7647 for instructions in FROM..TO range.
7649 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
7651 * cfgexpand.c (expand_one_var): Use specific wording in error message
7652 for non-local frame variables.
7653 * stor-layout.c (layout_decl): Do not issue a warning for them.
7655 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
7657 * haifa-sched.c (priority): Add force_recompute parameter.
7658 (apply_replacement): Call priority () with force_recompute = true.
7659 (restore_pattern): Likewise.
7661 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7663 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
7664 HOST_BITS_PER_WIDE_INT.
7665 (test_vector_ops_duplicate): Likewise.
7667 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7670 * config/i386/sse.md (VF_AVX512): New.
7671 (avx512bcst): Likewise.
7672 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7674 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7676 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7679 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
7681 * doc/invoke.texi (-dumpversion): Improve grammar.
7682 (-dumpfullversion): Make more consistent with -dumpversion.
7684 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
7686 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
7687 Set JUMP_LABEL to the jump insn.
7688 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
7689 Predict emitted jump and add label to jump insn.
7691 2018-10-18 David Malcolm <dmalcolm@redhat.com>
7693 PR tree-optimization/87562
7694 * input.c (get_substring_ranges_for_loc): Use
7695 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
7696 getting the linemap for the endpoint. Verify that it's either
7697 in the same linemap as the start point's spelling location, or
7698 at least in the same file.
7700 2018-10-18 Richard Biener <rguenther@suse.de>
7702 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
7703 feed width-specific load/store costs through ix86_vec_cost.
7704 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
7705 (k8_cost): Likewise.
7706 (bdver_cost): Likewise.
7707 (znver1_cost): Likewise.
7708 (btver1_cost): Likewise.
7709 (btver2_cost): Likewise.
7711 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7713 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
7714 to simplify subreg of vec_merge.
7716 2018-10-18 Richard Biener <rguenther@suse.de>
7718 * config/i386/i386.c: Fix costing of vector FMA.
7720 2018-10-18 Richard Biener <rguenther@suse.de>
7722 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
7724 (ix86_builtin_vectorization_cost): For vec_construct properly
7725 cost insertion into SSE regs.
7726 (...): Adjust calls to ix86_vec_cost.
7728 2018-10-18 Richard Biener <rguenther@suse.de>
7732 2018-02-07 Richard Biener <rguenther@suse.de>
7734 PR tree-optimization/84204
7735 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7738 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7741 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
7743 (test_vector_ops_duplicate): Add test for a scalar subreg of a
7744 VEC_MERGE of a VEC_DUPLICATE.
7746 2018-10-17 Joseph Myers <joseph@codesourcery.com>
7748 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
7749 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
7750 * doc/standards.texi (C Language): Document C2X.
7751 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
7752 (rl78_option_override): Handle "GNU C2X" language name.
7754 2018-10-17 Joseph Myers <joseph@codesourcery.com>
7756 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
7757 Document C17 as published in 2018.
7759 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
7762 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
7763 bail out if both sides do not have the same storage order.
7765 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
7767 * bitmap.c (bitmap_head::dump): New.
7768 * bitmap.h (bitmap_head): Add dump().
7769 * gimple-ssa-evrp-analyze.c
7770 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
7771 (evrp_range_analyzer::set_ssa_range_info): Same.
7772 (evrp_range_analyzer::record_ranges_from_phis): Same.
7773 (evrp_range_analyzer::record_ranges_from_stmt): Same.
7774 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
7775 * gimple-ssa-sprintf.c (get_int_range): Same.
7776 (format_integer): Same.
7777 (sprintf_dom_walker::handle_gimple_call): Same.
7778 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
7779 (ipcp_vr_lattice::top_p): Same.
7780 (ipcp_vr_lattice::bottom_p): Same.
7781 (ipcp_vr_lattice::set_to_bottom): Same.
7782 (ipa_vr_operation_and_type_effects): Same.
7783 (propagate_vr_across_jump_function): Same.
7784 (ipcp_store_vr_results): Same.
7785 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
7786 (ipa_print_node_jump_functions_for_edge): Same.
7787 (ipa_get_value_range): Same.
7788 (ipa_compute_jump_functions_for_edge): Same.
7789 (ipa_write_jump_function): Same.
7790 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
7791 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
7793 * vr-values.c (set_value_range_to_nonnegative): Same.
7794 (set_value_range_to_truthvalue): Same.
7795 (vr_values::get_value_range): Same.
7796 (vr_values::set_defs_to_varying): Same.
7797 (vr_values::update_value_range): Same.
7798 (symbolic_range_based_on_p): Same.
7799 (vr_values::op_with_boolean_value_range_p): Same.
7800 (vr_values::extract_range_for_var_from_comparison_expr): Same.
7801 (vr_values::extract_range_from_ssa_name): Same.
7802 (vr_values::extract_range_from_binary_expr): Same.
7803 (vr_values::extract_range_from_unary_expr): Same.
7804 (vr_values::extract_range_from_cond_expr): Same.
7805 (vr_values::extract_range_from_comparison): Same.
7806 (vr_values::check_for_binary_op_overflow): Same.
7807 (vr_values::extract_range_basic): Same.
7808 (vr_values::extract_range_from_assignment): Same.
7809 (compare_ranges): Same.
7810 (compare_range_with_value): Same.
7811 (vr_values::adjust_range_with_scev): Same.
7812 (vrp_valueize): Same.
7813 (vrp_valueize_1): Same.
7814 (vr_values::get_vr_for_comparison): Same.
7815 (vr_values::compare_name_with_value): Same.
7816 (vr_values::compare_names): Same.
7817 (vr_values::vrp_evaluate_conditional): Same.
7818 (find_case_label_ranges): Same.
7819 (vr_values::vrp_visit_switch_stmt): Same.
7820 (vr_values::extract_range_from_phi_node): Same.
7821 (vr_values::simplify_div_or_mod_using_ranges): Same.
7822 (vr_values::simplify_bit_ops_using_ranges): Same.
7823 (test_for_singularity): Same.
7824 (range_fits_type_p): Same.
7825 (vr_values::simplify_cond_using_ranges_1): Same.
7826 (vr_values::simplify_switch_using_ranges): Same.
7827 (vr_values::simplify_float_conversion_using_ranges): Same.
7828 (vr_values::two_valued_val_range_p): Same.
7829 (vr_values::add_equivalence): Move to value_range::equiv_add.
7830 * vr-values.h (vr_values::add_equivalence): Remove.
7831 (VR_INITIALIZER): Remove.
7832 * tree-vrp.c (value_range::set): New.
7833 (value_range::equiv_add): New.
7834 (value_range::value_range): New.
7835 (value_range::deep_copy): New.
7836 (value_range::check): New.
7837 (value_range::equal_p): New.
7838 (value_range::ignore_equivs_equal_p): New.
7839 (value_range::operator==): New.
7840 (value_range::operator!=): New.
7841 (value_range::symbolic_p): New.
7842 (value_range::numeric_p): New.
7843 (value_range::set_undefined): New.
7844 (value_range::set_varying): New.
7845 (value_range::may_contain_p): New.
7846 (value_range::equiv_clear): New.
7847 (value_range::singleton_p): New.
7848 (value_range::intersect): New.
7849 (value_range::dump): New.
7850 (value_range::set_and_canonicalize): New.
7851 (set_value_range): Adjust for value_range API.
7852 (set_value_range_to_undefined): Same.
7853 (set_value_range_to_varying): Same.
7854 (set_and_canonicalize_value_range): Same.
7855 (set_value_range_to_nonnull): Same.
7856 (set_value_range_to_null): Same.
7857 (range_is_null): Same.
7858 (range_is_nonnull): Same.
7859 (range_int_cst_p): Same.
7860 (range_int_cst_singleton_p): Same.
7861 (symbolic_range_p): Same.
7862 (range_includes_zero_p): Same.
7863 (value_range_constant_singleton): Same.
7864 (vrp_set_zero_nonzero_bits): Same.
7865 (ranges_from_anti_range): Same.
7866 (extract_range_into_wide_ints): Same.
7867 (extract_range_from_multiplicative_op): Same.
7868 (set_value_range_with_overflow): Same.
7869 (extract_range_from_binary_expr_1): Same.
7870 (extract_range_from_unary_expr): Same.
7871 (dump_value_range): Same.
7872 (debug_value_range): Same.
7873 (vrp_prop::check_array_ref): Same.
7874 (vrp_prop::check_mem_ref): Same.
7875 (vrp_prop::vrp_initialize): Same.
7876 (vrp_prop::visit_stmt): Same.
7877 (intersect_ranges): Same.
7878 (vrp_prop::visit_phi): Same.
7879 (vrp_prop::vrp_finalize): Same.
7880 (determine_value_range_1): Same.
7881 (determine_value_range): Same.
7882 (vrp_intersect_ranges_1): Rename to...
7883 (vrp_intersect_1): this.
7884 (vrp_intersect_ranges): Rename to...
7885 (value_range::intersect_helper): ...this.
7886 (vrp_meet_1): Rename to...
7887 (value_range::union_helper): ...this.
7888 (vrp_meet): Rename to...
7889 (value_range::union_): ...this.
7890 (copy_value_range): Remove.
7891 * tree-vrp.h (struct value_range): Rewrite into a proper class.
7892 (value_range::vrtype): New.
7893 (value_range::type): New.
7894 (value_range::equiv): New.
7895 (value_range::min): New.
7896 (value_range::max): New.
7897 (value_range::varying_p): New.
7898 (value_range::undefined_p): New.
7899 (value_range::null_p): New.
7900 (value_range::equiv_add): New.
7901 (copy_value_range): Remove.
7903 2018-10-17 David Malcolm <dmalcolm@redhat.com>
7905 * Makefile.in (SELFTEST_TARGETS): New.
7906 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
7907 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
7908 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
7910 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
7911 (selftest-c++-gdb, selftest-c++-valgrind): Move to
7913 * configure: Regenerate.
7914 * configure.ac (selftest_languages): New.
7916 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
7918 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
7919 overflow wraps argument.
7920 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
7921 wide_int_range_multiplicative_op.
7922 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
7923 overflow wraps argument.
7924 (wide_int_range_multiplicative_op): Same.
7925 (wide_int_range_lshift): Same.
7926 (wide_int_range_div): Same.
7927 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
7928 (wide_int_range_lshift): Same.
7929 (wide_int_range_div): Same.
7931 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
7933 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
7934 use sign as argument.
7935 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
7936 wide_int_range_shift_undefined_p.
7938 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7940 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
7942 (@despeculate_copy<ALLI_TI:mode>): ... This.
7943 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
7946 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7948 * config.gcc: Obsolete *-*-solaris2.10*.
7949 * doc/install.texi (Specific, *-*-solaris2*): Document it.
7951 2018-10-12 Jeff Law <law@redhat.com>
7953 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
7954 reg + sym +- const_int addressing modes.
7956 2018-10-15 David Malcolm <dmalcolm@redhat.com>
7958 * common.opt (fdiagnostics-minimum-margin-width=): New option.
7959 * diagnostic-show-locus.c (layout::layout): Apply the minimum
7961 (layout::start_annotation_line): Only print up to 3 of the
7962 margin character, to avoid touching the left-hand side.
7963 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
7964 minimum margin width, as set by test_diagnostic_context's ctor.
7965 (selftest::test_fixit_insert_containing_newline): Likewise.
7966 (selftest::test_fixit_insert_containing_newline_2): Likewise.
7967 (selftest::test_line_numbers_multiline_range): Clear
7968 dc.min_margin_width.
7969 * diagnostic.c (diagnostic_initialize): Initialize
7971 * diagnostic.h (struct diagnostic_context): Add field
7973 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
7974 * opts.c (common_handle_option): Handle
7975 OPT_fdiagnostics_minimum_margin_width_.
7976 * selftest-diagnostic.c
7977 (selftest::test_diagnostic_context::test_diagnostic_context):
7978 Initialize min_margin_width to 6.
7979 * toplev.c (general_init): Initialize global_dc->min_margin_width.
7981 2018-10-15 David Malcolm <dmalcolm@redhat.com>
7983 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
7984 Fix usage of "error_at_rich_loc" in the comment.
7986 2018-10-15 Renlin Li <renlin.li@arm.com>
7989 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
7990 if-conversioned loop when it contains ifn with types not
7991 supported by backend.
7992 * internal-fn.c (expand_direct_optab_fn): Add an assert.
7993 (direct_internal_fn_supported_p): New helper function.
7994 * internal-fn.h (direct_internal_fn_supported_p): Declare.
7996 2018-10-15 Jakub Jelinek <jakub@redhat.com>
7999 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
8000 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
8001 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
8003 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
8005 PR tree-optimization/87022
8006 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
8007 bits in dist vector rather than the first one.
8009 2018-10-15 Richard Biener <rguenther@suse.de>
8012 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
8013 (visit_loadstore): When a used restrict tag escaped verify that
8014 the points-to solution of "other" pointers do not include
8016 (compute_dependence_clique): If a used restrict tag escaped
8017 communicated that down to visit_loadstore.
8019 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
8021 * config/s390/s390.c (s390_expand_vec_init): Force vector element
8022 into reg if it isn't a general operand.
8024 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
8027 * config/i386/sse.md (*vec_dupv2di): Add register source to
8030 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
8033 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
8034 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
8035 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
8036 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
8038 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
8040 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
8041 (notice_args_size): Set it in the current trace if no insn that can
8042 throw internally has been seen yet.
8043 (connect_traces): When connecting args_size between traces, allow the
8044 incoming values not to match if there is an insn setting it before the
8045 first insn that can throw internally; in that case, force the creation
8046 of a CFI note on this latter insn.
8048 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
8050 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
8051 the base class, because using the injected-class-name was not clearly
8052 specified until DR 176.
8054 2018-10-12 Paul Koning <ni1d@arrl.net>
8056 * config/pdp11/pdp11.md (doloop_end): New expander.
8057 (doloop_end_insn): renamed from "doloop_end".
8058 (addqi3): New pattern.
8059 (subqi3): New pattern.
8060 * config/pdp11/predicates.md (incdec_operand): New predicate.
8062 2018-10-12 Yury Gribov <tetra2005@gmail.com>
8065 * real.c (format_helper::can_represent_integral_type_p): New function
8066 * real.h (format_helper::can_represent_integral_type_p): Ditto.
8067 * match.pd: New pattern.
8069 2018-10-12 Alexandre Oliva <oliva@adacore.com>
8071 * configure.ac: Introduce --enable-large-address-aware
8072 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
8073 * doc/install.texi: Document it.
8074 * configure, config.in: Rebuilt.
8075 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
8076 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
8077 (LINK_SPEC): Insert it.
8078 * config/i386/mingw-w64.h: Likewise.
8080 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
8082 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
8084 PR rtl-optimization/87600
8085 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
8087 2018-10-12 Paul Koning <ni1d@arrl.net>
8089 * doc/md.texi (doloop_end): Document that the pattern code may
8090 need to check operand mode.
8092 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
8094 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
8095 to zero-extend between int and floating-point registers.
8096 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
8097 ldp into floating-point registers. Add type and arch attributes.
8098 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
8099 Use f_loads for type attribute.
8101 2018-10-11 Martin Sebor <msebor@redhat.com>
8103 * doc/extend.texi (attribute packed): Correct typos.
8105 2018-10-11 Martin Sebor <msebor@redhat.com>
8107 * doc/extend.texi (attribute flatten): Mention interaction with
8110 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
8113 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8114 Set new_decl virtual flag to zero.
8116 2018-10-11 Martin Sebor <msebor@redhat.com>
8119 * doc/extend.texi (attribute format_arg): Discuss using multiple
8120 attributes on a single function.
8122 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
8124 PR tree-optimization/86829
8125 * match.pd (sin (atan (x))): New simplification rules.
8126 (cos (atan (x))): Likewise.
8127 * real.c (build_sinatan_real): New function.
8128 * real.h (build_sinatan_real): Prototype.
8130 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
8132 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
8134 (fold_mergeeo_helper): New helper function.
8135 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
8136 intrinsics. Correct some whitespace indentation issues.
8138 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
8141 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
8142 Use HOST_WIDE_INT_1U for shift.
8144 2018-10-11 Doug Rupp <rupp@adacore.com>
8145 Olivier Hainque <hainque@adacore.com>
8147 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
8148 Pass --relax to the linker for RTPs.
8149 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
8151 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
8152 Jan Hubicka <jh@suse.cz>
8153 Martin Jambor <mjambor@suse.cz>
8155 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
8156 the same elements are repeated rather than printing all of them.
8157 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
8158 "repeated" elements.
8159 * read-rtl-function.c (test_loading_repeat): New function.
8160 (read_rtl_function_c_tests): Call test_loading_repeat.
8161 * rtl-tests.c (test_dumping_repeat): New function.
8162 (rtl_tests_c_tests): Call test_dumping_repeat.
8164 2018-10-11 Richard Biener <rguenther@suse.de>
8166 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
8167 bdver?_cost): Unify to ...
8168 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
8169 * config/i386/i386.c (processor_cost_table): Adjust.
8171 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
8174 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
8175 the thunk when expanding to GIMPLE.
8177 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
8180 * varasm.c (mergeable_string_section): Don't try to move zero-length
8181 strings to the merge section.
8183 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
8186 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
8188 2018-10-10 Jakub Jelinek <jakub@redhat.com>
8191 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
8192 to special_args set.
8194 2018-10-10 Richard Biener <rguenther@suse.de>
8196 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
8197 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
8198 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
8199 and recursing and pattern terminating the recursion on SSE
8200 vector width using ix86_expand_reduc.
8201 (reduc_sminmax_scal_<mode>): Split into part reducing to half
8202 width and recursing and SSE2 vector variant doing the final
8203 reduction with ix86_expand_reduc.
8204 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
8205 with terminating the recursion at AVX level, splitting that
8208 2018-10-09 David Malcolm <dmalcolm@redhat.com>
8210 * genmatch.c (error_cb): Rename to...
8211 (diagnostic_cb): ...this, converting int params to enums.
8212 (fatal_at): Update for renaming.
8213 (warning_at): Likewise.
8215 * input.c (selftest::ebcdic_execution_charset::apply):
8216 Update for renaming of...
8217 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
8219 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
8220 converting level and reason to enums.
8221 (class selftest::lexer_error_sink): Rename to...
8222 (class selftest::lexer_test_options): ...this, renaming field
8223 "m_errors" to "m_diagnostics".
8224 (selftest::lexer_test_options::apply): Update for renaming of...
8225 (selftest::lexer_test_options::on_error): ...this, renaming to...
8226 (selftest::lexer_test_options::on_diagnostic): ...this
8227 converting level and reason to enums.
8228 (selftest::test_lexer_string_locations_raw_string_unterminated):
8229 Update for renamings.
8230 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
8233 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
8235 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
8236 * config/rs6000/pmmintrin.h: New file.
8238 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
8240 PR tree-optimization/86659
8241 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
8243 2018-10-09 Richard Biener <rguenther@suse.de>
8245 PR tree-optimization/63155
8246 * tree-ssa-structalias.c: Include tree-ssa.h.
8247 (get_constraint_for_ssa_var): For undefs return nothing_id.
8248 (find_func_aliases): Cleanup PHI handling.
8250 2018-10-09 Richard Biener <rguenther@suse.de>
8252 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
8255 2018-10-09 Martin Liska <mliska@suse.cz>
8257 * asan.c (asan_emit_stack_protection): If a stack variable
8258 is located in a same file as current function, then emit
8259 line info into variable definition string.
8261 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
8263 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
8266 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
8268 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
8271 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
8273 PR tree-optimization/86659
8274 * gimple-match.h (struct gimple_match_op): Add reverse field.
8275 (gimple_match_op::set_op): New overloaded method.
8276 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
8277 the REF_REVERSE_STORAGE_ORDER flag on the value.
8278 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
8279 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
8281 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
8284 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
8285 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
8287 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
8290 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
8291 Defined with __builtin_ia32_vfmaddsubpd512_mask.
8293 2018-10-08 Richard Biener <rguenther@suse.de>
8295 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
8296 cost the same as AVX128 ones.
8298 2018-10-08 Paul Koning <ni1d@arrl.net>
8300 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
8301 (expand_block_move): New function.
8302 * config/pdp11/pdp11.c (output_block_move): Remove.
8303 (expand_block_move): New function.
8304 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
8305 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
8306 (*movmemhi1): Remove.
8308 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
8310 * config/s390/2827.md: Increase latencies for some FP instructions.
8312 2018-10-08 Richard Biener <rguenther@suse.de>
8314 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8316 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
8317 * tree-vectorizer.h (dump_stmt_cost): Adjust.
8318 (add_stmt_cost): Dump return value of the hook.
8320 2018-10-08 Richard Biener <rguenther@suse.de>
8322 PR tree-optimization/63155
8323 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
8324 (ssa_propagation_engine::ssa_propagate): Remove redundant
8325 bitmap bit clearing.
8327 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
8329 PR rtl-optimization/86939
8330 PR rtl-optimization/87479
8331 * ira.h (non_conflicting_reg_copy_p): New prototype.
8332 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
8333 (make_hard_regno_dead): Don't add conflicts for register
8334 ignore_reg_for_conflicts.
8335 (make_object_dead): Likewise.
8336 (non_conflicting_reg_copy_p): New function.
8337 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
8338 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
8339 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
8340 (make_hard_regno_dead): Don't add conflicts for register
8341 ignore_reg_for_conflicts. Remove special conflict handling of
8342 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
8343 check_pic_pseudo_p and update callers.
8344 (mark_pseudo_dead): Don't add conflicts for register
8345 ignore_reg_for_conflicts.
8346 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
8348 2018-10-05 Andrew Waterman <andrew@sifive.com>
8349 Jim Wilson <jimw@sifive.com>
8351 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
8352 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
8353 new pattern using HONOR_SNANS that emits one extra instruction.
8355 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
8357 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
8358 patterns): Merge SI and DI patterns to a GPR pattern.
8359 (unnamed define_insn and define_split for record form of that): Merge
8360 to a single define_insn_and_split pattern.
8362 2018-10-05 David Malcolm <dmalcolm@redhat.com>
8365 * input.c (expand_location_to_spelling_point): Add param "aspect"
8366 and use rather than hardcoding LOCATION_ASPECT_CARET.
8367 (get_substring_ranges_for_loc): Handle the case of a single token
8368 within a macro expansion.
8369 * input.h (expand_location_to_spelling_point): Add "aspect" param,
8370 defaulting to LOCATION_ASPECT_CARET.
8372 2018-10-05 Paul Koning <ni1d@arrl.net>
8374 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
8375 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
8376 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
8377 (pdp11_guard_type): New function.
8379 2018-10-05 Paul Koning <ni1d@arrl.net>
8381 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
8382 * config/pdp11/pdp11.opt (mfloat32): Remove.
8384 * doc/invoke.texi (pdp11 -mfloat32): Remove:
8385 (pdp11 -mfloat64): Remove.
8387 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
8389 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
8390 (*cmp<mode>_cc_i387): Ditto.
8391 (*cmpu<mode>_cc_i387): Ditto.
8392 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
8393 * config/i386/i386.c (ix86_expand_fp_compare): Remove
8395 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
8396 Emit x86_sahf_1 pattern.
8397 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
8398 (ix86_expand_carry_flag_compare): Ditto.
8400 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
8402 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
8403 to reg_or_0_operand. Add "C" constraint.
8404 (*cmpxf_cc_i387): Ditto.
8405 (*cmp<mode>_i387): Change operand 2 predicate
8406 to nonimm_or_0_operand. Add "C" constraint.
8407 (*cmp<mode>_cc_i387): Ditto.
8408 (*cmp<mode>_0_i387): Remove insn pattern.
8409 (*cmp<mode>_0_cc_i387): Ditto.
8411 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
8413 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
8414 * config/i386/predicates.md (nonimm_or_0_operand): Rename
8415 from vector_move_operand. Update all uses.
8417 2018-10-05 Martin Sebor <msebor@redhat.com>
8419 PR tree-optimization/87490
8420 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
8423 2018-10-05 Richard Biener <rguenther@suse.de>
8425 PR tree-optimization/63155
8426 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
8427 vertical space in dumpfiles.
8428 * tree-ssa-propagate.h
8429 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
8430 * tree-ssa-propagate.c (cfg_blocks_back): New global.
8431 (ssa_edge_worklist_back): Likewise.
8432 (curr_order): Likewise.
8433 (cfg_blocks_get): Remove abstraction.
8434 (cfg_blocks_add): Likewise.
8435 (cfg_blocks_empty_p): Likewise.
8436 (add_ssa_edge): Add to current or next worklist based on
8438 (add_control_edge): Likewise.
8439 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
8441 (ssa_propagation_engine::ssa_propagate): ... here. Unify
8442 iteration from CFG and SSA edge worklist so we process
8443 everything in RPO order, prioritizing forward progress
8445 (ssa_prop_init): Allocate new worklists, do not dump
8447 (ssa_prop_fini): Free new worklists.
8449 2018-10-05 Richard Biener <rguenther@suse.de>
8451 * tree-core.h (tree_block::abstract_flag): Remove.
8452 (tree_block::block_num): Make full 32bits.
8453 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
8454 * tree.h (BLOCK_ABSTRACT): Remove.
8455 * dwarf2out.c (gen_lexical_block_die): Remove dead code
8456 resulting from BLOCK_ABSTRACT being always false.
8457 (gen_inlined_subroutine_die): Likewise.
8458 (gen_block_die): Likewise.
8459 * tree.c (block_ultimate_origin): Likewise.
8460 * tree-pretty-print.c (dump_block_node): Remove code dealing
8461 with BLOCK_ABSTRACT.
8462 * tree-ssa-live.c (dump_scope_block): Likewise.
8463 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
8464 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
8466 2018-10-05 Richard Biener <rguenther@suse.de>
8468 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
8469 is asked for initialize mode to the component mode of the
8472 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
8475 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
8476 assembler for -mavx.
8477 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
8479 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
8482 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
8484 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
8485 to be DImode. When using mffscrn, force the operand to a register.
8487 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
8489 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
8490 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
8491 X87MODEF mode iterator.
8492 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
8493 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
8494 X87MODEF mode iterator.
8496 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
8498 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
8499 -Wno-prio-ctor-dtor.
8501 2018-10-04 David Malcolm <dmalcolm@redhat.com>
8503 * Makefile.in (OBJS): Add opt-problem.o.
8504 * dump-context.h: Include "selftest.h.
8505 (selftest::temp_dump_context): New forward decl.
8506 (class dump_context): Make friend of class
8507 selftest::temp_dump_context.
8508 (dump_context::dump_loc_immediate): New decl.
8509 (class dump_pretty_printer): Move here from dumpfile.c.
8510 (class temp_dump_context): Move to namespace selftest.
8511 (temp_dump_context::temp_dump_context): Add param
8512 "forcibly_enable_dumping".
8513 (selftest::verify_dumped_text):
8514 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
8515 (selftest::verify_item):
8516 (ASSERT_IS_TEXT): Move here from dumpfile.c.
8517 (ASSERT_IS_TREE): Likewise.
8518 (ASSERT_IS_GIMPLE): Likewise.
8519 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
8521 (dump_context::dump_loc_immediate): ...this new function.
8522 (class dump_pretty_printer): Move to dump-context.h.
8523 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
8524 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
8525 (temp_dump_context::temp_dump_context): Move to "selftest"
8526 namespace. Add param "forcibly_enable_dumping", and use it to
8527 conditionalize the use of m_pp;
8528 (selftest::verify_dumped_text): Make non-static.
8529 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
8530 (selftest::verify_item): Make non-static.
8531 (ASSERT_IS_TEXT): Move to dump-context.h.
8532 (ASSERT_IS_TREE): Likewise.
8533 (ASSERT_IS_GIMPLE): Likewise.
8534 (selftest::test_capture_of_dump_calls): Pass "true" for new
8535 param of temp_dump_context.
8536 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
8537 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
8539 * opt-problem.cc: New file.
8540 * opt-problem.h: New file.
8541 * optinfo-emit-json.cc
8542 (selftest::test_building_json_from_dump_calls): Pass "true" for
8543 new param of temp_dump_context.
8544 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
8545 (optinfo::emit_for_opt_problem): New function.
8546 (optinfo::emit): Clarity which emit_item is used.
8547 * optinfo.h (optinfo::get_dump_location): New accessor.
8548 (optinfo::emit_for_opt_problem): New decl.
8549 (optinfo::emit): Make const.
8550 * selftest-run-tests.c (selftest::run_tests): Call
8551 selftest::opt_problem_cc_tests.
8552 * selftest.h (selftest::opt_problem_cc_tests): New decl.
8553 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
8554 bool to opt_result, converting fprintf messages to
8555 opt_result::failure_at calls. Add "stmt" param for use by the
8557 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
8558 (runtime_alias_check_p): Convert return type from bool to
8559 opt_result, converting dump_printf calls to
8560 opt_result::failure_at, using the statement DDR_A for their
8562 (find_data_references_in_stmt): Convert return type from bool to
8563 opt_result, converting "return false" to opt_result::failure_at
8565 * tree-data-ref.h: Include "opt-problem.h".
8566 (dr_analyze_innermost): Convert return type from bool to opt_result,
8567 and add a const gimple * param.
8568 (find_data_references_in_stmt): Convert return type from bool to
8570 (runtime_alias_check_p): Likewise.
8571 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
8572 dr_analyze_innermost.
8573 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
8574 Convert return type from bool to opt_result, adding a message for
8575 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
8576 (vect_analyze_data_ref_dependence): Convert return type from bool
8577 to opt_result. Change sense of return type from "false"
8578 effectively meaning "no problems" to "false" meaning a problem,
8579 so that "return false" becomes "return opt_result::success".
8580 Convert "return true" calls to opt_result::failure_at, using
8581 the location of statement A rather than vect_location.
8582 (vect_analyze_data_ref_dependences): Convert return type from bool
8584 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
8585 calls to opt_result::failure_at, using the stmt location rather
8587 (vect_verify_datarefs_alignment): Convert return type from bool
8589 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
8590 into multiple more-tightly-scoped copies.
8591 (vect_analyze_data_refs_alignment): Convert return type from bool
8593 (vect_analyze_data_ref_accesses): Likewise, converting a
8594 "return false" to a "return opt_result::failure_at", adding a
8596 (vect_prune_runtime_alias_test_list): Convert return type from
8597 bool to opt_result, converting dump_printf_loc to
8598 opt_result::failure_at. Add a %G to show the pertinent statement,
8599 and use the stmt's location rather than vect_location.
8600 (vect_find_stmt_data_reference): Convert return type from
8601 bool to opt_result, converting dump_printf_loc to
8602 opt_result::failure_at, using stmt's location.
8603 (vect_analyze_data_refs): Convert return type from bool to
8604 opt_result. Convert "return false" to "return
8605 opt_result::failure_at", adding messages as needed.
8606 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
8607 type from bool to opt_result.
8608 (vect_determine_vf_for_stmt): Likewise.
8609 (vect_determine_vectorization_factor): Likewise, converting
8610 dump_printf_loc to opt_result::failure_at, using location of phi
8611 rather than vect_location.
8612 (vect_analyze_loop_form_1): Convert return type from bool to
8613 opt_result, converting dump_printf_loc calls, retaining the use of
8615 (vect_analyze_loop_form): Convert return type from loop_vec_info
8616 to opt_loop_vec_info.
8617 (vect_analyze_loop_operations): Convert return type from bool to
8618 opt_result, converting dump_printf_loc calls, using the location
8619 of phi/stmt rather than vect_location where available. Convert
8620 various "return false" to "return opt_result::failure_at" with
8621 "unsupported phi" messages.
8622 (vect_get_datarefs_in_loop): Convert return type from bool to
8623 opt_result. Add a message for the
8624 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
8625 (vect_analyze_loop_2): Convert return type from bool to
8626 opt_result. Ensure "ok" is set to a opt_result::failure_at before
8627 each "goto again;", adding new messages where needed.
8628 Add "unsupported grouped {store|load}" messages.
8629 (vect_analyze_loop): Convert return type from loop_vec_info to
8631 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
8633 * tree-vect-stmts.c (process_use): Likewise, converting
8634 dump_printf_loc call and using stmt location, rather than
8636 (vect_mark_stmts_to_be_vectorized): Likeise.
8637 (vect_analyze_stmt): Likewise, adding a %G.
8638 (vect_get_vector_types_for_stmt): Convert return type from bool to
8639 opt_result, converting dump_printf_loc calls and using stmt
8640 location, rather than vect_location.
8641 (vect_get_mask_type_for_stmt): Convert return type from tree to
8642 opt_tree, converting dump_printf_loc calls and using stmt location.
8643 * tree-vectorizer.c: Include "opt-problem.h.
8644 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
8645 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
8646 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
8647 enabled, use it to report at the top level "couldn't vectorize
8648 loop" followed by the problem.
8649 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
8650 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
8652 (vect_analyze_stmt): Likewise.
8653 (vect_get_vector_types_for_stmt): Likewise.
8654 (tree vect_get_mask_type_for_stmt): Likewise.
8655 (vect_analyze_data_ref_dependences): Likewise.
8656 (vect_enhance_data_refs_alignment): Likewise.
8657 (vect_analyze_data_refs_alignment): Likewise.
8658 (vect_verify_datarefs_alignment): Likewise.
8659 (vect_analyze_data_ref_accesses): Likewise.
8660 (vect_prune_runtime_alias_test_list): Likewise.
8661 (vect_find_stmt_data_reference): Likewise.
8662 (vect_analyze_data_refs): Likewise.
8663 (vect_analyze_loop): Convert return type from loop_vec_info to
8665 (vect_analyze_loop_form): Likewise.
8666 (vect_analyze_slp): Convert return type from bool to opt_result.
8668 2018-10-04 David Malcolm <dmalcolm@redhat.com>
8670 * doc/invoke.texi (-fopt-info): Document new "internals"
8672 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
8673 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
8675 (optinfo_verbosity_options): Add "internals".
8676 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
8677 (dump_context::apply_dump_filter_p): New member function.
8678 (dump_context::dump_loc): Use apply_dump_filter_p rather than
8679 explicitly masking the dump_kind.
8680 (dump_context::begin_scope): Increment the scope depth first. Use
8681 apply_dump_filter_p rather than explicitly masking the dump_kind.
8682 (dump_context::emit_item): Use apply_dump_filter_p rather than
8683 explicitly masking the dump_kind.
8684 (dump_dec): Likewise.
8685 (dump_hex): Likewise.
8686 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
8687 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
8688 (opt_info_switch_p): Update handling of default
8689 MSG_OPTIMIZED_LOCATIONS to cope with default of
8690 MSG_PRIORITY_USER_FACING.
8691 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
8692 masking the dump_kind.
8693 (selftest::test_capture_of_dump_calls): Update test_dump_context
8694 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
8695 than MSG_ALL. Generalize scope test to be run at all four
8696 combinations of with/without MSG_PRIORITY_USER_FACING and
8697 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
8698 for each of the two values.
8699 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
8700 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
8701 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
8702 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
8703 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
8704 with MSG_PRIORITY_*.
8705 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
8706 dump messages as MSG_PRIORITY_USER_FACING.
8707 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
8708 about the interaction with MSG_PRIORITY_*.
8710 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
8712 * varasm.c (output_constant): Add new parameter merge_strings.
8713 Make strings properly zero terminated in merge string sections.
8714 (mergeable_string_section): Don't fail if the last char is non-zero.
8715 (assemble_variable_contents): Handle merge string sections.
8716 (assemble_variable): Likewise.
8717 (assemble_constant_contents): Likewise.
8718 (output_constant_def_contents): Likewise.
8719 (output_constructor_array_range,
8720 output_constructor_regular_field): Adjust call to output_constant.
8721 (output_object_block): Adjust call to assemble_constant_contents
8722 and assemble_variable_contents.
8724 2018-10-04 Martin Liska <mliska@suse.cz>
8727 * cgraphunit.c (process_function_and_variable_attributes):
8728 Warn about a function with alias attribute and a body.
8730 2018-10-04 Martin Liska <mliska@suse.cz>
8733 * multiple_target.c (redirect_to_specific_clone): New function.
8734 (ipa_target_clone): Use it.
8735 * tree-inline.c: Fix comment.
8737 2018-10-04 David Malcolm <dmalcolm@redhat.com>
8739 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
8741 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
8742 (gcc::dump_manager::register_pass): New member function, adapted
8743 from loop body in gcc::pass_manager::register_pass, adding a
8744 call to update_dfi_for_opt_info.
8745 (gcc::dump_manager::opt_info_enable_passes): Store the
8746 -fopt-info options into the new fields. Move the loop
8748 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
8750 * dumpfile.h (struct opt_pass): New forward decl.
8751 (gcc::dump_manager::register_pass): New decl.
8752 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
8753 (class gcc::dump_manager): Add fields "m_optgroup_flags",
8754 "m_optinfo_flags", and "m_optinfo_filename".
8755 * passes.c (gcc::pass_manager::register_pass): Move all of the
8756 dump-handling code to gcc::dump_manager::register_pass.
8758 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
8760 PR rtl-optimization/87466
8761 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
8762 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
8763 * doc/tm.texi: Regenerate.
8764 * ira-lives.c (process_bb_node_lives): Use the new target hook.
8765 * lra-lives.c (process_bb_lives): Likewise.
8766 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
8769 2018-10-04 Tamar Christina <tamar.christina@arm.com>
8771 * params.c (add_params): Fix initialization.
8773 2018-10-04 Martin Liska <mliska@suse.cz>
8775 PR gcov-profile/84107
8776 * tree-profile.c (init_ic_make_global_vars):
8777 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
8778 Come up with new ic_tuple* variables. Emit
8779 __gcov_indirect_call{,_topn} variables.
8780 (gimple_gen_ic_profiler): Access the variable
8782 (gimple_gen_ic_func_profiler): Access
8783 __gcov_indirect_call.callee field.
8784 (gimple_init_gcov_profiler): Use ptr_type_node.
8785 * value-prof.c (gimple_ic): Use ptr_type_node.
8787 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8789 PR tree-optimization/85787
8790 * ipa-pure-const.c (malloc_candidate_p_1): Move most of
8791 malloc_candidate_p into this function and add support for
8792 detecting multiple phis.
8793 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
8795 2018-10-04 Martin Liska <mliska@suse.cz>
8798 * ipa-inline.c (inline_to_all_callers_1):
8799 Call ultimate_alias_target for node being inlined.
8801 2018-10-03 Jeff Law <law@redhat.com>
8803 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
8805 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
8806 * tree.h (get_typenode_from_name): Prototype.
8808 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
8810 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
8811 Change operand 2 predicate to nonimmediate_operand.
8812 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
8814 2018-10-03 Martin Sebor <msebor@redhat.com>
8815 Jeff Law <law@redhat.com>
8817 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
8819 (get_string_length): Detect unterminated arrays.
8820 (format_string): Same.
8821 (format_directive): Warn about unterminated arrays.
8822 (handle_gimple_call): Mark statements with no_warning as needed.
8824 2018-10-03 Jim Wilson <jimw@sifive.com>
8826 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
8827 also define __riscv_abi_rve. Delete trailing white space.
8829 2018-10-03 Paul Koning <ni1d@arrl.net>
8831 Enable LRA register allocator for PDP11.
8832 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
8835 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
8836 * config/pdp11/pdp11.opt (-mlra): New option.
8837 * doc/invoke.texi (PDP-11 Options): Document -mlra.
8839 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
8841 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
8842 (*<absneg:code>extend<mode>xf2): Ditto.
8844 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
8846 PR tree-optimization/87415
8847 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
8850 2018-10-02 Jeff Law <law@redhat.com>
8852 * gimple-fold.c (get_range_strlen): Only set *nonstr when
8853 an unterminated string is discovered. Bubble up range
8854 even for unterminated strings.
8855 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
8856 indicates the string was not terminated via NONSTR.
8858 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
8860 * tree-vrp.c (extract_range_from_unary_expr): Special case all
8861 pointer conversions.
8862 Do not do anything special for anti-ranges.
8864 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
8866 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
8867 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
8869 2018-10-03 Martin Liska <mliska@suse.cz>
8871 PR gcov-profile/86109
8872 * coverage.c (coverage_begin_function): Do not
8873 mark lambdas as artificial.
8874 * tree-core.h (struct GTY): Remove tm_clone_flag
8875 and introduce new lambda_function.
8876 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
8878 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
8881 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
8882 P8_VECTOR and VSX are enabled.
8884 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
8886 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
8887 0x3907 as CPU model number.
8889 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
8891 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
8892 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
8893 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
8894 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
8895 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
8896 * config/s390/s390.md: Likewise. Rename also the cpu attribute
8897 value from arch12 to z14.
8899 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
8901 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
8903 (isinf<mode>2): Ditto.
8905 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
8907 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
8908 before emitting fxam. Perform calculations in XFmode.
8910 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
8912 * match.pd (((X /[ex] A) +- B) * A): New transformation.
8914 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
8917 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
8918 * tree.c (signed_or_unsigned_type_for): Handle complex.
8920 2018-10-02 Jeff Law <law@redhat.com>
8922 * gimple-fold.c (get_range_strlen): Remove dead code.
8924 2018-10-02 Martin Sebor <msebor@redhat.com>
8925 Jeff Law <law@redhat.com>
8927 * builtins.c (unterminated_array): Add new arguments.
8928 If argument is not terminated, bubble up size and exact
8930 (expand_builtin_strnlen): Detect, avoid expanding
8931 and diagnose unterminated arrays.
8932 (c_strlen): Fill in offset of start of unterminated strings.
8933 * builtins.h (unterminated_array): Update prototype.
8935 2018-10-02 Richard Biener <rguenther@suse.de>
8937 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
8938 of haddv4df, first reduce to SSE width and exploit the fact
8939 that we only need element zero with the reduction result.
8940 (reduc_plus_scal_v2df): Likewise.
8942 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
8944 * dojump.h (do_jump): Delete.
8945 (do_jump_1): Likewise.
8946 (split_comparison): Move around.
8947 * dojump.c (do_jump): Make static.
8948 (do_jump_1): Likewise.
8949 (jumpifnot): Move around.
8950 (jumpifnot_1): Likewise.
8952 (jumpif_1): Likewise.
8953 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
8955 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
8957 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
8958 insns in the delay slot and add_insn_after for the jump insn.
8960 2018-10-02 Richard Biener <rguenther@suse.de>
8962 * tree-inline.c (expand_call_inline): Use the location of
8963 the callee declaration for the inline-entry marker.
8964 * final.c (notice_source_line): Remove special-casing of
8965 NOTE_INSN_INLINE_ENTRY.
8967 2018-10-01 Carl Love <cel@us.ibm.com>
8970 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
8971 (__builtin_mtfsb0): New.
8972 (__builtin_mtfsb1): New.
8973 ( __builtin_set_fpscr_rn): New.
8974 (__builtin_set_fpscr_drn): New.
8975 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
8976 (rs6000_expand_set_fpscr_rn_builtin): Add.
8977 (rs6000_expand_set_fpscr_drn_builtin): Add.
8978 (rs6000_expand_builtin): Add case statement entries for
8979 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
8980 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
8981 RS6000_BUILTIN_MFFSL.
8982 (rs6000_init_builtins): Add ftype initialization and def_builtin
8983 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
8984 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
8985 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
8986 rs6000_mffscdrn): Add define_insn.
8987 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
8988 * doc/extend.texi: Add documentation for the builtins.
8990 2018-10-01 Richard Biener <rguenther@suse.de>
8992 PR tree-optimization/87465
8993 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
8994 causing branch miscounts.
8996 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8998 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
8999 aarch64_option_default_param): New.
9000 (params.h): Include.
9001 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
9002 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
9003 stack-clash protection validation code.
9005 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9007 * params.c (validate_param): New.
9008 (add_params): Use it.
9009 (set_param_value): Refactor param validation into validate_param.
9010 (diagnostic.h): Include.
9011 * diagnostic.h (diagnostic_ready_p): New.
9013 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9015 * params.c (set_param_value):
9016 Add index of parameter being validated.
9017 * common/common-target.def (option_validate_param): New.
9018 * common/common-targhooks.h (default_option_validate_param): New.
9019 * common/common-targhooks.c (default_option_validate_param): New.
9020 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
9021 * doc/tm.texi: Regenerate.
9023 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9026 * config/aarch64/aarch64.c (aarch64_override_options_internal):
9027 Add validation for stack-clash parameters and set defaults.
9029 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9032 * configure.ac: Add stack-clash-protection-guard-size.
9033 * doc/install.texi: Document it.
9034 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
9035 * params.def: Update comment for guard-size.
9036 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
9037 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
9038 * configure: Regenerate.
9040 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9043 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
9044 STACK_DYNAMIC_OFFSET): New.
9045 * config/aarch64/aarch64.c (aarch64_layout_frame):
9046 Update outgoing args size.
9047 (aarch64_stack_clash_protection_alloca_probe_range,
9048 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
9050 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9053 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
9055 * target.def (stack_clash_protection_alloca_probe_range): New.
9056 (stack_clash_protection_final_dynamic_probe): Remove.
9057 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
9058 (default_stack_clash_protection_final_dynamic_probe): Remove.
9059 * targhooks.c: Likewise.
9060 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
9061 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
9062 * doc/tm.texi: Regenerate.
9064 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9067 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
9068 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
9069 aarch64_clamp_to_uimm12_shift): New.
9070 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
9071 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
9073 2018-10-01 Tamar Christina <tamar.christina@arm.com>
9076 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
9078 2018-10-01 Jeff Law <law@redhat.com>
9079 Richard Sandiford <richard.sandiford@linaro.org>
9080 Tamar Christina <tamar.christina@arm.com>
9083 * config/aarch64/aarch64.md
9084 (probe_stack_range): Add k (SP) constraint.
9085 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
9086 STACK_CLASH_MAX_UNROLL_PAGES): New.
9087 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
9088 stack probes for stack clash.
9089 (aarch64_allocate_and_probe_stack_space): New.
9090 (aarch64_expand_prologue): Use it.
9091 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
9092 (aarch64_sub_sp): Add emit_move_imm optional param.
9094 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
9096 PR tree-optimization/87261
9097 * match.pd: Remove trailing whitespace.
9098 Add (x & y) | ~(x | y) -> ~(x ^ y),
9099 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
9101 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
9103 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
9105 (ashlsi3_insn): Update instruction constraints.
9106 (ashrsi3_insn): Likewise.
9107 (rotrsi3): Likewise.
9108 (add_shift): Likewise.
9109 * config/arc/constraints.md (Csz): New 32 bit constraint. It
9110 avoids placing in the limm field small constants which, otherwise,
9111 could end into a small instruction.
9113 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
9115 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
9116 destination register is not odd-even.
9117 (umaddsidi4_split): Likewise.
9119 2018-10-01 Richard Biener <rguenther@suse.de>
9121 * tree-inline.c (expand_call_inline): Store origin of fn
9122 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
9123 * tree.c (block_ultimate_origin): Simplify and do some
9126 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
9128 * config/i386/mmx.md (EMMS): New int iterator.
9129 (emms): New int attribute.
9130 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
9131 EMMS int iterator. Explicitly declare clobbers.
9132 (mmx_emms): Remove expander.
9134 * config/i386/predicates.md (emms_operation): Remove predicate.
9135 (vzeroall_pattern): New predicate.
9136 (vzeroupper_pattern): Rename from vzeroupper_operation.
9137 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
9138 vzeroupper_pattern and vzeroall_pattern predicates.
9140 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
9142 PR rtl-optimization/86939
9143 * ira-lives.c (make_hard_regno_born): Rename from this...
9144 (make_hard_regno_live): ... to this. Remove update to conflict
9145 information. Update function comment.
9146 (make_hard_regno_dead): Add conflict information update. Update
9148 (make_object_born): Rename from this...
9149 (make_object_live): ... to this. Remove update to conflict information.
9150 Update function comment.
9151 (make_object_dead): Add conflict information update. Update function
9153 (mark_pseudo_regno_live): Call make_object_live.
9154 (mark_pseudo_regno_subword_live): Likewise.
9155 (mark_hard_reg_dead): Update function comment.
9156 (mark_hard_reg_live): Call make_hard_regno_live.
9157 (process_bb_node_lives): Likewise.
9158 * lra-lives.c (make_hard_regno_born): Rename from this...
9159 (make_hard_regno_live): ... to this. Remove update to conflict
9160 information. Remove now uneeded check_pic_pseudo_p argument.
9161 Update function comment.
9162 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
9163 to conflict information. Update function comment.
9164 (mark_pseudo_live): Remove update to conflict information. Update
9166 (mark_pseudo_dead): Add conflict information update.
9167 (mark_regno_live): Call make_hard_regno_live.
9168 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
9169 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
9171 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
9174 * config/i386/i386.c (construct_container): Use TImode for
9175 BLKmode values in 2 integer registers.
9177 2018-09-29 Jeff Law <law@redhat.com>
9179 * builtins.c (unterminated_array): Pass in c_strlen_data * to
9180 c_strlen rather than just a tree *.
9181 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
9182 Update recursive calls appropriately. If caller did not provide a
9183 suitable data pointer, create a local one. When a non-terminated
9184 string is discovered, bubble up information about the string via the
9185 c_strlen_data object.
9186 * builtins.h (c_strlen): Update prototype.
9187 (c_strlen_data): New structure.
9188 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
9189 For a type 2 call, if c_strlen indicates a non-terminated string
9190 use the length of the non-terminated string.
9191 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
9193 2018-09-29 Jakub Jelinek <jakub@redhat.com>
9196 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
9197 __m512d type for __A argument rather than __m512.
9199 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
9201 * match.pd (simple_comparison): Don't optimize if either operand is
9202 a function pointer when target needs function pointer canonicalization.
9204 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
9206 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
9207 power5 .. power9 to remove indirection.
9208 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
9209 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
9210 ASM_CPU_476_SPEC): Delete.
9211 (ASM_CPU_SPEC): Adjust.
9212 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
9213 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
9215 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
9217 * config.in: Delete HAVE_AS_DCI.
9218 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
9219 * config/rs6000/rs6000.h: Ditto.
9220 * configure.ac: Delete HAVE_AS_DCI.
9221 * configure: Regenerate.
9223 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
9225 * config.in (HAVE_AS_LWSYNC): Delete.
9226 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
9227 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
9229 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
9230 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
9232 * configure.ac: Delete HAVE_AS_LWSYNC.
9233 * configure: Regenerate.
9235 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
9236 Pierre-Marie de Rodat <derodat@adacore.com>
9238 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
9239 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
9240 (cgraph_node::create_thunk): Add indirect_offset parameter.
9241 (thunk_adjust): Likewise.
9242 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
9243 and initialize the corresponding field with it.
9244 (cgraph_node::dump): Dump indirect_offset field.
9245 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
9246 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
9247 (thunk_adjust): Add indirect_offset parameter and deal with it.
9248 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
9249 pass it to thunk_adjust. Do not call the target hook if it's non-zero
9250 or if the thunk is external or local. Fix formatting. Do not chain
9251 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
9252 if any, in the GIMPLE representation.
9253 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
9254 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
9255 (input_node): Read indirect_offset field.
9256 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
9257 call to thunk_adjust.
9258 * tree-nested.c (struct nesting_info): Add thunk_p field.
9259 (create_nesting_tree): Set it.
9260 (convert_all_function_calls): Copy static chain from targets to thunks.
9261 (finalize_nesting_tree_1): Return early for thunks.
9262 (unnest_nesting_tree_1): Do not finalize thunks.
9263 (gimplify_all_functions): Do not gimplify thunks.
9265 2018-09-28 David Malcolm <dmalcolm@redhat.com>
9267 * opt-suggestions.c (option_proposer::build_option_suggestions):
9268 Release "option_values".
9270 2018-09-28 David Malcolm <dmalcolm@redhat.com>
9272 * coverage.c (get_coverage_counts): Convert problem-reporting dump
9273 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
9274 * dumpfile.c (kind_as_string): New function.
9275 (dump_loc): Rather than a hardcoded prefix of "note: ", use
9276 kind_as_string to vary the prefix based on dump_kind.
9277 (selftest::test_capture_of_dump_calls): Update for above.
9279 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
9281 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
9282 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
9284 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
9286 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
9287 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
9288 INVALID_REGNUM instead of FPSR_REG.
9289 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
9290 * config/i386/i386.md: Update comment of FP compares.
9291 (fldenv): Do not clobber FPSR_REG.
9293 2018-09-28 Richard Biener <rguenther@suse.de>
9295 * tree.h (BLOCK_ORIGIN): New.
9296 * omp-expand.c (grid_expand_target_grid_body): Assign
9297 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
9298 * tree-inline.c (remap_block): Likewise.
9299 * auto-profile.c (get_function_decl_from_block): Simplify
9300 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
9301 * langhooks.c (lhd_print_error_function): Likewise.
9302 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
9304 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9305 * tree.c (block_nonartificial_location): Likewise.
9306 (block_ultimate_origin): Likewise.
9307 * tree-pretty-print.c (percent_K_format): Likewise. Remove
9308 no longer needed LTO case.
9310 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
9311 Jan Hubicka <jh@suse.cz>
9312 Martin Jambor <mjambor@suse.cz>
9314 * simplify-rtx.c (simplify_merge_mask): New function.
9315 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
9316 same masks are used in op1 or op2.
9317 (test_vec_merge): New function.
9318 (test_vector_ops): Call test_vec_merge.
9320 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
9322 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
9323 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
9324 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
9325 (cypress_costs): Set it.
9326 (supersparc_costs): Likewise.
9327 (hypersparc_costs): Likewise.
9328 (leon_cost): Likewise.
9329 (leon3_costs): Likewise.
9330 (sparclet_costs): Likewise.
9331 (ultrasparc_costs): Likewise.
9332 (ultrasparc_costs): Likewise.
9333 (niagara_costs): Likewise.
9334 (niagara2_costs): Likewise.
9335 (niagara3_costs): Likewise.
9336 (niagara4_costs): Likewise.
9337 (niagara7_costs): Likewise.
9338 (m8_costs): Likewise.
9339 (TARGET_CAN_FOLLOW_JUMP): Define.
9340 (pass_work_around_errata::gate): Minor tweak.
9341 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
9342 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
9343 Automaitcally clear MASK_FSMULD mask for V7 processors.
9344 (sparc_can_follow_jump): New static function.
9345 (output_ubranch): Deal with CROSSING_JUMP_P.
9346 (sparc_use_sched_lookahead): Rewrite using switch statement.
9347 (sparc_issue_rate): Reorder.
9348 (sparc_branch_cost): New function.
9350 2018-09-27 Martin Sebor <msebor@redhat.com>
9352 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
9353 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
9354 (int_fits_type_p): Same.
9356 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
9358 * config/i386/i386.md (FPCR_REG): Remove.
9359 (UNSPEC_FLDCW): Remove.
9360 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
9361 (x86_fldcw_1): Remove insn pattern.
9362 (fnstenv): Do not clobber FPCR_REG.
9364 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
9365 (CALL_USED_REGISTERS): Ditto.
9366 (REG_ALLOC_ORDER): Ditto.
9367 (REG_CLASS_CONTENTS): Ditto.
9368 (HI_REGISTER_NAMES): Ditto.
9369 (ADDITIONAL_REGISTER_NAMES): Use defines instead
9370 of numerical constants.
9371 * config/i386/i386.c (regclass_map): Remove fpsr register.
9372 (dbx_register_map): Ditto.
9373 (dbx64_register_map): Ditto.
9374 (svr4_dbx_register_map): Ditto.
9375 (print_reg): Do not handle FPCR_REG.
9377 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
9380 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
9381 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
9382 Delete, always treat as true.
9383 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9384 Ditto. Simplify remaining code.
9385 * config/powerpcspe/powerpcspe.h: Ditto.
9386 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
9387 Simplify remaining code.
9388 (rs6000_expand_builtin): Ditto.
9389 * config/rs6000/rs6000.h: Ditto.
9390 * configure.ac: Ditto.
9391 * configure: Regenerate.
9393 2018-09-27 Martin Liska <mliska@suse.cz>
9395 * coverage.c (get_coverage_counts): Revert the formatting
9396 of missing profile opt info.
9398 2018-09-27 Richard Biener <rguenther@suse.de>
9402 * dwarf2out.c (set_block_origin_self): Do not mark outermost
9403 block as we do not output that.
9404 (gen_inlined_subroutine_die): Elide the originally outermost
9405 block, matching what we do for concrete instances.
9406 (decls_for_scope): Add parameter specifying whether to recurse
9409 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
9410 Tom de Vries <tom@codesourcery.com>
9414 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
9415 STORE_FLAG_VALUE == 1.
9417 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
9419 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
9420 constant definitions.
9421 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
9422 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
9423 ("speculation_barrier"): New expander definition.
9425 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
9427 PR gcov-profile/86957
9428 * common.opt: New warning option -Wmissing-profile.
9429 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
9430 * doc/invoke.texi: Document -Wmissing-profile.
9432 2018-09-26 Jim Wilson <jimw@sifive.com>
9434 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
9435 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
9437 2018-09-26 Martin Sebor <msebor@redhat.com>
9439 * tree.c (zerop): Change return type to bool.
9440 (integer_zerop, integer_onep, integer_each_onep): Same.
9441 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
9442 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
9443 (real_onep, real_minus_onep, chain_index): Same.
9444 (print_type_hash_statistics, type_list_equal): Same.
9445 * tree.h (zerop): Same.
9446 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
9447 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
9448 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
9449 (real_onep, real_minus_onep, chain_index): Same.
9450 (print_type_hash_statistics, type_list_equal): Same.
9452 2018-09-26 Jim Wilson <jimw@sifive.com>
9454 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
9456 2018-09-26 Jakub Jelinek <jakub@redhat.com>
9459 * config/i386/i386.c: Include debug.h and dwarf2out.h.
9460 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
9463 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
9465 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
9467 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
9469 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
9470 and FP_SECOND_SSE_REGS.
9471 (REG_CLASS_NAMES): Ditto.
9472 (REG_CLASS_CONTENTS): Ditto.
9473 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
9474 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
9475 (ix86_preferred_output_reload_class): Ditto.
9476 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
9477 clobber constraint to "=&f".
9478 (fix_truncdi_i387): Ditto.
9479 (lrintxfdi2): Ditto.
9480 (fistdi2_<rounding>): Ditto.
9481 (fpremxf4_i387): Change "=u" constraint to "=f".
9482 (fprem1xf4_i387): Ditto.
9484 (fptanxf4_i387): Ditto.
9485 (fxtractxf3_i387): Ditto.
9486 (fscalexf4_i387): Ditto.
9487 (atan2xf3): Change "u" constraint to "f".
9488 (fyl2xxf3_i387): Ditto.
9489 (fyl2xp1xf3_i387): Ditto.
9491 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
9494 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
9495 for removed I387_MASK_PM entity.
9498 2018-09-26 Jeff Law <law@redhat.com>
9500 2018-09-26 Alexey Neyman <stilor@att.net>
9502 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
9503 headers are no longer pulled in by <isl/val.h>.
9505 2018-09-26 Richard Biener <rguenther@suse.de>
9508 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
9509 or concrete instance DIE to the tree. Create abstract origin
9510 attributes also for concrete instances.
9512 2018-09-26 Alexey Neyman <stilor@att.net>
9514 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
9515 headers are no longer pulled in by <isl/val.h>.
9517 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
9519 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
9520 Use new helper functions.
9521 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
9522 Use new helper functions.
9523 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
9524 aarch_mm_needs_release): New declarations.
9525 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
9526 aarch_mm_needs_release): New.
9528 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
9530 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
9531 (arm32_output_mi_thunk): Deal with long calls.
9533 2018-09-26 Richard Biener <rguenther@suse.de>
9537 * tree-inline.c (expand_call_inline): When the location
9538 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
9539 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
9540 the inserted BLOCK to make inlined_function_outer_scope_p
9542 * dwarf2out.c (add_call_src_coords_attributes): Do not add
9543 coords for reserved locations.
9545 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
9547 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
9548 (*call_indirect_nonlocal_sysv<mode>): Ditto.
9549 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
9550 (*sibcall_nonlocal_sysv<mode>): Ditto.
9551 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
9552 (<bd>_<mode>): Ditto.
9553 (<bd>tf_<mode>): Ditto.
9555 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
9557 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
9558 control string as a list of templates instead of as C code.
9559 (*altivec_movti): Ditto.
9560 * config/rs6000/darwin.md (movdf_low_di): Ditto.
9562 2018-09-25 Jim Wilson <jimw@sifive.com>
9564 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
9565 when target symbol is weak.
9567 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
9570 * builtins.c (unterminated_array): Simplify.
9571 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
9572 where pointer arithmetic is safe.
9574 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
9577 * config/rs6000/altivec.md (altivec_vspltb): Use
9578 const_0_to_15_operand instead of u5bit_cint_operand.
9579 (*altivec_vspltb_internal): Ditto.
9580 (altivec_vspltb_direct): Ditto.
9581 (altivec_vsplth): Use const_0_to_7_operand instead of
9583 (*altivec_vsplth_internal): Ditto.
9584 (altivec_vsplth_direct): Ditto.
9585 (altivec_vspltw): Use const_0_to_3_operand instead of
9587 (*altivec_vspltw_internal): Ditto.
9588 (altivec_vspltw_direct): Ditto.
9589 (altivec_vspltsf): Ditto.
9590 (*altivec_vspltsf_internal): Ditto.
9591 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
9592 various splats with the proper size immediate. Reorder the various
9593 cases by ascending size of immediate, and put all such together.
9595 2018-09-25 Richard Biener <rguenther@suse.de>
9598 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
9600 (maybe_create_die_with_external_ref): Do not create
9601 DW_TAG_imported_unit here.
9602 (add_abstract_origin_attribute): Handle external BLOCK refs.
9603 (dwarf2out_abstract_function): Simplify LTO case.
9604 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
9605 rather than using maybe_create_die_with_external_ref.
9607 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
9610 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
9611 (frndintxf2_mask_pm_i387): Ditto.
9612 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
9613 Enable for !flag_trapping_math.
9614 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
9615 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
9616 Change operand 1 predicate to nonimmediate_operand.
9617 (attr "i387_cw"): Remove mask_pm.
9618 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
9619 (enum ix86_entity): Remove I387_MASK_PM.
9620 * config/i386/i386.c (ix86_i387_mode_needed): Do not
9621 handle I387_MASK_PM.
9622 (ix86_mode_needed): Ditto.
9623 (ix86_mode_after): Ditto.
9624 (ix86_mode_entry): Ditto.
9625 (ix86_mode_exit): Ditto.
9626 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
9628 2018-09-25 Jakub Jelinek <jakub@redhat.com>
9630 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
9631 to_update_switch_stmts to vNULL instead of calling create on them
9634 2018-09-25 Richard Biener <rguenther@suse.de>
9636 PR tree-optimization/87402
9637 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
9638 (visit_phi): Re-instantiate handling of supposed to be VARYING
9639 but non-VARYING backedge value.
9641 2018-09-25 Richard Biener <rguenther@suse.de>
9644 * dwarf2out.c (struct sym_off_pair): New.
9645 (external_die_map): New global.
9646 (lookup_decl_die): When in LTO create DIEs lazily from the
9648 (lookup_block_die): New function, create DIEs lazily in LTO.
9649 (equate_block_to_die): New function.
9650 (dwarf2out_die_ref_for_decl): During WPA get the association
9651 from the external DIE map.
9652 (dwarf2out_register_external_die): Record mapping into the
9654 (maybe_create_die_with_external_ref): New function split out from
9655 DIE generation part of old dwarf2out_register_external_die.
9656 (add_abstract_origin_attribute): Do not return the DIE. When
9657 in LTO reference externals directly.
9658 (dwarf2out_abstract_function): When in LTO ignore calls for
9659 decls with external DIEs (already present abstract instances).
9660 (gen_call_site_die): Adjust.
9661 (add_high_low_attributes): Likewise.
9662 (gen_lexical_block_die): Likewise.
9663 (gen_inlined_subroutine_die): Likewie.
9664 (gen_block_die): Likewise.
9665 (dwarf2out_inline_entry): Likewise.
9666 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
9669 2018-09-25 Martin Liska <mliska@suse.cz>
9671 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
9672 integers and not by a float value.
9674 2018-09-25 Martin Liska <mliska@suse.cz>
9677 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
9679 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
9680 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
9681 * godump.c (go_format_type): Remove extra parenthesis.
9683 2018-09-25 Martin Liska <mliska@suse.cz>
9685 * alias.c (set_dest_equal_p): Remove unused function.
9686 * config/i386/i386.c (def_builtin_pure2): Likewise.
9687 * diagnostic-show-locus.c (class layout): Remove
9689 (layout::layout): Likewise here.
9690 * dump-context.h (class temp_dump_context): Likewise.
9691 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
9692 (add_AT_loclistsptr): Likewise.
9693 (add_AT_offset): Likewise.
9694 (get_AT_hi_pc): Likewise.
9695 (is_comdat_die): Likewise.
9696 (type_is_enum): Likewise.
9697 (ceiling): Likewise.
9698 (add_AT_vms_delta): Likewise.
9699 (is_class_die): Likewise.
9700 * edit-context.c (class line_event): Remove unused field.
9701 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
9703 * ipa-cp.c (ipa_get_vr_lat): Likewise.
9704 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
9705 (ok_for_base_p_nonstrict): Likewise.
9706 * tree-chrec.c (is_not_constant_evolution): Likewise.
9707 (chrec_fold_poly_cst): Likewise.
9708 * tree-if-conv.c (has_pred_critical_p): Likewise.
9709 * tree-ssa-coalesce.c (print_exprs): Likewise.
9710 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
9711 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
9712 * tree-vrp.c (value_ranges_intersect_p): Likewise.
9713 (value_range_nonnegative_p): Likewise.
9715 2018-09-25 Martin Liska <mliska@suse.cz>
9717 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
9718 Do not handle "GNU Pascal".
9719 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
9721 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
9722 from documentation. Likewise.
9723 * dbxout.c (dbxout_range_type): Likewise.
9724 * doc/cpp.texi: Likewise.
9725 * doc/extend.texi: Likewise.
9726 * doc/frontends.texi: Likewise.
9727 * doc/invoke.texi: Remove Pascal entry.
9728 * tree.def (CLEANUP_POINT_EXPR): Likewise.
9729 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
9731 2018-09-25 Martin Liska <mliska@suse.cz>
9734 * doc/invoke.texi: Document all parameters and remove default
9737 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
9740 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
9741 contains HOST_WIDE_INTs when computing its size.
9743 2018-09-24 Jim Wilson <jimw@sifive.com>
9746 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
9748 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
9750 2018-09-24 Andrew Pinski <apinski@marvell.com>
9752 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
9753 access prev before checking it for NULLness in the
9754 AARCH64_FUSE_CMP_BRANCH case.
9756 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
9759 * config/i386/i386.c (rest_of_insert_endbranch): Set
9760 endbr_queued_at_entrance to true and don't insert ENDBR if
9761 x86_function_profiler will be called.
9762 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
9764 * config/i386/i386.h (machine_function): Add
9765 endbr_queued_at_entrance.
9767 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
9769 * genattrtab.c (mk_attr_alt): Use alternative_mask.
9770 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
9772 (check_attr_test): Use alternative_mask.
9773 (get_attr_value): Likewise.
9774 (compute_alternative_mask): Use alternative_mask and XWINT.
9775 (make_alternative_compare): Use alternative_mask.
9776 (attr_alt_subset_p): Use XWINT.
9777 (attr_alt_subset_of_compl_p): Likewise.
9778 (attr_alt_intersection): Use alternative_mask and XWINT.
9779 (attr_alt_union): Likewise.
9780 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
9781 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
9782 (simplify_test_exp): Use alternative_mask and XWINT.
9783 (write_test_expr): Use alternative_mask and XWINT, adjust bit
9784 number calculation to support 64 bits. Generate code that
9785 checks 64-bit masks.
9786 (main): Use alternative_mask.
9787 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
9789 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
9792 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
9793 RETURN+USE when returning via %r14.
9795 2018-09-24 Martin Liska <mliska@suse.cz>
9797 * gcov.c (output_lines): Print colorization legend
9798 for both flag_use_colors and flag_use_hotness_colors.
9801 2018-09-24 Martin Liska <mliska@suse.cz>
9803 * coverage.c (get_coverage_counts): Use warning_at
9804 with current_function_decl location. Use %qD in warning
9807 2018-09-24 Martin Liska <mliska@suse.cz>
9809 * memory-block.h (memory_block_pool::release): Annotate with
9810 valgrind that the memory is not accessible.
9812 2018-09-24 Martin Liska <mliska@suse.cz>
9815 * asan.c: Make asan_handled_variables extern.
9817 * cfgexpand.c (expand_stack_vars): Make sure
9818 a representative is unpoison if another
9819 variable in the partition is handled by
9820 use-after-scope sanitization.
9822 2018-09-24 Richard Biener <rguenther@suse.de>
9824 PR tree-optimization/63155
9825 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
9826 the worklist when the edge of the respective argument isn't
9829 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
9831 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
9832 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
9833 (MASK_CLASS_P): Update for rename.
9834 (MAYBE_MASK_CLASS_P): Ditto.
9835 (REG_CLASS_NAMES): Update.
9836 (REG_CLASS_CONTENT): Update.
9837 * config/i386/i386.c (regclass_map): Update for MASK_REG
9838 and ALL_MASK_REGS rename.
9839 * config/i386/constraints.md (Yk): Update for rename.
9842 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
9844 * config/i386/i386.h (enum reg_class): Remove
9845 EVEX_SSE_REGS and MOD4_SSE_REGS.
9846 (REG_CLASS_NAMES): Update.
9847 (REG_CLASS_CONTENT): Update.
9848 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
9849 registers as ALL_SSE_REGS.
9850 (ix86_additional_allocno_class_p): Remove.
9851 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
9852 (ix86_register_priority): Lower priority of EVEX SSE registers.
9853 Use IN_RANGE macro where appropriate.
9854 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
9855 AVX-5124VNNIW checks.
9856 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
9857 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
9858 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
9859 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
9860 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
9861 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
9862 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
9863 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
9864 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
9865 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
9866 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
9867 * config/i386/constraints.md (Yh): Remove.
9869 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
9871 * config/i386/i386.c (regclass_map): Declare integer REX registers
9874 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
9876 * doc/service.texi (Service): Switch the fsf.org link to https.
9878 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
9881 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9882 Define to speculation_safe_value_not_needed.
9884 2018-09-21 Florian Weimer <fweimer@redhat.com>
9887 * doc/extend.texi (Common Function Attributes): Mention that
9888 noreturn suppresses tail call optimization.
9890 2018-09-21 Jeff Law <law@redhat.com>
9892 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
9893 vr_values::cleanup_edges_and_switches.
9894 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
9896 (identify_jump_threads): Remove EDGE_IGNORE handling.
9897 (execute_vrp): Move handling of to_remove_edges and
9898 to_update_switch_stmts into vr_values class member functions.
9899 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
9900 (to_update_switch_stmts): Likewise.
9901 * vr-values.c: Include cfghooks.h.
9902 (vr_values::vr_values): Initialize to_remove_edges and
9903 to_update_switch_stmts.
9904 (vr_values::~vr_values): Verify to_remove_edges and
9905 to_update_switch_stmts are empty.
9906 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
9907 (vr_values::cleanup_edges_and_switches): New member function.
9908 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
9909 function. Add new data members.
9911 2018-09-21 David Malcolm <dmalcolm@redhat.com>
9913 PR tree-optimization/87309
9914 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
9915 calls with pflags and alt_flags.
9916 (selftest::test_capture_of_dump_calls): Add test of interaction of
9917 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
9919 2018-09-21 Olivier Hainque <hainque@adacore.com>
9921 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
9923 2018-09-21 Olivier Hainque <hainque@adacore.com>
9925 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
9927 2018-09-21 Olivier Hainque <hainque@adacore.com>
9929 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
9930 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
9931 (VXWORKS_LIBS_RTP): Minor reordering.
9933 2018-09-21 Olivier Hainque <hainque@adacore.com>
9935 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
9936 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
9938 2018-09-21 Olivier Hainque <hainque@adacore.com>
9940 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
9941 (PTRDIFF_TYPE): Likewise.
9943 2018-09-21 Olivier Hainque <hainque@adacore.com>
9945 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
9946 triplet, similar to support for VxWorks7.
9947 * config/vxworks-dummy.h: Provide a default definition
9948 of TARGET_VXWORKS64 to 0.
9950 2018-09-21 Olivier Hainque <hainque@adacore.com>
9952 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
9953 * config/vxworks-dummy.h: here.
9955 2018-09-21 Olivier Hainque <hainque@adacore.com>
9957 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
9959 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
9960 Bo Zhou <zbo.zhou@hisilicon.com>
9962 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
9963 * config/aarch64/aarch64-tune.md: Regenerated.
9964 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
9965 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
9966 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
9968 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
9969 Julian Brown <julian@codesourcery.com>
9971 * builtins.c (get_builtin_sync_mem): Handle address spaces.
9973 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
9975 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
9976 if the call takes a static chain.
9978 2018-09-21 Martin Liska <mliska@suse.cz>
9980 * auto-profile.c (autofdo_source_profile::read): Do not
9982 (read_profile): Do not add working sets.
9983 (read_autofdo_file): Remove sum_all.
9984 (afdo_callsite_hot_enough_for_early_inline): Remove const
9986 * coverage.c (struct counts_entry): Remove gcov_summary.
9987 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
9988 do not support GCOV_TAG_PROGRAM_SUMMARY.
9989 (get_coverage_counts): Remove summary and expected
9991 * coverage.h (get_coverage_counts): Likewise.
9992 * doc/gcov-dump.texi: Remove -w option.
9993 * gcov-dump.c (dump_working_sets): Remove.
9994 (main): Do not support '-w' option.
9995 (print_usage): Likewise.
9996 (tag_summary): Likewise.
9997 * gcov-io.c (gcov_write_summary): Do not dump
9999 (gcov_read_summary): Likewise.
10000 (gcov_histo_index): Remove.
10001 (gcov_histogram_merge): Likewise.
10002 (compute_working_sets): Likewise.
10003 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
10005 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
10006 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
10007 (GCOV_HISTOGRAM_SIZE): Remove.
10008 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
10009 (struct gcov_summary): Simplify rapidly just
10010 to runs and sum_max fields.
10011 (gcov_histo_index): Remove.
10012 (NUM_GCOV_WORKING_SETS): Likewise.
10013 (compute_working_sets): Likewise.
10014 * gcov-tool.c (print_overlap_usage_message): Remove
10015 trailing empty line.
10016 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
10017 (output_lines): Remove program related line.
10018 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
10019 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
10021 (input_profile_summary): Do not read it.
10022 (merge_profile_summaries): And do not merge it.
10023 (input_symtab): Do not call removed function.
10024 * modulo-sched.c (sms_schedule): Do not print sum_max.
10025 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
10026 removed when histogram method was invented.
10027 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
10029 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
10031 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
10032 and dump selected value.
10033 * profile.c (add_working_set): Remove.
10034 (get_working_sets): Likewise.
10035 (find_working_set): Likewise.
10036 (get_exec_counts): Do not work with working sets.
10037 (read_profile_edge_counts): Do not inform as sum_max is removed.
10038 (compute_branch_probabilities): Likewise.
10039 (compute_value_histograms): Remove argument for call of
10040 get_coverage_counts.
10041 * profile.h: Do not make gcov_summary const.
10043 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
10045 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
10047 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
10049 PR tree-optimization/86990
10050 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
10051 Check that the entire merged store group is made of constants only for
10052 overlapping stores.
10054 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
10056 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
10057 (VTABLE_VERIFICATION_SPEC): Likewise.
10058 (SANITIZER_EARLY_SPEC): Likewise.
10059 (SANITIZER_SPEC): Likewise.
10060 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
10061 * doc/invoke.texi (Link Options): Document -r.
10063 2018-09-20 Richard Biener <rguenther@suse.de>
10065 PR middle-end/87054
10066 * gimplify.c (gimplify_expr): Retain alignment of
10067 addressable lvalue in dereference.
10069 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
10072 * configure.ac: Check for .loc is_stmt support.
10073 * configure, config.in: Rebuilt.
10074 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
10077 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
10079 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
10080 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
10083 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
10085 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
10087 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
10090 2018-09-20 Marek Polacek <polacek@redhat.com>
10092 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
10094 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
10096 PR tree-optimization/87288
10097 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
10098 into account when determining PEELING_FOR_NITERS.
10100 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
10102 PR tree-optimization/86877
10103 * tree-vect-loop.c (vect_analyze_loop_2): Call
10104 vect_verify_datarefs_alignment.
10106 2018-09-19 Marek Polacek <polacek@redhat.com>
10108 * doc/invoke.texi: Document -Wclass-conversion.
10110 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
10112 * config/pa/pa.c (pa_adjust_priority): Delete.
10113 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
10115 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
10116 (atomic_storehi): Likewise.
10117 (atomic_storesi): Likewise.
10118 (atomic_loaddi): Restore compare and swap exchange loop code.
10120 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
10122 PR rtl-optimization/86902
10123 * combine.c (try_combine): When changing the CC mode used, don't change
10124 an unrelated mode in other_insn to that new CC mode.
10126 2018-09-19 David Malcolm <dmalcolm@redhat.com>
10128 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
10129 with %T in place of calls to dump_generic_expr.
10130 (prune_runtime_alias_test_list): Likewise.
10131 (create_runtime_alias_checks): Likewise.
10132 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
10133 (vect_analyze_data_ref_dependence): Likewise.
10134 (vect_slp_analyze_data_ref_dependence): Likewise.
10135 (vect_record_base_alignment): Likewise. Use %G in place of call
10136 to dump_gimple_stmt.
10137 (vect_compute_data_ref_alignment): Likewise.
10138 (verify_data_ref_alignment): Likewise.
10139 (vect_find_same_alignment_drs): Likewise.
10140 (vect_analyze_group_access_1): Likewise.
10141 (vect_analyze_data_ref_accesses): Likewise.
10142 (dependence_distance_ge_vf): Likewise.
10143 (dump_lower_bound): Likewise.
10144 (vect_prune_runtime_alias_test_list): Likewise.
10145 (vect_find_stmt_data_reference): Likewise.
10146 (vect_analyze_data_refs): Likewise.
10147 (vect_create_addr_base_for_vector_ref): Likewise.
10148 (vect_create_data_ref_ptr): Likewise.
10149 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
10150 (vect_can_advance_ivs_p): Likewise.
10151 (vect_update_ivs_after_vectorizer): Likewise.
10152 (vect_gen_prolog_loop_niters): Likewise.
10153 (vect_prepare_for_masked_peels): Likewise.
10154 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
10155 (vect_determine_vectorization_factor): Likewise.
10156 (vect_is_simple_iv_evolution): Likewise.
10157 (vect_analyze_scalar_cycles_1): Likewise.
10158 (vect_analyze_loop_operations): Likewise.
10159 (report_vect_op): Likewise.
10160 (vect_is_slp_reduction): Likewise.
10161 (check_reduction_path): Likewise.
10162 (vect_is_simple_reduction): Likewise.
10163 (vect_create_epilog_for_reduction): Likewise.
10164 (vect_finalize_reduction:): Likewise.
10165 (vectorizable_induction): Likewise.
10166 (vect_transform_loop_stmt): Likewise.
10167 (vect_transform_loop): Likewise.
10168 (optimize_mask_stores): Likewise.
10169 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
10170 (vect_split_statement): Likewise.
10171 (vect_recog_over_widening_pattern): Likewise.
10172 (vect_recog_average_pattern): Likewise.
10173 (vect_determine_min_output_precision_1): Likewise.
10174 (vect_determine_precisions_from_range): Likewise.
10175 (vect_determine_precisions_from_users): Likewise.
10176 (vect_mark_pattern_stmts): Likewise.
10177 (vect_pattern_recog_1): Likewise.
10178 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10179 (vect_record_max_nunits): Likewise.
10180 (vect_build_slp_tree_1): Likewise.
10181 (vect_build_slp_tree_2): Likewise.
10182 (vect_print_slp_tree): Likewise.
10183 (vect_analyze_slp_instance): Likewise.
10184 (vect_detect_hybrid_slp_stmts): Likewise.
10185 (vect_detect_hybrid_slp_1): Likewise.
10186 (vect_slp_analyze_operations): Likewise.
10187 (vect_slp_analyze_bb_1): Likewise.
10188 (vect_transform_slp_perm_load): Likewise.
10189 (vect_schedule_slp_instance): Likewise.
10190 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
10191 (vect_mark_stmts_to_be_vectorized): Likewise.
10192 (vect_init_vector_1): Likewise.
10193 (vect_get_vec_def_for_operand): Likewise.
10194 (vect_finish_stmt_generation_1): Likewise.
10195 (vect_check_load_store_mask): Likewise.
10196 (vectorizable_call): Likewise.
10197 (vectorizable_conversion): Likewise.
10198 (vectorizable_operation): Likewise.
10199 (vectorizable_load): Likewise.
10200 (vect_analyze_stmt): Likewise.
10201 (vect_is_simple_use): Likewise.
10202 (vect_get_vector_types_for_stmt): Likewise.
10203 (vect_get_mask_type_for_stmt): Likewise.
10204 * tree-vectorizer.c (increase_alignment): Likewise.
10206 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
10208 * doc/rtl.texi: Adjust vec_select description.
10209 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
10210 non-constant selectors.
10212 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
10214 * config/aarch64/aarch64-protos.h
10215 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
10216 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
10217 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
10218 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
10219 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
10220 (AARCH64_FL_PROFILE): Move index so flags are ordered.
10221 (AARCH64_ISA_RCPC8_4): New flag.
10222 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
10223 to aarch64_offset_9bit_signed_unscaled_p.
10224 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
10226 * config/aarch64/constraints.md (Ust): New constraint.
10227 * config/aarch64/predicates.md.
10228 (aarch64_9bit_offset_memory_operand): New predicate.
10229 (aarch64_rcpc_memory_operand): New predicate.
10231 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
10233 PR rtl-optimization/87361
10234 * rtlanal.c (nonzero_bits1): Revert accidental change.
10236 2018-09-19 Richard Biener <rguenther@suse.de>
10238 PR tree-optimization/87349
10239 PR tree-optimization/87342
10240 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
10242 2018-09-18 Marek Polacek <polacek@redhat.com>
10244 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
10245 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
10247 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
10249 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
10251 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
10253 PR rtl-optimization/86882
10254 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
10256 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
10258 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
10259 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
10261 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
10264 * doc/invoke.texi (Link Options): Fix formatting and grammar.
10266 2018-09-18 Richard Biener <rguenther@suse.de>
10268 PR middle-end/63155
10269 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
10270 (compute_samebase_partition_bases): Likewise.
10271 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
10272 (gimple_can_coalesce_p): Simplify.
10274 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
10276 Handle a library implementation of ffs calling __builtin_ffs.
10277 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
10278 (mmix_init_libfuncs): New function: make __builtin_ffs expand
10281 2018-09-17 David Malcolm <dmalcolm@redhat.com>
10283 * diagnostic-show-locus.c (class layout_range): Add field
10285 (layout_range::layout_range): Add "original_idx" param and use it
10286 to initialize new field.
10287 (make_range): Use 0 for original_idx.
10288 (layout::layout): Pass in index to calls to
10289 maybe_add_location_range.
10290 (layout::maybe_add_location_range): Add param "original_idx" and
10291 pass it on to layout_range.
10292 (layout::print_any_labels): Pass on range->m_original_idx to
10294 (gcc_rich_location::add_location_if_nearby): Use 0 for
10296 * gcc-rich-location.h (text_range_label::get_text): Update for new
10298 (range_label_for_type_mismatch::get_text): Likewise.
10300 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
10302 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
10304 2018-09-17 David Malcolm <dmalcolm@redhat.com>
10306 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
10307 format_string_diagnostic_t.
10308 (fmtwarn_n): Likewise.
10309 * substring-locations.c
10310 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
10311 (format_warning_n_va): Convert to...
10312 (format_string_diagnostic_t::emit_warning_n_va): ...this.
10313 (format_warning_va): Convert to...
10314 (format_string_diagnostic_t::emit_warning_va): ...this.
10315 (format_warning_at_substring): Convert to...
10316 (format_string_diagnostic_t::emit_warning): ...this.
10317 (format_warning_at_substring_n): Convert to...
10318 (format_string_diagnostic_t::emit_warning_n): ...this.
10319 * substring-locations.h (class format_string_diagnostic_t): New
10321 (format_warning_va): Convert to
10322 format_string_diagnostic_t::emit_warning_va.
10323 (format_warning_n_va): Convert to
10324 format_string_diagnostic_t::emit_warning_n_va.
10325 (format_warning_at_substring): Convert to
10326 format_string_diagnostic_t::emit_warning.
10327 (format_warning_at_substring_n): Convert to
10328 format_string_diagnostic_t::emit_warning_n.
10330 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
10331 Bernd Schmidt <bernds_cb1@t-online.de>
10333 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
10336 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
10338 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
10339 operand 0 predicate to nonimmediate operand.
10340 (rint<mode>2_frndint): Remove insn pattern.
10341 (rint<mode>2): Change operand 1 predicate to general_operand.
10342 Extend operand 1 to XFmode and generate rintxf2 insn.
10343 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
10344 Do not use X87MODEF mode macro.
10345 (frndintxf2_<rounding>_i387): Rename from
10346 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
10347 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
10348 to XFmode and generate significandxf3 insn.
10350 2018-09-17 Richard Biener <rguenther@suse.de>
10352 PR tree-optimization/87328
10353 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
10354 visiting unexecutable backedges when not iterating.
10355 (do_rpo_vn): Mark all edges not executable even when not
10358 2018-09-17 Martin Jambor <mjambor@suse.cz>
10361 * doc/invoke.texi (Warning Options): Likewise.
10363 2018-09-17 Richard Biener <rguenther@suse.de>
10365 PR tree-optimization/87301
10366 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
10367 clean EH info from leftover copy assignments.
10369 2018-09-17 Martin Liska <mliska@suse.cz>
10371 PR gcov-profile/85871
10372 * gcov.c (output_intermediate_file): Fix out of bounds
10375 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
10377 * config/arc/arc.c: Object attributes for core4 not reflected
10379 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
10382 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
10384 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
10386 2018-09-17 Martin Liska <mliska@suse.cz>
10388 * doc/gcov.texi: Document new option --use-hotness-colors.
10389 * gcov.c (struct source_info): Declare new field.
10390 (source_info::source_info): Set default for maximum_count.
10391 (print_usage): Add new -q option.
10392 (process_args): Process it.
10393 (accumulate_line_info): Save src->maximum_count.
10394 (output_line_beginning): Make color line number if
10395 flag_use_hotness_colors is set.
10396 (output_line_details): Pass default argument value.
10397 (output_lines): Pass src->maximum_count.
10399 2018-09-17 Martin Liska <mliska@suse.cz>
10401 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
10402 Use processor_names table.
10403 * config/i386/i386.c (ix86_default_align): Use
10404 processor_cost_table for alignment values.
10405 (ix86_option_override_internal): Use processor_names.
10406 (ix86_function_specific_print): Likewise.
10407 * config/i386/i386.h (struct processor_costs):
10408 Add alignment values.
10409 (struct ptt): Remove and replace with const char *.
10410 * config/i386/x86-tune-costs.h (struct processor_costs):
10411 Declare default alignments for all costs.
10413 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
10415 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
10416 symbolics or VR_VARYING ranges for ABS_EXPR.
10417 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
10418 when range will wrap.
10420 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
10422 PR middle-end/86864
10423 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
10424 before and after a JUMP_TABLE_DATA.
10426 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
10428 PR middle-end/87188
10429 * dojump.c (do_compare_and_jump): Canonicalize function pointers
10430 when one operand is a function pointer. Use POINTER_TYPE_P and
10431 FUNC_OR_METHOD_TYPE_P.
10432 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
10433 * fold-const.c (build_range_check): Likewise.
10434 * match.pd (simple_comparison): Likewise.
10436 2018-09-14 David Malcolm <dmalcolm@redhat.com>
10439 * spellcheck.c (get_edit_distance_cutoff): New function.
10440 (selftest::test_edit_distance_unit_test_oneway): Rename to...
10441 (selftest::test_get_edit_distance_one_way): ...this.
10442 (selftest::test_get_edit_distance_unit): Rename to...
10443 (selftest::test_get_edit_distance_both_ways): ...this.
10444 (selftest::test_edit_distances): Move tests to this new function,
10445 and test some more pairs of strings. Update for above renaming.
10446 (selftest::get_old_cutoff): New function.
10447 (selftest::test_get_edit_distance_cutoff): New function.
10448 (selftest::assert_suggested_for): New function.
10449 (ASSERT_SUGGESTED_FOR): New macro.
10450 (selftest::assert_not_suggested_for): New function.
10451 (ASSERT_NOT_SUGGESTED_FOR): New macro.
10452 (selftest::test_suggestions): New function.
10453 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
10454 tests to selftest::test_edit_distances and call it. Add calls to
10455 selftest::test_get_edit_distance_cutoff and
10456 selftest::test_suggestions.
10457 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
10458 (best_match::consider): Replace hard-coded cutoff calculation with
10460 (best_match::get_cutoff): New declaration.
10461 (best_match::get_best_meaningful_candidate): Likewise.
10463 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
10465 * builtins.c (fold_builtin_strlen): Remove TODO comment.
10467 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
10470 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
10472 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
10473 terminated string literal.
10475 2018-09-14 Martin Sebor <msebor@redhat.com>
10477 * builtins.c (unterminated_array): Handle ARRAY_REF.
10478 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
10479 * builtins.h (unterminated_array): Declare extern.
10480 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
10482 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
10485 2018-09-14 Martin Sebor <msebor@redhat.com>
10486 Jeff Law <law@redhat.com>
10488 * builtins.c (unterminated_array): New.
10489 (expand_builtin_strcpy): Adjust.
10490 (expand_builtin_strcpy_args): Detect unterminated arrays.
10491 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
10492 unterminated arrays.
10493 * gimple-fold.h (get_maxval_strlen): Add argument.
10494 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
10496 * gimple-fold.c (get_range_strlen): Add argument.
10497 (get_maxval_strlen): Adjust.
10498 * gimple-fold.h (get_range_strlen): Add argument.
10500 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
10502 * config/i386/movdirintrin.h: Fix copyright year.
10504 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
10506 * reg-stack.c: Include regs.h.
10507 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
10508 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
10509 FIRST_STACK_REG, not DFmode.
10510 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
10511 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
10512 (change stack): Default register mode to the reg_raw_mode of
10513 FIRST_STACK_REG, not DFmode.
10514 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
10515 (*swapxf): Rename from swapxf.
10517 2018-09-14 Carl Love <cel@us.ibm.com>
10519 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
10520 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
10522 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
10525 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
10528 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
10531 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
10533 2018-09-14 Jason Merrill <jason@redhat.com>
10535 Fix --enable-gather-detailed-mem-stats.
10536 * hash-table.c (hash_table_usage): Change from variable to function.
10537 * hash-table.h: Adjust.
10538 * Makefile.in: Add missing dependencies on hash-table.h.
10540 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10542 PR tree-optimization/87259
10544 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
10545 execute_cse_reciprocals_1 has tried transforming.
10547 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
10549 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
10550 VR_VARYING for PLUS/MINUS_EXPR.
10552 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
10554 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
10557 2018-09-14 Richard Biener <rguenther@suse.de>
10559 PR middle-end/63155
10560 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
10561 bits for the merged partition.
10563 2018-09-13 Martin Sebor <msebor@redhat.com>
10564 Bernd Edlinger <bernd.edlinger@hotmail.de>
10566 * builtins.h (c_srlen): Add argument.
10567 * builtins.c (warn_string_no_nul): New function.
10568 (c_strlen): Add argument and use it. Update recursive calls.
10569 Pass DECL argument to string_constant to get info on non
10570 terminated strings. Update *NONSTR as needed.
10571 (fold_builtin_strlen): Add argument to calls to c_strlen.
10572 Warn for unterminated arrays.
10573 (warn_string_no_null): Add prototype.
10574 * expr.c (string_constant): Update arguments. Update recursive
10575 calls appropriately. Detect missing NUL terminator and outermost
10576 declaration its missing in.
10577 Improve checks for arrays with nonzero lower bound or elements
10578 that are not a single byte. Simplify offset computation.
10579 Simplify checks for non-NUL terminated strings.
10580 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
10581 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
10583 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
10585 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
10587 * fold-const.c (c_getstr): Fix function comment. Remove unused third
10588 argument. Fix range checks.
10589 * fold-const.h (c_getstr): Adjust protoype.
10590 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
10591 string is constant but contains no NUL byte.
10593 * expr.c (string_constant): Adjust function comment.
10594 Remove bogus check for zero termination.
10596 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
10598 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
10599 (get_constant_size): Don't make STRING_CSTs larger than they are.
10600 (check_string_literal): New check function for STRING_CSTs.
10601 (output_constant): Use it.
10603 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
10606 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
10608 2018-09-13 Richard Biener <rguenther@suse.de>
10610 PR tree-optimization/87263
10611 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
10612 (struct unwind_state): Add max_rpo field.
10613 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
10614 Compute max_rpo, the max RPO number a block can be backwards reached
10615 from. Re-write non-iterating mode to a RPO ordered worklist approach,
10616 separating it from the iterating mode.
10618 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
10620 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
10621 (rfs_decision): New scheduling decision.
10623 2018-09-13 Richard Biener <rguenther@suse.de>
10626 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
10627 (vn_nary_op_insert_pieces_predicated): Do not write useless
10628 valid_dominated_by_p entry outside of the allocated storage.
10630 2018-09-13 Omar Sandoval <osandov@osandov.com>
10631 Tom de Vries <tdevries@suse.de>
10634 * dwarf2out.c (is_c): New function.
10635 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
10637 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
10640 * config/aarch64/aarch64.md (*aarch64_bfxil):
10642 * config/aarch64/constraints.md (Ulc): Define.
10643 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
10645 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
10648 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
10650 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
10651 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
10652 aarch64_layout_frame call.
10653 (aarch64_expand_epilogue): Likewise.
10654 (aarch64_initial_elimination_offset): Likewise.
10655 (aarch64_get_separate_components): Likewise.
10656 (aarch64_use_return_insn_p): Likewise.
10657 (aarch64_layout_frame): Remove unneeded check.
10659 2018-09-13 Jakub Jelinek <jakub@redhat.com>
10661 * configure.ac: Only append
10662 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
10663 gcc_config_arguments if it was never reconfigured or last reconfigure
10664 was with different arguments.
10665 * configure: Regenerated.
10667 2018-09-13 Jakub Jelinek <jakub@redhat.com>
10668 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10670 PR middle-end/87290
10671 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
10672 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
10674 2018-09-13 Jakub Jelinek <jakub@redhat.com>
10676 PR tree-optimization/87287
10677 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
10678 X % C == 0 to X % (unsigned) C == 0 optimization to ...
10679 * match.pd (X % C == 0): ... here. New optimization.
10681 2018-09-12 Jakub Jelinek <jakub@redhat.com>
10683 PR middle-end/82853
10684 * expr.h (maybe_optimize_mod_cmp): Declare.
10685 * expr.c (mod_inv): New function.
10686 (maybe_optimize_mod_cmp): New function.
10687 (do_store_flag): Use it.
10688 * cfgexpand.c (expand_gimple_cond): Likewise.
10690 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
10691 Julian Brown <julian@codesourcery.com>
10693 PR middle-end/86336
10694 * gimplify.c (gimplify_scan_omp_clauses): Set
10695 target_firstprivatize_array_bases in OpenACC parallel and kernels
10696 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
10697 OpenACC data regions.
10699 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
10701 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
10702 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
10704 2018-09-12 Richard Biener <rguenther@suse.de>
10706 PR tree-optimization/87280
10707 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
10708 edge but unreachable target.
10709 (do_rpo_vn): For conservatively handling a PHI only mark
10710 the backedge executable but not the block reachable.
10712 2018-09-12 Richard Biener <rguenther@suse.de>
10714 PR tree-optimization/87266
10715 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
10718 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
10720 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
10722 ("trunc<BFP:mode><DFP_ALL:mode>2")
10723 ("trunc<DFP_ALL:mode><BFP:mode>2")
10724 ("extend<BFP:mode><DFP_ALL:mode>2")
10725 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
10726 according to the target operand type.
10728 2018-09-12 Jakub Jelinek <jakub@redhat.com>
10729 Andreas Krebbel <krebbel@linux.ibm.com>
10731 PR tree-optimization/86844
10732 * gimple-ssa-store-merging.c
10733 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
10734 there are any overlapping stores in between them, make sure they are
10735 also coalesced or we give up completely.
10737 2018-09-12 Jakub Jelinek <jakub@redhat.com>
10739 PR middle-end/87248
10740 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
10741 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
10743 2018-09-12 Tom de Vries <tdevries@suse.de>
10745 * common.opt (gdescribe-dies): Add option.
10746 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
10747 attribute for artifical and nameless decls.
10748 (dwarf2out_register_external_die): Add description attribute to
10749 external reference die.
10750 (add_desc_attribute): New functions.
10751 (gen_subprogram_die): Add description attribute to
10752 DW_TAG_call_site_parameter.
10753 * tree-pretty-print.c (print_generic_expr_to_str): New function.
10754 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
10755 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
10756 -gno-describe-dies.
10757 (@item -gdescribe-dies): Add.
10759 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
10761 * tree-vrp.c (vrp_shift_undefined_p): Remove.
10762 (extract_range_from_binary_expr_1: Call
10763 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
10764 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
10767 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
10769 * gimple-ssa-warn-alloca.c
10770 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
10771 field for ALLOCA_BOUND_*_LARGE.
10773 2018-09-11 Nathan Sidwell <nathan@acm.org>
10775 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
10777 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
10779 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
10780 for clobbers. Remove obsolete comment.
10782 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
10784 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
10785 mpxchk, mpxld and mpxst types.
10786 (define_attr length_immediate): Remove all processing of mpx types.
10787 (define_attr prefix_0f): Ditto.
10788 (define_attr memory): Ditto.
10790 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
10792 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
10793 (log<mode>2): Change operand 1 predicate to general_operand.
10794 Extend operand 1 to XFmode and generate logxf3 insn.
10795 (log10<mode>2): Change operand 1 predicate to general_operand.
10796 Extend operand 1 to XFmode and generate log10xf3 insn.
10797 (log2<mode>2): Change operand 1 predicate to general_operand.
10798 Extend operand 1 to XFmode and generate log2xf3 insn.
10799 (fyl2xp1_extend<mode>xf3_i387): Remove.
10800 (log1p<mode>2): Change operand 1 predicate to general_operand.
10801 Extend operand 1 to XFmode and generate log1pxf3 insn.
10802 (fxtract_extend<mode>xf3_i387): Remove.
10803 (logb<mode>2): Change operand 1 predicate to general_operand.
10804 Extend operand 1 to XFmode and generate logbxf3 insn.
10805 (ilogb<mode>2): Change operand 1 predicate to general_operand.
10806 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
10807 (significand<mode>2): Change operand 1 predicate to general_operand.
10808 Extend operand 1 to XFmode and generate significandxf3 insn.
10810 2018-09-11 Nathan Sidwell <nathan@acm.org>
10812 * gcc.c (perror_with_name, pfatal_with_name): Delete.
10813 (load_specs): Use fatal_error.
10814 (DELETE_IF_ORDINARY, process_command): Use error.
10815 (execute, run_attempt): Use fatal_error.
10817 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
10819 * diagnostic-core.h (sorry_at): New prototype.
10820 * diagnostic.c (sorry_at): New function.
10822 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
10824 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
10825 by zero as VR_UNDEFINED.
10827 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
10829 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
10830 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
10831 (<sincos>mode2): New expander.
10832 (sincos_extend<mode>xf3_i387): Remove insn pattern.
10833 (sincos -> sin, cos splitters): Remove splitter patterns.
10834 (sincos<mode>3): Change operand 2 predicate to general_operand.
10835 Extend operand 2 to XFmode and generate sincosxf3 insn.
10836 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
10837 Change operand 3 predicate to const1_operand.
10838 (fptan_extend<mode>xf4_i387): Remove insn pattern.
10839 (tanxf2): Update operands in the call to fptanxf4_i387.
10840 (tan<mode>2): Change operand 1 predicate to general_operand.
10841 Extend operand 1 to XFmode and generate tanxf3 insn.
10842 (atan2xf3): Rename from *fpatanxf3_i387.
10843 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
10844 (atan2xf3): Remove expander.
10845 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
10846 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
10847 (atan<mode>2): Change operand 1 predicate to general_operand.
10848 Extend operand 1 to XFmode and generate atanxf3 insn.
10850 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
10852 * config/i386/i386.md (x87/SSE constant load splitter): Use
10853 memory_operand instead of nonimmediate_operand for input operand
10856 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
10858 * config/i386/i386.md (float partial SSE register stall splitter): Move
10859 splitter near its instruction pattern.
10860 (float_extend partial SSE register stall splitter): Ditto.
10861 (float_truncate partial SSE register stall splitter): Ditto.
10863 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
10866 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10867 to speculation_safe_value_not_needed.
10870 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
10871 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
10872 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
10873 leaf_function_p, instead use has_hard_reg_initial_val.
10875 2018-09-09 Nathan Sidwell <nathan@acm.org>
10877 * gcc.h (pfatal_with_name): Don't declare here.
10878 * gcc.c (pfatal_with_name): Make static.
10880 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
10882 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
10885 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
10887 PR rtl-optimization/85458
10888 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
10889 priority hook to reduce the priority of EXPR.
10891 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
10893 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
10894 DImode for x87 on 32bit targets. Conditionally disable x87 modes
10895 with X87_ENABLE_FLOAT. Remove preparation code.
10896 (*float<SWI48:mode><MODEF:mode>2): Rename from
10897 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
10898 math using "enabled" attribute.
10899 (*floatdi<MODEF:mode>2_i387): Rename from
10900 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
10901 enable for 32bit targets only.
10902 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
10904 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
10905 as operand 1 predicate. Rewrite as define_insn_and_split.
10906 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
10908 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
10910 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
10911 to fallthru to FLOAT case.
10913 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
10916 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
10917 around folding of vec_sl to handle out of range shift values.
10919 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
10921 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
10922 Update callers to gen_fix_trunc<mode>_i387_fisttp
10923 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
10924 nonimmediate_operand.
10925 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
10926 and corresponding splitters.
10927 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
10928 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
10929 (fix_truncdi_i387_with_temp): Remove insn pattern
10930 and corresponding splitters.
10931 (fix_trunc<mode>_i387): Change operand 0 predicate to
10932 nonimmediate_operand.
10933 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
10934 and corresponding splitters.
10935 (*fistdi2_1): Remove.
10937 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
10938 (lrintxfdi2): Remove expander. Reimplement as define_insn.
10939 (*fist<mode>2_1): Remove.
10940 (fist<mode>2): Ditto.
10941 (fist<mode>2_with_temp): Remove insn pattern and corresponding
10943 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
10944 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
10945 (fistdi2_<rounding>): Change operand 0 predicate to
10946 nonimmediate_operand.
10947 (fistdi2_<rounding>_with_temp): Remove insn pattern
10948 and corresponding splitters.
10949 (fist<mode>2_<rounding>): Change operand 0 predicate to
10950 nonimmediate_operand.
10951 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
10952 and corresponding splitters.
10954 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
10956 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
10958 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
10961 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
10963 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10964 early gimple folding of vec_splat().
10965 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
10966 * gimple-fold.h: Add an extern define for tree_vec_extract().
10968 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
10970 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
10971 wrappers around TREE_TYPE comparisons.
10973 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
10976 * config/s390/predicates.md: Add nonsym_memory_operand.
10977 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
10978 contains a SYMBOL_REF, load it into an intermediate pseudo.
10979 (s390_emit_compare_and_swap): Legitimize operand.
10980 * config/s390/s390.md: Use the new nonsym_memory_operand
10981 with UNSPECV_CAS patterns.
10983 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
10986 * config/s390/s390-passes.def: New file.
10987 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
10989 (make_pass_s390_early_mach): Add declaration.
10990 * config/s390/s390.c (make_pass_s390_early_mach):
10991 (s390_option_override): Remove dynamic registration.
10992 * config/s390/t-s390: Add s390-passes.def.
10994 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
10996 * config/s390/s390.c (s390_decompose_constant_pool_ref):
10997 Remove UNSPEC_LTREL_BASE check.
10998 (annotate_constant_pool_refs): Likewise.
10999 (find_constant_pool_ref): Likewise.
11000 (find_ltrel_base): Removed.
11001 (replace_ltrel_base): Removed.
11002 (s390_mainpool_finish): Remove replace_ltrel_base call.
11003 (s390_chunkify_start): Remove pending LTREL_BASE logic.
11004 (s390_chunkify_finish): Remove replace_ltrel_base call.
11005 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
11007 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
11010 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
11011 to speculation_safe_value_not_needed.
11013 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
11014 Bernd Schmidt <bernds_cb1@t-online.de>
11016 * config/nvptx/nvptx-opts.h: New file.
11017 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
11018 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
11019 (ASM_SPEC): Define.
11020 (TARGET_SM35): New macro.
11021 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
11023 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
11025 (misa=): New option.
11026 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
11028 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
11030 * config/i386/i386.md (truncdfsf2): Remove expander.
11031 (truncdfsf2_with_temp): Ditto.
11032 (truncxf<mode>2): Ditto.
11033 (*truncdfsf_fast_mixed): Remove insn pattern.
11034 (*truncdfsf_fast_i387): Ditto.
11035 (*truncdfsf_mixed): Ditto.
11036 (*truncdfsf_i387): Ditto.
11037 (*truncdfsf2_i387_1): Ditto.
11038 (*truncxfsf2_mixed): Ditto.
11039 (*truncxfdf2_mixed): Ditto.
11040 (*truncxf<mode>2_i387_noop): Ditto. Update callers
11041 to call gen_truncxf<mode>2 instead.
11042 (*truncxf<mode>2_i387): Remove.
11043 (reg->reg splitters): Remove splitter pattern.
11044 (reg->mem splitters): Ditto.
11046 (truncdfsf2): New insn pattern.
11047 (truncxf<mode>2): Ditto.
11049 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11051 * tree-ssa-math-opts.c (is_mult_by): New function.
11052 (is_square_of): Use the above.
11053 (optimize_recip_sqrt): New function.
11054 (pass_cse_reciprocals::execute): Use the above.
11056 2018-09-05 Richard Biener <rguenther@suse.de>
11059 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
11060 to zero-init the emplaced vec.
11062 2018-09-05 Martin Liska <mliska@suse.cz>
11064 PR tree-optimization/87205
11065 * tree-switch-conversion.c (pass_lower_switch::execute):
11066 Group cases for switch statements.
11068 2018-09-05 Richard Biener <rguenther@suse.de>
11070 PR tree-optimization/87217
11071 * tree-ssa-sccvn.c (vuse_valueize): New.
11072 (vn_reference_lookup_pieces): Use it.
11073 (vn_reference_lookup): Likewise.
11075 2018-09-05 Nathan Sidwell <nathan@acm.org>
11078 * stor-layout.c (place_field): Scan forwards to check last
11079 bitfield when ms_bitfield_placement is in effect.
11081 2018-09-05 Richard Biener <rguenther@suse.de>
11084 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
11087 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
11088 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
11090 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
11091 * config.gcc (extra_objs): Build it.
11092 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
11094 * config/aarch64/aarch64-passes.def
11095 (pass_tag_collision_avoidance): New pass.
11096 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
11097 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
11098 (aarch64_classify_address): Remove static qualifier.
11099 (aarch64_address_info, aarch64_address_type): Move to...
11100 * config/aarch64/aarch64-protos.h: ... here.
11101 (make_pass_tag_collision_avoidance): New function.
11102 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
11105 2018-09-05 Martin Liska <mliska@suse.cz>
11107 * doc/gcov.texi: Update documentation of humar
11109 * gcov.c (format_count): Print one decimal place, it provides
11110 more fine number of situations like '1G' vs. '1.4G'.
11112 2018-09-05 Martin Liska <mliska@suse.cz>
11115 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
11116 * optc-gen.awk: Allow 'Var' for Deprecated options in order
11117 to generate a MASK value.
11119 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
11122 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
11123 if frame pointer isn't used.
11124 (compute_frame_pointer_to_fb_displacement): Likewise.
11126 2018-09-04 Jakub Jelinek <jakub@redhat.com>
11129 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
11130 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
11131 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
11132 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
11133 and OPTION_MASK_ISA_XSAVEC_UNSET.
11135 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
11137 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
11138 XOR operations in NAND case.
11140 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
11142 * wide-int-range.cc (wide_int_range_convert): New.
11143 * wide-int-range.h (wide_int_range_convert): New.
11144 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
11145 code into wide_int_range_convert.
11146 (extract_range_into_wide_ints): Do not munge anti range constants
11147 into the entire domain. Just return the range back.
11149 2018-09-04 Martin Liska <mliska@suse.cz>
11151 * genmatch.c (output_line_directive): Add new argument
11153 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
11155 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
11157 * doc/invoke.texi (Option Summary): Add whitespace.
11159 * doc/invoke.texi (Option Summary): Add -Waligned-new.
11161 2018-09-04 Richard Biener <rguenther@suse.de>
11163 PR tree-optimization/87211
11164 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
11165 backedge value we're supposed to treat as VARYING also number
11166 the PHI to VARYING in case it got a different value-number already.
11168 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
11170 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
11171 (extract_range_from_binary_expr_1): Do not call
11172 vrp_can_optimize_bit_op.
11173 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
11175 (wide_int_range_get_mask_and_bounds): New.
11176 (wide_int_range_optimize_bit_op): New.
11177 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
11178 (wide_int_range_bit_and): Same.
11179 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
11180 (wide_int_range_optimize_bit_op): New.
11181 (wide_int_range_get_mask_and_bounds): New.
11183 2018-09-04 Richard Biener <rguenther@suse.de>
11185 PR tree-optimization/87176
11186 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
11187 variable. When value-numbering a virtual PHI node make sure
11188 to not value-number to the backedge value.
11190 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
11192 * doc/extend.texi (Long Long, Hex Floats): Document support for
11193 long long and hex floats in more recent versions of ISO C++.
11195 2018-09-03 Richard Biener <rguenther@suse.de>
11197 PR tree-optimization/87177
11198 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
11201 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11203 * bb-reorder.c (edge_order): Convert to C-qsort-style
11204 tri-state comparator.
11205 (reorder_basic_blocks_simple): Change std::stable_sort to
11208 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11210 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
11211 tri-state comparator.
11212 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
11214 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11216 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
11217 (mergesort): ... here as maximum count for using netsort.
11218 (gcc_qsort): Set nlim to 3 if stable sort is requested.
11219 (gcc_stablesort): New.
11220 * system.h (gcc_stablesort): Declare.
11222 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11224 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
11225 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
11226 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
11228 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
11230 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
11231 lxsdx and stxsdx alternatives.
11232 (*mov<mode>_hardfloat64): Ditto.
11233 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
11235 2018-09-03 Richard Biener <rguenther@suse.de>
11237 PR tree-optimization/87200
11238 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
11241 2018-09-03 Martin Liska <mliska@suse.cz>
11243 PR tree-optimization/87201
11244 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
11245 Fix parenthesis in an expression.
11247 2018-09-03 Richard Biener <rguenther@suse.de>
11249 PR tree-optimization/87197
11250 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
11251 visited. CSE the VN_INFO hashtable lookup.
11253 PR tree-optimization/87169
11254 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
11255 iterating make sure there's no extra backedges from irreducible
11256 regions feeding the header. Mark the destination block
11259 2018-09-03 Martin Liska <mliska@suse.cz>
11262 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
11263 * common/common-targhooks.c (default_get_valid_option_values):
11265 * common/common-targhooks.h (default_get_valid_option_values):
11267 * common/config/i386/i386-common.c: Move processor_target_table
11269 (ix86_get_valid_option_values): New function.
11270 (TARGET_GET_VALID_OPTION_VALUES): New macro.
11271 * config/i386/i386.c (struct ptt): Move to i386-common.c.
11272 (PTA_*): Move all defined masks into i386-common.c.
11273 (ix86_function_specific_restore): Use new processor_cost_table.
11274 * config/i386/i386.h (struct ptt): Moved from i386.c.
11275 (struct pta): Likewise.
11276 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
11277 * doc/tm.texi.in: Likewise.
11278 * opt-suggestions.c (option_proposer::suggest_option):
11279 Pass prefix to build_option_suggestions.
11280 (option_proposer::get_completions): Likewise.
11281 (option_proposer::build_option_suggestions): Use the new target
11283 * opts.c (struct option_help_tuple): New struct.
11284 (print_filtered_help): Use the new target hook.
11286 2018-09-03 Martin Liska <mliska@suse.cz>
11288 PR middle-end/59521
11289 * predict.c (set_even_probabilities): Add likely_edges
11290 argument and handle cases where we have precisely one
11292 (combine_predictions_for_bb): Catch also likely_edges.
11293 (tree_predict_by_opcode): Handle gswitch statements.
11294 * tree-cfg.h (find_case_label_for_value): New declaration.
11295 (find_taken_edge_switch_expr): Likewise.
11296 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
11297 Find pivot in decision tree based on probabily, not by number of
11300 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
11302 * doc/standards.texi (Standards): Update Objective-C reference.
11304 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
11306 * doc/install.texi (Prerequisites): Update link for MPC.
11308 2018-09-01 Michael Matz <matz@suse.de>
11310 PR tree-optimization/87074
11311 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
11312 PHIs for outer-loop uses.
11314 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
11316 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
11317 * doc/invoke.texi (C Dialect Options): Ditto.
11319 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
11321 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
11323 2018-08-31 Richard Biener <rguenther@suse.de>
11325 PR tree-optimization/87168
11326 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
11327 (rpo_elim::eliminate_avail): When OP was not visited it must
11330 2018-08-31 David Malcolm <dmalcolm@redhat.com>
11332 * tree-vrp.c (copy_value_range): Convert param "from" from
11333 "value_range *" to "const value_range *".
11334 (range_is_null): Likewise for param "vr".
11335 (range_int_cst_p): Likewise.
11336 (range_int_cst_singleton_p): Likewise.
11337 (symbolic_range_p): Likewise.
11338 (value_ranges_intersect_p): Likewise for both params.
11339 (value_range_nonnegative_p): Likewise for param "vr".
11340 (value_range_constant_singleton): Likewise.
11341 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
11342 (extract_range_into_wide_ints): Likewise for param "vr".
11343 (extract_range_from_multiplicative_op): Likewise for params "vr0"
11345 (vrp_can_optimize_bit_op): Likewise.
11346 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
11348 (extract_range_from_unary_expr): Likewise.
11349 (debug_value_range): Likewise for param "vr".
11350 (value_range::dump): Add "const" qualifier.
11351 (vrp_prop::check_array_ref): Convert local "vr" from
11352 "value_range *" to "const value_range *".
11353 (vrp_prop::check_mem_ref): Likewise.
11354 (vrp_prop::visit_stmt): Likewise for local "old_vr".
11355 (vrp_intersect_ranges_1): Likewise for param "vr_1".
11356 (vrp_intersect_ranges): Likewise.
11357 (simplify_stmt_for_jump_threading): Likewise for local "vr".
11358 (vrp_prop::vrp_finalize): Likewise.
11359 * tree-vrp.h (value_range::dump): Add "const" qualifier.
11360 (vrp_intersect_ranges): Add "const" qualifier to params as above.
11361 (extract_range_from_unary_expr): Likewise.
11362 (value_range_constant_singleton): Likewise.
11363 (symbolic_range_p): Likewise.
11364 (copy_value_range): Likewise.
11365 (extract_range_from_binary_expr_1): Likewise.
11366 (range_int_cst_p): Likewise.
11367 (vrp_set_zero_nonzero_bits): Likewise.
11368 (range_int_cst_singleton_p): Likewise.
11370 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
11372 * config/aarch64/arm_neon.h (vabsd_s64): New.
11373 (vnegd_s64): Likewise.
11375 2018-08-31 Martin Jambor <mjambor@suse.cz>
11377 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
11379 2018-08-31 Martin Liska <mliska@suse.cz>
11381 * ipa-icf.c (sem_item::add_type): Use
11382 sem_item::m_type_hash_cache.
11383 * ipa-icf.h: Move the cache from sem_item_optimizer
11386 2018-08-31 Nathan Sidwell <nathan@acm.org>
11388 * doc/extend.texi (Backwards Compatibility): Remove implicit
11389 extern C leeway of () being (...).
11391 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11393 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
11395 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
11399 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
11401 2018-08-31 Jakub Jelinek <jakub@redhat.com>
11403 PR middle-end/87138
11404 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
11405 gen_int_mode. Formatting fixes.
11407 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
11409 * target.def (custom_function_descriptors): Improve documentation.
11410 * doc/tm.texi.in (Trampolines): Expand discussion of function
11411 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
11412 beginning of the section.
11413 * doc/tm.texi: Regenerated.
11415 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
11417 * cfg.h (class auto_edge_flag): Spell out the template-id of the
11418 base class in the initializer list. This is a workaround for
11419 building with older GCC.
11420 (class auto_bb_flag): Likewise.
11422 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
11424 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
11425 (altivec_vcmpequ<VI_char>_p): Remove star.
11426 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
11428 (expand_strncmp_vec_sequence): New function.
11429 (emit_final_str_compare_vec): New function.
11430 (expand_strn_compare): Add support for vector strncmp.
11431 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
11432 length specification to bytes.
11433 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
11434 (vcmpnezb_p): New pattern.
11435 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
11436 for option -mstring-compare-inline-limit.
11438 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
11440 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
11441 (PTA_SKYLAKE): Add PTA_AES.
11442 (PTA_GOLDMONT): Likewise.
11444 2018-08-29 Jan Hubicka <jh@suse.cz>
11447 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
11448 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
11450 2018-08-29 Jan Hubicka <jh@suse.cz>
11452 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
11454 (hash_tree): Do not visit TYPE_STUB_DECL.
11455 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
11456 stream TYPE_STUB_DECL.
11457 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
11458 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
11459 after free_lang_data.
11460 (type_in_anonymous_namespace_p): Likewise.
11462 2018-08-29 Jan Hubicka <jh@suse.cz>
11464 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
11465 comment that it has to be even number.
11466 (class sreal): Change m_sig type to int32_t.
11467 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
11468 int64_t for temporary calculations.
11469 (sreal_verify_basics): Drop one bit from minimum and maximum.
11471 2018-08-30 Richard Biener <rguenther@suse.de>
11473 PR tree-optimization/87147
11474 * tree-ssa-sccvn.c (SSA_VISITED): New function.
11475 (visit_phi): When the degenerate result is from the backedge and
11476 we didn't visit its definition yet drop to VARYING.
11477 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
11479 2018-08-29 Jan Hubicka <jh@suse.cz>
11481 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
11483 (hash_tree): Likewise.
11485 2018-08-29 Jan Hubicka <jh@suse.cz>
11487 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
11488 and TYPE_NEXT_REF_TO.
11490 2018-08-29 Jan Hubicka <jh@suse.cz>
11492 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
11493 comment that it has to be even number.
11494 (class sreal): Change m_sig type to int32_t.
11495 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
11496 int64_t for temporary calculations.
11497 (sreal_verify_basics): Drop one bit from minimum and maximum.
11499 2018-08-30 Tamar Christina <tamar.christina@arm.com>
11501 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
11503 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
11505 PR middle-end/86995
11506 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
11507 if to_add is negative.
11509 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
11511 PR middle-end/87053
11512 * builtins.c (c_strlen): Improve range checks.
11514 2018-08-29 Martin Sebor <msebor@redhat.com>
11515 Jeff Law <law@redhat.com>
11517 PR tree-optimization/86714
11518 PR tree-optimization/86711
11519 * builtins.c (c_strlen): Add arguments to call to string_constant.
11520 * expr.c (string_constant): Add argument. Detect missing nul
11521 terminator and outermost declaration it's missing in.
11522 * expr.h (string_constant): Add argument.
11523 * fold-const.c (read_from_constant_string): Add arguments to call to
11525 (c_getstr): Likewise.
11526 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
11527 to string_constant.
11528 * tree-ssa-strlen.c (get_stridx): Likewise.
11530 2018-08-29 Jan Hubicka <jh@suse.cz>
11532 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
11533 Do not stream DECL_VINDEX.
11534 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
11535 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
11536 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
11538 2018-08-29 Richard Biener <rguenther@suse.de>
11540 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
11541 virtual operands that are not default defs to honor region
11543 (rpo_vn_valueize): Remove ineffective code here.
11545 2018-08-29 Richard Biener <rguenther@suse.de>
11547 PR tree-optimization/87132
11548 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
11549 when skipping defs reachable over backedges.
11551 2018-08-29 Richard Biener <rguenther@suse.de>
11553 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
11554 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
11555 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
11556 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
11557 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
11558 (vn_reference_lookup_3): Remove use of const_parms.
11559 (free_rpo_vn): Do not free const_parms.
11560 (do_rpo_vn): Do not call init_const_parms.
11561 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
11562 SSA_NAME_POINTS_TO_READONLY_MEMORY.
11563 (call_may_clobber_ref_p_1): Likewise.
11565 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
11568 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
11570 (-ftree-scev-cprop): Document.
11572 2018-08-29 Jan Hubicka <jh@suse.cz>
11574 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
11576 (sreal constructor): Update.
11577 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
11578 sreal:operator/): Update.
11580 2018-08-29 Martin Liska <mliska@suse.cz>
11582 * tree-switch-conversion.c (switch_conversion::expand):
11583 Strenghten assumption about gswitch statements.
11585 2018-08-29 Richard Biener <rguenther@suse.de>
11587 PR tree-optimization/87117
11588 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
11589 re-value-number released SSA VDEFs.
11591 2018-08-29 Richard Biener <rguenther@suse.de>
11593 PR tree-optimization/87126
11594 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
11596 2018-08-28 Jim Wilson <jimw@sifive.com>
11598 * config/riscv/pic.md: Rewrite.
11599 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
11601 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
11602 (SOFTF, default_load, softload, softstore): New.
11604 2018-08-28 Jeff Law <law@redhat.com>
11606 * fold-const.c (fold_binary_loc): Remove recently added assert.
11608 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
11610 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
11611 to OP parmeter of generated function.
11613 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
11615 PR tree-optimization/87009
11616 * match.pd: Add boolean optimizations.
11618 2018-08-28 Martin Sebor <msebor@redhat.com>
11620 PR middle-end/86631
11621 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
11622 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
11623 (pass_walloca::gate): Use it.
11624 (alloca_call_type): Same.
11625 (pass_walloca::execute): Same.
11626 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
11628 2018-08-28 David Malcolm <dmalcolm@redhat.com>
11630 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
11631 GCC_VERSION for usage of "__gcc_dump_printf__" format from
11632 >= 3005 to >= 9000.
11634 2018-08-28 Richard Biener <rguenther@suse.de>
11636 PR tree-optimization/87124
11637 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
11638 constants before looking up avail.
11640 2018-08-28 Jakub Jelinek <jakub@redhat.com>
11642 PR middle-end/87099
11643 * calls.c (maybe_warn_nonstring_arg): Punt early if
11644 warn_stringop_overflow is zero. Don't call get_range_strlen
11645 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
11646 Swap comparison operands to have constants on rhs. Only use
11647 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
11648 increment lenrng[0].
11650 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
11652 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
11653 use of tree_to_shwi. Remove duplicated test for the size being
11654 a whole number of bytes.
11656 2018-08-28 Richard Biener <rguenther@suse.de>
11658 PR tree-optimization/87117
11659 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
11660 Handle removed stmt without LHS (GIMPLE_NOP).
11662 2018-08-28 Richard Biener <rguenther@suse.de>
11664 PR tree-optimization/87117
11665 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
11666 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
11668 2018-08-28 Richard Biener <rguenther@suse.de>
11670 PR tree-optimization/87117
11671 * tree-ssa-pre.c (compute_avail): Do not make expressions
11672 with predicated values available.
11673 (get_expr_value_id): Assert we do not run into predicated value
11676 2018-08-28 Richard Biener <rguenther@suse.de>
11678 PR tree-optimization/87117
11679 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
11680 get virtual operands.
11681 (get_expr_operands): Handle STRING_CST like other decls.
11683 2018-08-28 Martin Liska <mliska@suse.cz>
11685 * tree.h: Update documentation of fndecl_built_in_p
11689 2018-08-27 Jeff Law <law@redhat.com>
11690 PR tree-optimization/87110
11691 * tree-ssa-dse.c (compute_trims): Handle non-constant
11694 2018-08-27 Martin Sebor <msebor@redhat.com>
11696 PR tree-optimization/86914
11697 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
11699 2018-08-27 Martin Sebor <msebor@redhat.com>
11701 PR tree-optimization/87112
11702 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
11703 the type of the bound argument.
11705 2018-08-27 Jeff Law <law@redhat.com>
11707 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
11708 type does not have a TYPE_SIZE_UNIT.
11710 2018-08-27 Steve Ellcey <sellcey@cavium.com>
11712 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
11713 with include of backend.h.
11715 2018-08-27 Richard Biener <rguenther@suse.de>
11717 PR tree-optimization/86927
11718 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
11719 use const cond reduction code.
11721 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
11723 PR tree-optimization/85758
11724 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
11726 2018-08-27 David Malcolm <dmalcolm@redhat.com>
11729 * diagnostic-show-locus.c (class layout_range): Update for
11730 conversion of show_caret_p to a tri-state.
11731 (layout_range::layout_range): Likewise.
11732 (make_range): Likewise.
11733 (layout::maybe_add_location_range): Likewise.
11734 (layout::should_print_annotation_line_p): Don't show annotation
11735 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
11736 (layout::get_state_at_point): Update for conversion of
11737 show_caret_p to a tri-state. Bail out early for
11738 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
11739 underlining or source colorization.
11740 (gcc_rich_location::add_location_if_nearby): Update for conversion
11741 of show_caret_p to a tri-state.
11742 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
11743 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
11745 (selftest::test_one_liner_labels): Likewise.
11746 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
11747 conversion of show_caret_p to a tri-state.
11748 * pretty-print.c (text_info::set_location): Likewise.
11749 * pretty-print.h (text_info::set_location): Likewise.
11750 * substring-locations.c (format_warning_n_va): Likewise.
11751 * tree-diagnostic.c (default_tree_printer): Likewise.
11752 * tree-pretty-print.c (newline_and_indent): Likewise.
11754 2018-08-27 David Malcolm <dmalcolm@redhat.com>
11757 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
11758 line above for line-insertion fix-it hints.
11759 (selftest::test_fixit_insert_containing_newline): Update the
11760 expected results, and add a test with line-numbering enabled.
11762 2018-08-27 Martin Liska <mliska@suse.cz>
11765 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
11766 params with DECL_HAS_VALUE_EXPR_P.
11768 2018-08-27 Martin Liska <mliska@suse.cz>
11770 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
11771 selected expansion strategy.
11773 2018-08-27 Martin Liska <mliska@suse.cz>
11775 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
11776 * builtins.c (is_builtin_fn): Likewise.
11777 * attribs.c (diag_attr_exclusions): Use new function
11778 fndecl_built_in_p and remove check for FUNCTION_DECL if
11780 (builtin_mathfn_code): Likewise.
11781 (fold_builtin_expect): Likewise.
11782 (fold_call_expr): Likewise.
11783 (fold_builtin_call_array): Likewise.
11784 (fold_call_stmt): Likewise.
11785 (set_builtin_user_assembler_name): Likewise.
11786 (is_simple_builtin): Likewise.
11787 * calls.c (gimple_alloca_call_p): Likewise.
11788 (maybe_warn_nonstring_arg): Likewise.
11789 * cfgexpand.c (expand_call_stmt): Likewise.
11790 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
11791 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
11792 (cgraph_node::verify_node): Likewise.
11793 * cgraphclones.c (build_function_decl_skip_args): Likewise.
11794 (cgraph_node::create_clone): Likewise.
11795 * config/arm/arm.c (arm_insert_attributes): Likewise.
11796 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
11797 * dse.c (scan_insn): Likewise.
11798 * expr.c (expand_expr_real_1): Likewise.
11799 * fold-const.c (operand_equal_p): Likewise.
11800 (fold_binary_loc): Likewise.
11801 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
11802 * gimple-low.c (lower_stmt): Likewise.
11803 * gimple-pretty-print.c (dump_gimple_call): Likewise.
11804 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11806 * gimple.c (gimple_build_call_from_tree): Likewise.
11807 (gimple_call_builtin_p): Likewise.
11808 (gimple_call_combined_fn): Likewise.
11809 * gimplify.c (gimplify_call_expr): Likewise.
11810 (gimple_boolify): Likewise.
11811 (gimplify_modify_expr): Likewise.
11812 (gimplify_addr_expr): Likewise.
11813 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
11814 * ipa-cp.c (determine_versionability): Likewise.
11815 * ipa-fnsummary.c (compute_fn_summary): Likewise.
11816 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
11817 * ipa-split.c (visit_bb): Likewise.
11818 (split_function): Likewise.
11819 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
11820 * lto-cgraph.c (input_node): Likewise.
11821 * lto-streamer-out.c (write_symbol): Likewise.
11822 * omp-low.c (setjmp_or_longjmp_p): Likewise.
11823 (lower_omp_1): Likewise.
11824 * predict.c (strip_predict_hints): Likewise.
11825 * print-tree.c (print_node): Likewise.
11826 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
11827 * trans-mem.c (is_tm_irrevocable): Likewise.
11828 (is_tm_load): Likewise.
11829 (is_tm_simple_load): Likewise.
11830 (is_tm_store): Likewise.
11831 (is_tm_simple_store): Likewise.
11832 (is_tm_abort): Likewise.
11833 (tm_region_init_1): Likewise.
11834 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
11835 * tree-cfg.c (verify_gimple_call): Likewise.
11836 (move_stmt_r): Likewise.
11837 (stmt_can_terminate_bb_p): Likewise.
11838 * tree-eh.c (lower_eh_constructs_2): Likewise.
11839 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
11840 * tree-inline.c (remap_gimple_stmt): Likewise.
11841 (copy_bb): Likewise.
11842 (estimate_num_insns): Likewise.
11843 (fold_marked_statements): Likewise.
11844 * tree-sra.c (scan_function): Likewise.
11845 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
11846 (optimize_stack_restore): Likewise.
11847 (pass_fold_builtins::execute): Likewise.
11848 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11849 (mark_all_reaching_defs_necessary_1): Likewise.
11850 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
11851 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
11852 (pass_forwprop::execute): Likewise.
11853 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11854 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
11855 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
11856 * tree-ssa-strlen.c (get_string_length): Likewise.
11857 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
11858 (find_func_aliases_for_call): Likewise.
11859 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
11860 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
11861 * tree-tailcall.c (find_tail_calls): Likewise.
11862 * tree.c (need_assembler_name_p): Likewise.
11863 (free_lang_data_in_decl): Likewise.
11864 (get_call_combined_fn): Likewise.
11865 * ubsan.c (is_ubsan_builtin_p): Likewise.
11866 * varasm.c (incorporeal_function_p): Likewise.
11867 * tree.h (DECL_BUILT_IN): Remove and replace with
11869 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
11870 (fndecl_built_in_p): New.
11872 2018-08-27 Martin Liska <mliska@suse.cz>
11874 PR tree-optimization/86847
11875 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
11876 Dump also subtree probability.
11877 (switch_decision_tree::do_jump_if_equal): New function.
11878 (switch_decision_tree::emit_case_nodes): Handle special
11879 situations in balanced tree that can be emitted much simpler.
11880 Fix calculation of probabilities that happen in tree expansion.
11881 * tree-switch-conversion.h (struct cluster): Add
11883 (struct simple_cluster): Likewise.
11884 (struct case_tree_node): Add new function has_child.
11885 (do_jump_if_equal): New.
11887 2018-08-27 Martin Liska <mliska@suse.cz>
11889 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
11890 Add new argument to bit_test_cluster constructor.
11891 (bit_test_cluster::emit): Set bits really number of values
11893 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
11894 probability argument.
11895 * tree-switch-conversion.h (struct bit_test_cluster):
11896 Add m_handles_entire_switch.
11898 2018-08-27 Martin Liska <mliska@suse.cz>
11900 PR tree-optimization/86702
11901 * tree-switch-conversion.c (jump_table_cluster::emit):
11902 Make probabilities even for values in jump table
11903 according to number of cases handled.
11904 (switch_decision_tree::compute_cases_per_edge): Pass
11905 argument to reset_out_edges_aux function.
11906 (switch_decision_tree::analyze_switch_statement): Likewise.
11907 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
11910 2018-08-27 Martin Liska <mliska@suse.cz>
11912 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
11913 cfun argument explicitly.
11914 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
11915 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
11916 function gimple_switch_default_bb.
11917 (convert_switch_statements):
11919 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
11920 * stmt.c (label_to_block_fn): Use label_to_block and pass
11921 cfun argument explicitly and use gimple_switch_label_bb.
11922 (expand_case): Likewise.
11923 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
11924 cfun argument explicitly. Likewise.
11925 (make_edges_bb): Likewise.
11926 (make_cond_expr_edges): Likewise.
11927 (get_cases_for_edge): Likewise.
11928 (make_gimple_switch_edges): Likewise.
11929 (label_to_block_fn): Likewise.
11930 (label_to_block): Likewise.
11931 (make_goto_expr_edges): Likewise.
11932 (make_gimple_asm_edges): Likewise.
11933 (main_block_label): Likewise.
11934 (group_case_labels_stmt): Likewise.
11935 (find_taken_edge_computed_goto): Likewise.
11936 (find_taken_edge_switch_expr): Likewise.
11937 (gimple_verify_flow_info): Likewise.
11938 (gimple_redirect_edge_and_branch): Likewise.
11939 (gimple_switch_label_bb): New function.
11940 (gimple_switch_default_bb): Likewise.
11941 (gimple_switch_edge): Likewise.
11942 (gimple_switch_default_edge): Likewise.
11943 * tree-cfg.h (label_to_block_fn): Remove and replace ...
11944 (label_to_block): ... with this.
11945 (gimple_switch_label_bb): New.
11946 (gimple_switch_default_bb): Likewise.
11947 (gimple_switch_edge): Likewise.
11948 (gimple_switch_default_edge): Likewise.
11949 * tree-cfgcleanup.c (convert_single_case_switch): Use
11950 new gimple functions and pass new argument to label_to_block.
11951 (cleanup_control_flow_bb):
11952 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
11953 cfun argument explicitly.
11954 (make_eh_edges): Likewise.
11955 (redirect_eh_dispatch_edge): Likewise.
11956 (lower_resx): Likewise.
11957 (lower_eh_dispatch): Likewise.
11958 (maybe_remove_unreachable_handlers): Likewise.
11959 (unsplit_eh): Likewise.
11960 (cleanup_empty_eh): Likewise.
11961 (verify_eh_edges): Likewise.
11962 (verify_eh_dispatch_edge): Likewise.
11963 * tree-ssa-dom.c (record_edge_info): Likewise.
11964 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
11965 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
11966 (thread_through_normal_block): Likewise.
11967 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
11968 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
11969 * tree-switch-conversion.c (switch_conversion::collect): Use new
11971 (switch_conversion::check_final_bb): Likewise.
11972 (switch_conversion::gather_default_values): Pass new argument
11974 (switch_conversion::build_constructors): Likewise.
11975 (switch_decision_tree::compute_cases_per_edge): Use new
11976 gimple_switch_edge function.
11977 (switch_decision_tree::analyze_switch_statement): Pass new argument
11979 (switch_decision_tree::try_switch_expansion): Use
11980 gimple_switch_default_edge.
11981 * tree-vrp.c (find_switch_asserts): Pass new argument
11983 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
11984 (vr_values::simplify_switch_using_ranges): Likewise.
11986 2018-08-27 Richard Biener <rguenther@suse.de>
11988 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
11989 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
11991 * tree-ssa-sccvn.h (struct vn_pval): New structure.
11992 (struct vn_nary_op_s): Add unwind_to member. Add
11993 predicated_values flag and put result into a union together
11994 with a linked list of vn_pval.
11995 (struct vn_ssa_aux): Add name member to make maintaining
11996 a map of SSA name to vn_ssa_aux possible. Remove no longer
11997 needed info, dfsnum, low, visited, on_sccstack, use_processed
11998 and range_info_anti_range_p members.
11999 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
12000 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
12002 (vn_valueize): New global.
12003 (vn_context_bb): Likewise.
12004 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
12005 VN_INFO_PTR_INFO): Remove.
12006 * tree-ssa-sccvn.c: ... (rewrite)
12007 (pass_fre::execute): For -O2+ initialize loops and run
12008 RPO VN in optimistic mode (iterating). For -O1 and -Og
12009 run RPO VN in non-optimistic mode.
12010 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
12011 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
12012 * doc/invoke.texi (sccvn-max-scc-size): Remove.
12013 (rpo-vn-max-loop-depth): Document.
12014 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
12015 when valuezing the VUSE signals we walked out of the region.
12016 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
12017 (phi_translate): Set VN context block to use for availability
12019 (compute_avail): Likewise.
12020 (pre_valueize): New function.
12021 (pass_pre::execute): Adjust to the RPO VN API.
12023 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
12024 (propagate_constants_for_unrolling): Remove.
12025 (tree_unroll_loops_completely): Perform value-numbering
12026 on the unrolled bodies loop parent.
12028 2018-08-27 Richard Biener <rguenther@suse.de>
12030 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
12031 for partial antic compute.
12033 2018-08-27 Jakub Jelinek <jakub@redhat.com>
12035 PR rtl-optimization/87065
12036 * combine.c (simplify_if_then_else): Formatting fix.
12037 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
12039 (known_cond): Don't return const_true_rtx for vector modes. Use
12040 CONST0_RTX instead of const0_rtx. Formatting fixes.
12042 2018-08-27 Martin Liska <mliska@suse.cz>
12044 PR gcov-profile/87069
12045 * gcov.c (process_file): Record files already processed
12046 and warn about a file being processed multiple times.
12048 2018-08-27 Martin Liska <mliska@suse.cz>
12051 * config/aarch64/aarch64.c (aarch64_override_options_internal):
12052 Set default values for x_aarch64_*_string strings.
12053 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
12054 prefix. For -mabi do not print '=ABI' in help and use
12055 <option_value> format for -msve-vector-bits and -moverride
12058 2018-08-26 Jeff Law <law@redhat.com>
12060 * config/mips/frame-header-opt.c: Include "backend.h" rather than
12063 2018-08-26 Marek Polacek <polacek@redhat.com>
12065 PR c++/87029, Implement -Wredundant-move.
12066 * doc/invoke.texi: Document -Wredundant-move.
12068 2018-08-25 Martin Sebor <msebor@redhat.com>
12070 PR tree-optimization/87059
12071 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
12072 to the same type as the other.
12073 * fold-const.c (fold_binary_loc): Assert expectation.
12075 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
12077 * config/darwin.c (machopic_legitimize_pic_address): Clean up
12078 extraneous parentheses, dead code section and formatting.
12080 2018-08-24 David Malcolm <dmalcolm@redhat.com>
12083 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
12084 wide enough for jumps in the line-numbering to be visible.
12085 (layout::print_gap_in_line_numbering): New member function.
12086 (layout::calculate_line_spans): When using line numbering, merge
12087 line spans that are only 1 line apart.
12088 (diagnostic_show_locus): When printing line numbers, show gaps in
12089 line numbering directly, rather than printing headers.
12090 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
12091 line-numbering with multiple line spans.
12092 (selftest::test_fixit_insert_containing_newline_2): Add test of
12093 line-numbering, in which the spans are close enough to be merged.
12095 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
12097 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
12098 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
12099 * tree-vrp.c (range_is_nonnull): Remove.
12100 (range_includes_zero_p): Accept value_range instead of min/max.
12101 (extract_range_from_binary_expr_1): Do not early bail on
12103 Use range_includes_zero_p instead of range_is_nonnull.
12104 (extract_range_from_unary_expr): Use range_includes_zero_p instead
12105 of range_is_nonnull.
12106 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
12107 special case VR_ANTI_RANGE.
12108 (vrp_finalize): Same.
12109 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
12110 instead of min/max.
12111 (range_is_nonnull): Remove.
12112 * vr-values.c (vrp_stmt_computes_nonzero): Use
12113 range_includes_zero_p instead of range_is_nonnull.
12114 (extract_range_basic): Pass value_range to range_includes_zero_p
12115 instead of range_is_nonnull.
12117 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
12119 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
12120 * emit-rtl.h (rtl_data): Remove return_bnd.
12121 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
12122 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
12123 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
12124 (POINTER_BOUNDS_MODE): Remove definition.
12125 (make_pointer_bounds_mode): Remove.
12126 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
12127 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
12128 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
12129 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
12130 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
12131 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
12132 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
12134 * config/i386/i386-modes.def (BND32, BND64): Remove.
12135 * config/i386/i386.c (dbx_register_map): Remove bound registers.
12136 (dbx64_register_map): Ditto.
12137 (svr4_dbx_register_map): Ditto.
12138 (indirect_thunk_bnd_needed): Remove.
12139 (indirect_thunks_bnd_used): Ditto.
12140 (indirect_return_bnd_needed): Ditto.
12141 (indirect_return_via_cx_bnd): Ditto.
12142 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
12143 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
12144 (output_indirect_thunk): Ditto. Remove need_prefix argument.
12145 (output_indirect_thunk_function): Remove handling of
12146 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
12147 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
12148 (ix86_save_reg): Remove handling of crtl->return_bnd.
12149 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
12150 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
12151 and UNSPEC_BNDLX_ADDR.
12152 (ix86_output_indirect_branch_via_reg): Remove handling of
12153 indirect_thunk_prefix_bnd.
12154 (ix86_output_indirect_branch_via_push): Ditto.
12155 (ix86_output_function_return): Ditto.
12156 (ix86_output_indirect_function_return): Ditto.
12157 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
12158 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
12159 (CALL_USED_REGISTERS): Ditto.
12160 (REG_ALLOC_ORDER): Update for removal of bound registers.
12161 (HI_REGISTER_NAMES): Ditto.
12162 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
12163 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
12164 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
12165 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
12166 (FIRST_PSEUDO_REG): Update.
12167 (BND): Remove mode iterator.
12168 * config/i386/predicates.md (bnd_mem_operator): Remove.
12170 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
12172 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
12175 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
12177 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
12178 the case in which the permute needs only a single element and
12179 repeats for every vector of the result. Extend that case to
12180 handle variable-length vectors.
12181 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
12183 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
12186 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
12187 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
12189 2018-08-24 Richard Biener <rguenther@suse.de>
12191 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
12192 bb_flags_allocated members.
12193 (auto_flag): New RAII class for allocating flags.
12194 (auto_edge_flag): New RAII class for allocating edge flags.
12195 (auto_bb_flag): New RAII class for allocating bb flags.
12196 * cfgloop.c (verify_loop_structure): Allocate temporary edge
12198 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
12199 in favor of temporarily allocated BB flag.
12200 * hsa-brig.c: Re-order includes.
12201 * hsa-dump.c: Likewise.
12202 * hsa-regalloc.c: Likewise.
12203 * print-rtl.c: Likewise.
12204 * profile-count.c: Likewise.
12206 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
12209 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
12212 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
12215 * wide-int-range.cc (wide_int_range_div): Do not ignore result
12216 from wide_int_range_multiplicative_op.
12218 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12220 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
12223 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
12225 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
12228 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
12230 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
12231 full GENERIC expression used for replacement.
12233 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
12235 * tree-vrp.c (abs_extent_range): Remove.
12236 (extract_range_into_wide_ints): Pass wide ints by reference.
12237 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
12238 Pass wide ints by reference in all calls to
12239 extract_range_into_wide_ints.
12240 * wide-int-range.cc (wide_int_range_div): New.
12241 * wide-int-range.h (wide_int_range_div): New.
12242 (wide_int_range_includes_zero_p): New.
12243 (wide_int_range_zero_p): New.
12245 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
12247 * config/aarch64/aarch64.md (arches): New enum.
12248 (arch): New enum attr.
12249 (arch_enabled): New attr.
12250 (enabled): Now uses arch_enabled only.
12251 (simd, sve, fp16): Removed attribute.
12252 (fp): Attr now defined in terms of 'arch'.
12253 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
12254 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
12255 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
12256 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
12257 attributes into 'arch'.
12258 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
12259 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
12260 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
12261 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
12262 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
12263 'simd' attribute into 'arch'.
12264 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
12265 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
12266 Convert use of 'fp' attribute to 'arch'.
12267 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
12268 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
12270 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
12271 (different modes) Merge 'fp' and 'simd' into 'arch'.
12272 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
12273 'simd' into 'arch'.
12275 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
12277 PR rtl-optimization/87026
12278 * expmed.c (canonicalize_comparison): If we can no longer create
12279 pseudoregisters, don't.
12281 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
12284 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
12286 * config/arm/arm.c (speculation_barrier_libfunc): New static
12288 (arm_init_libfuncs): Initialize it.
12289 (arm_emit_speculation_barrier): New function.
12290 * config/arm/arm.md (speculation_barrier): Call
12291 arm_emit_speculation_barrier for architectures that do not have
12293 (speculation_barrier_insn): Only match on Armv7 or later.
12295 2018-08-23 Richard Biener <rguenther@suse.de>
12297 PR middle-end/87024
12298 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
12301 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
12303 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
12304 of single-vector TBLs.
12305 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
12306 one input is given.
12308 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
12311 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
12312 aarch64_evpc_tbl guard.
12314 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
12316 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
12319 2018-08-22 Martin Sebor <msebor@redhat.com>
12321 PR middle-end/87052
12322 * tree-pretty-print.c (pretty_print_string): Add argument.
12323 (dump_generic_node): Call to pretty_print_string with string size.
12325 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
12327 PR rtl-optimization/86771
12328 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
12329 of two SETs into those two SETs, one to be placed at i2, if that SETs
12330 destination is modified between i2 and i3.
12332 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12334 PR tree-optimization/86725
12335 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
12337 (vect_analyze_scalar_cycles_1): Check it.
12339 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12341 PR tree-optimization/86725
12342 * tree-vect-loop.c (vect_is_simple_reduction): When treating
12343 an outer loop phi as a double reduction, make sure that the
12344 single user of the phi result is an inner loop phi.
12346 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12348 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
12349 grouped stores with gaps to a strided group.
12351 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12353 * tree-vect-stmts.c (get_group_load_store_type)
12354 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
12355 first statement in a group.
12357 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12359 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
12360 the sequence used in gcc/gcc.c.
12362 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12365 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
12368 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12370 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
12371 Darwin10-specific unwinder-shim.
12372 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
12373 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
12374 New to cater for Darwin10 Rosetta.
12376 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12378 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
12379 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
12381 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12386 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
12387 (dwarf2out_switch_text_section): Generate a local label for the
12388 second function sub-section and apply it as the second FDE start
12390 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
12391 second sub-section start.
12393 2018-08-22 Richard Biener <rguenther@suse.de>
12395 PR tree-optimization/86988
12396 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
12398 2018-08-22 Richard Biener <rguenther@suse.de>
12400 PR tree-optimization/86945
12401 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
12403 2018-08-22 Alexandre Oliva <oliva@adacore.com>
12405 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
12406 a comment about how uses of r2 for .sdata2 come about.
12408 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
12410 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
12412 2018-08-21 Marek Polacek <polacek@redhat.com>
12414 PR c++/86981, Implement -Wpessimizing-move.
12415 * doc/invoke.texi: Document -Wpessimizing-move.
12417 2018-08-21 Jan Hubicka <jh@suse.cz>
12419 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
12420 * tree.h (is_redundant_typedef): Remove.
12421 * dwarf2out.c (is_redundant_typedef): Turn into static function.
12423 2018-08-21 Jan Hubicka <jh@suse.cz>
12425 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
12428 2018-08-21 Tamar Christina <tamar.christina@arm.com>
12430 * expmed.c (extract_low_bits): Reject invalid subregs early.
12432 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
12434 PR middle-end/86121
12435 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
12438 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12440 * config/vxworks.h: Guard vxworks_asm_out_constructor and
12441 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
12442 * config/vxworks.c: Likewise.
12444 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12446 * config/vxworks.c: Set targetm.have_ctors_dtors
12447 if HAVE_INITFINI_ARRAY_SUPPORT.
12448 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
12449 if HAVE_INITFINI_ARRAY_SUPPORT.
12451 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12453 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
12454 default search path for VxWorks < 7.
12456 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
12458 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
12459 (get_format_string): Refer to c_getstr.
12461 2018-08-21 Tom de Vries <tdevries@suse.de>
12463 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
12464 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
12465 (debuginfo_early_stop): Declare.
12466 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
12467 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
12468 (debuginfo_early_stop): New function.
12469 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
12470 and debuginfo_early_stop.
12471 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
12472 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
12473 (general_init): Call debuginfo_early_init.
12474 (finalize): Call debuginfo_fini.
12475 (do_compile): Call debuginfo_init.
12476 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
12477 -fdump-early-debug.
12478 (@item -fdump-debug, @item -fdump-earlydebug): Add.
12480 2018-08-21 Tom de Vries <tdevries@suse.de>
12482 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
12483 flag_dump_noaddr and flag_dump_unnumbered.
12485 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
12487 * wide-int-range.cc (wide_int_range_abs): New.
12488 (wide_int_range_order_set): Rename from wide_int_range_min_max.
12489 * wide-int-range.h (wide_int_range_abs): New.
12490 (wide_int_range_min_max): New.
12491 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
12492 case to call wide_int_range_abs.
12493 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
12494 (extract_range_from_abs_expr): Delete.
12496 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
12499 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
12500 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
12503 2018-08-20 Nathan Sidwell <nathan@acm.org>
12504 Jeff Law <law@redhat.com>
12506 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
12507 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
12509 2018-08-20 David Malcolm <dmalcolm@redhat.com>
12512 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
12513 (decl_attributes): Likewise.
12514 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
12516 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12517 * diagnostic-core.h (class auto_diagnostic_group): New class.
12518 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
12519 (diagnostic_report_diagnostic): Handle the first diagnostics within
12521 (emit_diagnostic): Add auto_diagnostic_group instance.
12522 (inform): Likewise.
12523 (inform_n): Likewise.
12524 (warning): Likewise.
12525 (warning_at): Likewise.
12526 (warning_n): Likewise.
12527 (pedwarn): Likewise.
12528 (permerror): Likewise.
12530 (error_n): Likewise.
12531 (error_at): Likewise.
12533 (fatal_error): Likewise.
12534 (internal_error): Likewise.
12535 (internal_error_no_backtrace): Likewise.
12536 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
12537 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
12538 * diagnostic.h (struct diagnostic_context): Add fields
12539 "diagnostic_group_nesting_depth",
12540 "diagnostic_group_emission_count", "begin_group_cb",
12542 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
12543 Add auto_diagnostic_group instance(s).
12544 (find_explicit_erroneous_behavior): Likewise.
12545 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
12546 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
12547 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
12548 (gimplify_va_arg_expr): Likewise.
12549 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
12550 (HSA_SORRY_AT): Likewise.
12551 * ipa-devirt.c (compare_virtual_tables): Likewise.
12552 (warn_odr): Likewise.
12553 * multiple_target.c (expand_target_clones): Likewise.
12554 * opts-common.c (cmdline_handle_error): Likewise.
12555 * reginfo.c (globalize_reg): Likewise.
12556 * substring-locations.c (format_warning_n_va): Likewise.
12557 * tree-inline.c (expand_call_inline): Likewise.
12558 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
12559 * tree-ssa-loop-niter.c
12560 (do_warn_aggressive_loop_optimizations): Likewise.
12561 * tree-ssa-uninit.c (warn_uninit): Likewise.
12562 * tree.c (warn_deprecated_use): Likewise.
12564 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
12567 * config/i386/i386.md (eh_return): Always update EH return
12568 address in word_mode.
12570 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
12572 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
12573 TARGET_SPLIT_COMPLEX_ARG is defined.
12575 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
12577 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
12579 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
12582 * expr.c (expand_assignment): Assert that bitpos is positive.
12583 (store_field): Likewise
12584 (expand_expr_real_1): Make sure that bitpos is positive.
12585 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
12588 2018-08-20 Nathan Sidwell <nathan@acm.org>
12590 * Makefile.in (CPP_ID_DATA_H): Delete.
12591 (CPP_INTERNAL_H): Don't add it.
12592 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
12593 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
12595 2018-08-20 Richard Biener <rguenther@suse.de>
12597 PR tree-optimization/78655
12598 * tree-vrp.c (extract_range_from_binary_expr_1): Make
12599 pointer + offset nonnull if either operand is nonnull work.
12601 2018-08-20 Tom de Vries <tdevries@suse.de>
12603 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
12604 unless the referenced die describes the added property using
12605 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
12606 Otherwise, add a DW_AT_location to the referenced die.
12608 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
12611 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
12612 register_operand when calling ix86_set_reg_reg_cost.
12613 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
12614 Set *total to 0 for operands that satisfy x86_64_immediate_operand
12615 predicate and to 1 otherwise.
12617 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
12619 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
12620 emit a diagnostic that it is not supported and reset the option.
12621 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
12622 supported and consume the option. (ASM_FINAL_SPEC): New.
12624 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
12626 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
12629 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
12631 C-SKY port: Documentation
12633 * doc/extend.texi (C-SKY Function Attributes): New section.
12634 * doc/invoke.texi (Option Summary): Add C-SKY options.
12635 (C-SKY Options): New section.
12636 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
12638 2018-08-17 Jojo <jijie_rong@c-sky.com>
12639 Huibin Wang <huibin_wang@c-sky.com>
12640 Sandra Loosemore <sandra@codesourcery.com>
12641 Chung-Lin Tang <cltang@codesourcery.com>
12643 C-SKY port: Backend implementation
12645 * config/csky/*: New.
12646 * common/config/csky/*: New.
12648 2018-08-17 Jojo <jijie_rong@c-sky.com>
12649 Huibin Wang <huibin_wang@c-sky.com>
12650 Sandra Loosemore <sandra@codesourcery.com>
12651 Chung-Lin Tang <cltang@codesourcery.com>
12652 Andrew Jenner <andrew@codesourcery.com>
12654 C-SKY port: Configury
12656 * config.gcc (csky-*-*): New.
12657 * configure.ac: Add csky to targets for dwarf2 debug_line support.
12658 * configure: Regenerated.
12660 2018-08-17 David Malcolm <dmalcolm@redhat.com>
12662 * dump-context.h: Include "dumpfile.h".
12663 (dump_context::dump_printf_va): Convert final param from va_list
12664 to va_list *. Convert from ATTRIBUTE_PRINTF to
12665 ATTRIBUTE_GCC_DUMP_PRINTF.
12666 (dump_context::dump_printf_loc_va): Likewise.
12667 * dumpfile.c: Include "stringpool.h".
12668 (make_item_for_dump_printf_va): Delete.
12669 (make_item_for_dump_printf): Delete.
12670 (class dump_pretty_printer): New class.
12671 (dump_pretty_printer::dump_pretty_printer): New ctor.
12672 (dump_pretty_printer::emit_items): New member function.
12673 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
12675 (dump_pretty_printer::emit_item): New member function.
12676 (dump_pretty_printer::stash_item): New member function.
12677 (dump_pretty_printer::format_decoder_cb): New member function.
12678 (dump_pretty_printer::decode_format): New member function.
12679 (dump_context::dump_printf_va): Reimplement in terms of
12680 dump_pretty_printer.
12681 (dump_context::dump_printf_loc_va): Convert final param from va_list
12683 (dump_context::begin_scope): Reimplement call to
12684 make_item_for_dump_printf.
12685 (dump_printf): Update for change to dump_printf_va.
12686 (dump_printf_loc): Likewise.
12687 (selftest::test_capture_of_dump_calls): Convert "stmt" from
12688 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
12689 with %T, %E, and %G.
12690 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
12691 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
12692 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
12693 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
12694 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
12695 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
12696 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
12697 within a dump_printf_loc call to "%wu".
12698 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
12699 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
12700 missing space after "=".
12701 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
12702 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
12703 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
12704 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
12705 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
12706 duplicate "vectorized" from message.
12708 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
12710 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
12711 polyNxK_t element's TYPE_STRING_FLAG.
12713 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
12715 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
12716 (they were unnamed before). Fix comments.
12718 2018-08-17 Nathan Sidwell <nathan@acm.org>
12720 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
12722 2018-08-17 Richard Biener <rguenther@suse.de>
12724 PR tree-optimization/86841
12725 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
12727 2018-08-17 Martin Liska <mliska@suse.cz>
12729 * common.opt: Remove Warn, Init and Report for options with
12730 Ignore/Deprecated flag. Warning is done automatically for
12732 * config/i386/i386.opt: Likewise.
12733 * config/ia64/ia64.opt: Likewise.
12734 * config/rs6000/rs6000.opt: Likewise.
12735 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
12736 Remove usage of flag_check_pointer_bounds.
12737 * lto-wrapper.c (merge_and_complain): Do not handle
12738 OPT_fcheck_pointer_bounds.
12739 (append_compiler_options): Likewise.
12740 * opt-functions.awk: Do not handle Deprecated.
12741 * optc-gen.awk: Check that Var, Report and Init are not
12742 used for an option with Ignore/Deprecated flag.
12743 * opts-common.c (decode_cmdline_option): Do not report
12745 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
12747 * opts.h (struct cl_option): Remove cl_deprecated flag.
12748 (CL_ERR_DEPRECATED): Remove error enum value.
12750 2018-08-17 Richard Biener <rguenther@suse.de>
12752 PR middle-end/86505
12753 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
12754 across a va-arg-pack using call adjust its return value accordingly.
12756 2018-08-16 Martin Sebor <msebor@redhat.com>
12758 PR tree-optimization/86853
12759 * gimple-ssa-sprintf.c (struct format_result): Rename member.
12760 (struct fmtresult): Add member and initialize it in ctors.
12761 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
12762 (format_string): Handle %S the same as %ls. Set MAYFAIL.
12763 (format_directive): Set POSUNDER4K when MAYFAIL is set.
12764 (parse_directive): Handle %C same as %c.
12765 (sprintf_dom_walker::compute_format_length): Adjust.
12766 (is_call_safe): Adjust.
12768 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
12770 * builtins.c (c_strlen): Add new parameter eltsize. Use it
12771 for determining how to count the elements.
12772 * builtins.h (c_strlen): Adjust prototype.
12773 * expr.c (string_constant): Add new parameter mem_size.
12774 Set *mem_size appropriately.
12775 * expr.h (string_constant): Adjust protoype.
12776 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
12777 * gimple-fold.h (get_range_strlen): Adjust prototype.
12778 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
12779 (format_string): Call get_string_length with eltsize.
12781 2018-08-16 David Malcolm <dmalcolm@redhat.com>
12783 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
12784 to emit the span, rather than setting it as the prefix.
12786 2018-08-16 David Malcolm <dmalcolm@redhat.com>
12788 * diagnostic-show-locus.c (layout::start_annotation_line): Add
12789 "margin_char" parameter, defaulting to space. Use it in place
12790 of pp_space for the initial part of the margin.
12791 (layout::print_leading_fixits): Use '+' when filling the margin
12792 of line-insertion fix-it hints.
12794 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12796 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
12799 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12801 * config/rs6000/altivec.md: Don't set length attribute to the default
12803 * config/rs6000/darwin.md: Ditto.
12804 * config/rs6000/dfp.md: Ditto.
12805 * config/rs6000/htm.md: Ditto.
12806 * config/rs6000/rs6000.md: Ditto.
12807 * config/rs6000/sync.md: Ditto.
12808 * config/rs6000/vsx.md: Ditto.
12810 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12812 * config/rs6000/altivec.md: Don't set length attribute to the default
12813 value, for branch instructions.
12814 * config/rs6000/darwin.md: Ditto.
12815 * config/rs6000/rs6000.md: Ditto.
12817 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12819 * config/rs6000/rs6000.md (length): Always define as const_int 4.
12820 (unnamed conditional branch define_insn): Set length to 4 or 8
12821 depending on offset.
12822 (<bd>_<mode>): Similar, for alternative 0.
12823 (<bd>tf_<mode>): Ditto.
12825 2018-08-16 Tamar Christina <tamar.christina@arm.com>
12827 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
12829 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
12831 * doc/rtl.texi: Replace old RTX class names with new names.
12834 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
12836 * expmed.h (canonicalize_comparison): New declaration.
12837 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
12838 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
12839 * optabs.c (prepare_cmp_insn): Likewise.
12840 * rtl.h (unsigned_condition_p): New function which checks if a
12841 comparison operator is unsigned.
12843 2018-08-16 Nathan Sidwell <nathan@acm.org>
12845 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
12846 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
12848 2018-08-16 Tamar Christina <tamar.christina@arm.com>
12851 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
12852 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
12853 (mov<mov>): ..this and enable unconditionally.
12855 2018-08-16 Tamar Christina <tamar.christina@arm.com>
12857 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
12859 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
12861 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
12862 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
12863 "Common" with "Target".
12865 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
12867 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
12868 * doc/invoke.texi (mmitigate-rop): Remove.
12869 * config/i386/i386.c: Do not include "regrename.h".
12870 (ix86_rop_should_change_byte_p, reg_encoded_number)
12871 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
12873 (ix86_reorg): Remove call to ix86_mitigate_rop.
12874 * config/i386/i386.md (attr "modrm_class"): Remove.
12875 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
12876 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
12877 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
12879 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
12881 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
12882 allow folding of mergeh() and mergel() for the float and double types.
12883 (fold_mergehl_helper): Rework to handle building a permute tree
12886 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
12888 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
12891 2018-08-15 David Malcolm <dmalcolm@redhat.com>
12893 * common.opt (fdiagnostics-show-labels): New option.
12894 * diagnostic-show-locus.c (class layout_range): Add field
12896 (class layout): Add field "m_show_labels_p".
12897 (layout_range::layout_range): Add param "label" and use it to
12898 initialize m_label.
12899 (make_range): Pass in NULL for new "label" param of layout_range's
12901 (layout::layout): Initialize m_show_labels_p.
12902 (layout::maybe_add_location_range): Pass in loc_range->m_label
12903 when constructing layout_range instances.
12904 (struct line_label): New struct.
12905 (layout::print_any_labels): New member function.
12906 (layout::print_line): Call it if label-printing is enabled.
12907 (selftest::test_one_liner_labels): New test.
12908 (selftest::test_diagnostic_show_locus_one_liner): Call it.
12909 * diagnostic.c (diagnostic_initialize): Initialize
12910 context->show_labels_p.
12911 * diagnostic.h (struct diagnostic_context): Add field
12913 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
12914 -fno-diagnostics-show-labels.
12915 * dwarf2out.c (gen_producer_string): Add
12916 OPT_fdiagnostics_show_labels to the ignored options.
12917 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
12919 (gcc_rich_location::maybe_add_expr): Likewise.
12920 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
12921 label" param, defaulting to NULL.
12922 (gcc_rich_location::add_expr): Add "label" param.
12923 (gcc_rich_location::maybe_add_expr): Likewise.
12924 (class text_range_label): New class.
12925 (class range_label_for_type_mismatch): New class.
12926 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
12927 of format_warning_va.
12928 (fmtwarn_n): Likewise for new params of format_warning_n_va.
12929 * lto-wrapper.c (merge_and_complain): Add
12930 OPT_fdiagnostics_show_labels to the "pick one setting" options.
12931 (append_compiler_options): Likewise to the dropped options.
12932 (append_diag_options): Likewise to the passed-on options.
12933 * opts.c (common_handle_option): Handle the new option.
12934 * selftest-diagnostic.c
12935 (test_diagnostic_context::test_diagnostic_context): Enable
12937 * substring-locations.c: Include "gcc-rich-location.h".
12938 (format_warning_n_va): Add "fmt_label" and "param_label" params
12939 and use them as appropriate.
12940 (format_warning_va): Add "fmt_label" and "param_label" params,
12941 passing them on to format_warning_n_va.
12942 (format_warning_at_substring): Likewise.
12943 (format_warning_at_substring_n): Likewise.
12944 * substring-locations.h (format_warning_va): Add "fmt_label" and
12945 "param_label" params.
12946 (format_warning_n_va): Likewise.
12947 (format_warning_at_substring): Likewise.
12948 (format_warning_at_substring_n): Likewise.
12949 * toplev.c (general_init): Initialize global_dc->show_labels_p.
12951 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
12954 * builtins.c (expand_builtin_memcmp): Do not expand the call
12955 when overflow is detected.
12957 2018-08-15 Martin Sebor <msebor@redhat.com>
12959 PR tree-optimization/71625
12960 * config/aarch64/aarch64-builtins.c
12961 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
12963 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
12965 * config/s390/s390.c (s390_reorg): Remove loop.
12967 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
12970 (darwin_function_switched_text_sections): Delete.
12971 * gcc/config/darwin.h
12972 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
12974 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
12977 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
12978 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
12979 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
12981 2018-08-15 Martin Liska <mliska@suse.cz>
12983 PR tree-optimization/86925
12984 * predict.c (expr_expected_value_1): When taking
12985 later predictor, assign also probability.
12986 Use fold_build2_initializer_loc in order to fold
12987 the expression in -frounding-math.
12989 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
12991 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
12993 (expand_vec_perm_1): Try the new method.
12995 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
12998 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
12999 Check whether lra_live_max_point is 0 before dividing.
13001 2018-08-14 Martin Sebor <msebor@redhat.com>
13003 PR tree-optimization/86650
13004 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
13005 (vrp_prop::check_mem_ref): Same.
13007 2018-08-13 Liu Hao <lh_mouse@126.com>
13009 * pretty-print.c (eat_esc_sequence): Swap the foreground and
13010 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
13011 and clear it thereafter, as it only works for DBCS.
13013 2018-08-13 Liu Hao <lh_mouse@126.com>
13015 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
13016 handle returned by _get_osf_handle().
13018 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
13020 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
13021 for folding vec_perm.
13023 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
13025 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
13026 Add support for gimple-folding of vec_pack() and vec_unpack()
13029 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
13031 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
13032 vec_xst variants to the list.
13033 (rs6000_gimple_fold_builtin): Add support for folding unaligned
13034 vector loads and stores.
13036 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
13038 * config.gcc (rs6000-ibm-aix4.x): Delete.
13039 (rs6000-ibm-aix5.1): Delete.
13040 (rs6000-ibm-aix5.2): Delete.
13041 (rs6000-ibm-aix5.3): Delete.
13042 * config/rs6000/aix43.h: Delete.
13043 * config/rs6000/aix51.h: Delete.
13044 * config/rs6000/aix52.h: Delete.
13045 * config/rs6000/t-aix43: Delete.
13047 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
13049 * config/s390/s390.c (s390_decompose_constant_pool_ref):
13051 (s390_decompose_address): Factor out constant pool ref
13054 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
13056 * config/nds32/nds32-predicates.c
13057 (nds32_can_use_bclr_p): Change return type as bool.
13058 (nds32_can_use_bset_p): Ditto.
13059 (nds32_can_use_btgl_p): Ditto.
13060 (nds32_can_use_bitci_p): Ditto.
13061 * config/nds32/nds32-protos.h
13062 (nds32_can_use_bclr_p): Change declaration.
13063 (nds32_can_use_bset_p): Ditto.
13064 (nds32_can_use_btgl_p): Ditto.
13065 (nds32_can_use_bitci_p): Ditto.
13067 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
13069 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
13070 Support -msched-prolog-epilog option.
13071 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
13073 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
13075 * common/config/nds32/nds32-common.c
13076 (nds32_option_optimization_table): Enalbe -malways-align.
13078 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
13080 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
13082 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
13083 OPT_misr_secure_ case.
13084 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
13085 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
13086 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
13088 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
13089 (nds32_isr_info): New field security_level.
13090 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
13091 * config/nds32/nds32.md (return_internal): Consider critical attribute.
13092 * config/nds32/nds32.opt (misr-secure): New option.
13093 * config/nds32/nds32_init.inc: New file.
13094 * config/nds32/nds32_isr.h: New file.
13096 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
13098 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
13099 Update comment for atomic instructions.
13100 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
13101 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
13103 (atomic_loaddi): Revise fence expansion to only emit fence prior to
13104 load for __ATOMIC_SEQ_CST model.
13105 (atomic_loaddi_1): Remove float register target.
13106 (atomic_storedi): Handle CONST_INT values.
13107 (atomic_storedi_1): Remove float register source. Add special case
13109 (memory_barrier): New expander and insn.
13111 2018-08-11 Jakub Jelinek <jakub@redhat.com>
13113 PR tree-optimization/86835
13114 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
13115 new_stmt after def_gsi, make sure to insert new_square_stmt after
13116 that stmt, not 2 stmts before it.
13118 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
13121 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
13124 2018-08-10 Martin Liska <mliska@suse.cz>
13127 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
13129 * builtins.c (expand_builtin_expect_with_probability):
13131 (expand_builtin_expect_with_probability): New function.
13132 (build_builtin_expect_predicate): Add new argumnet probability
13133 for BUILT_IN_EXPECT_WITH_PROBABILITY.
13134 (fold_builtin_expect):
13137 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
13138 * builtins.h (fold_builtin_expect): Set new argument.
13139 * doc/extend.texi: Document __builtin_expect_with_probability.
13140 * doc/invoke.texi: Likewise.
13141 * gimple-fold.c (gimple_fold_call): Pass new argument.
13142 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
13143 also BUILT_IN_EXPECT_WITH_PROBABILITY.
13144 * predict.c (get_predictor_value): New function.
13145 (expr_expected_value): Add new argument probability. Assume
13146 that predictor and probability are always non-null.
13147 (expr_expected_value_1): Likewise. For __builtin_expect and
13148 __builtin_expect_with_probability set probability. Handle
13149 combination in binary expressions.
13150 (tree_predict_by_opcode): Simplify code by simply calling
13151 get_predictor_value.
13152 (pass_strip_predict_hints::execute): Add handling of
13153 BUILT_IN_EXPECT_WITH_PROBABILITY.
13154 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
13156 * tree.h (DECL_BUILT_IN_P): New function.
13158 2018-08-10 Martin Liska <mliska@suse.cz>
13160 PR tree-optimization/85799
13161 * passes.def: Add argument for pass_strip_predict_hints.
13162 * predict.c (class pass_strip_predict_hints): Add new argument
13164 (strip_predictor_early): New function.
13165 (pass_strip_predict_hints::execute): Call the function to
13167 (strip_predict_hints): New function.
13168 * predict.def: Fix comment.
13170 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
13172 * Makefile.in: Clarify which tm.texi to copy over to assert the
13173 right to grant a GFDL license for all.
13175 2018-08-09 Jeff Law <law@redhat.com>
13177 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
13180 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
13182 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
13185 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
13187 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
13188 reductions for variable-length vectors.
13190 2018-08-09 David Malcolm <dmalcolm@redhat.com>
13193 * common.opt (fdiagnostics-show-line-numbers): New option.
13194 * diagnostic-show-locus.c (class layout): Add fields
13195 "m_show_line_numbers_p" and "m_linenum_width";
13196 (num_digits): New function.
13197 (test_num_digits): New function.
13198 (layout::layout): Initialize new fields. Update m_x_offset
13199 logic to handle any left margin.
13200 (layout::print_source_line): Print line number when requested.
13201 (layout::start_annotation_line): New member function.
13202 (layout::print_annotation_line): Call it.
13203 (layout::print_leading_fixits): Likewise.
13204 (layout::print_trailing_fixits): Likewise. Update calls to
13205 move_to_column for new parameter.
13206 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
13207 it to potentially call start_annotation_line.
13208 (layout::show_ruler): Call start_annotation_line.
13209 (selftest::test_line_numbers_multiline_range): New selftest.
13210 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
13211 and selftest::test_line_numbers_multiline_range.
13212 * diagnostic.c (diagnostic_initialize): Initialize
13213 show_line_numbers_p.
13214 * diagnostic.h (struct diagnostic_context): Add field
13215 "show_line_numbers_p".
13216 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
13217 -fno-diagnostics-show-line-numbers.
13218 * dwarf2out.c (gen_producer_string): Add
13219 OPT_fdiagnostics_show_line_numbers to the ignored options.
13220 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
13221 one setting" options.
13222 (append_compiler_options): Likewise to the dropped options.
13223 (append_diag_options): Likewise to the passed-on options.
13224 * opts.c (common_handle_option): Handle the new option.
13225 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
13227 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
13229 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
13230 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
13231 third argument of type "const signed char" to descriptions of
13232 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
13233 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
13234 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
13235 __builtin_bcdsub_ov functions.
13237 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
13239 PR tree-optimization/86858
13240 * tree-vect-loop.c (vect_is_simple_reduction): Restore
13241 flow_bb_inside_loop_p calls.
13243 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
13245 PR tree-optimization/86871
13246 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
13247 instead of gimple_assign_lhs.
13249 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
13252 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
13253 register constraint to operand 0.
13254 (add<mode>3_carryinC): Likewise.
13255 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
13257 2018-08-09 Martin Liska <mliska@suse.cz>
13260 * common.opt: Remove extra line.
13262 2018-08-09 Martin Liska <mliska@suse.cz>
13264 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
13265 at the end of a line, make first letter capital and end up
13266 a sentence with a dot.
13267 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
13268 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
13269 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
13270 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
13271 (PARAM_MAX_ISL_OPERATIONS): Likewise.
13272 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
13273 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
13274 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
13275 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
13276 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
13277 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
13278 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
13279 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
13280 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
13281 (PARAM_TREE_REASSOC_WIDTH): Likewise.
13282 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
13283 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
13284 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
13286 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
13289 * config/s390/s390.c (s390_option_override_internal): Reduce the
13290 stack-clash-protection-probe-interval param if it would be too big
13293 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
13296 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
13297 * config/m68k/m68k.c (handle_move_double): Don't call
13298 m68k_final_prescan_insn.
13299 (m68k_adjust_decorated_operand): Renamed from
13300 m68k_final_prescan_insn, remove first and third operand and
13302 (print_operand): Call it.
13303 (print_operand_address): Call it.
13305 2018-08-08 Nathan Sidwell <nathan@acm.org>
13307 * diagnostic.c (diagnostic_report_current_module): Use
13308 linemap_included_from & linemap_included_from_linemap.
13310 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
13312 * config/aarch64/aarch64-cores.def: Add phecda core.
13313 * config/aarch64/aarch64-tune.md: Regenerate.
13314 * doc/invoke.texi: Add phecda core.
13316 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
13319 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
13321 * config/s390/s390.md ("movti"): Add more alternatives for
13322 constant to GPR copies.
13324 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
13326 * config/s390/s390.c: Fix whitespace damage throughout the file.
13327 * config/s390/s390.h: Likewise.
13328 * config/s390/tpf.h: Likewise.
13330 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
13332 * config/s390/s390.c (s390_loadrelative_operand_p):
13333 Remove TARGET_CPU_ZARCH usages.
13334 (s390_rtx_costs): Likewise.
13335 (s390_legitimate_constant_p): Likewise.
13336 (s390_cannot_force_const_mem): Likewise.
13337 (legitimate_reload_constant_p): Likewise.
13338 (s390_preferred_reload_class): Likewise.
13339 (legitimize_pic_address): Likewise.
13340 (legitimize_tls_address): Likewise.
13341 (s390_split_branches): Removed.
13342 (s390_add_execute): Removed.
13343 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
13344 (s390_mainpool_start): Likewise.
13345 (s390_mainpool_finish): Likewise.
13346 (s390_mainpool_cancel): Removed.
13347 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
13348 (s390_chunkify_cancel): Likewise.
13349 (s390_return_addr_rtx): Likewise.
13350 (s390_register_info): Remove split_branches_pending_p uages.
13351 (s390_optimize_register_info): Likewise.
13352 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
13353 split_branches_pending_p usages.
13354 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
13355 (s390_load_got): Likewise.
13356 (s390_expand_split_stack_prologue): Likewise.
13357 (output_asm_nops): Likewise.
13358 (s390_function_profiler): Likewise.
13359 (s390_emit_call): Likewise.
13360 (s390_conditional_register_usage): Likewise.
13361 (s390_optimize_prologue): Likewise.
13362 (s390_reorg): Remove TARGET_CPU_ZARCH and
13363 split_branches_pending_p usages.
13364 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
13366 (s390_output_indirect_thunk_function): Likewise.
13367 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
13368 (TARGET_CPU_ZARCH_P): Removed.
13369 (struct machine_function): Remove split_branches_pending_p.
13370 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
13372 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
13374 * common/config/s390/s390-common.c (processor_flags_table):
13376 * config.gcc: Remove with_arch/with_tune support.
13377 * config/s390/2064.md: Remove cpu attribute comparisons.
13378 * config/s390/driver-native.c (s390_host_detect_local_cpu):
13380 * config/s390/linux.h (ASM_SPEC):
13381 Remove -march support.
13382 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
13383 Use a table to get an arch level.
13384 * config/s390/s390-opts.h (enum processor_type):
13385 Remove enum values.
13386 * config/s390/s390.c
13387 (processor_table): Remove entries, add arch_level values.
13388 (s390_issue_rate): Remove cases.
13389 (s390_option_override): Adjust
13390 s390_option_override_internal() call.
13391 (s390_option_override_internal): Remove deprecation warning.
13392 (s390_valid_target_attribute_tree): Adjust
13393 s390_option_override_internal() call.
13394 * config/s390/s390.h (struct s390_processor):
13395 Share with s390-c.c, add arch_level field.
13396 * config/s390/s390.md:
13397 Remove occurrences in cpu attribute.
13398 * config/s390/s390.opt: Remove -march/-mtune support.
13399 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
13400 * doc/invoke.texi: Remove deprecation warning.
13402 2018-08-08 Luis Machado <luis.machado@linaro.org>
13404 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
13406 (qdf24xx_tunings): Set vector cost structure to
13407 qdf24xx_vector_cost.
13409 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
13410 <register_sextend>: Set to 3.
13412 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
13415 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
13416 * config/aarch64/aarch64-simd.md
13417 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
13418 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
13419 (aarch64_frecpx<mode>): ...this new pattern.
13420 * config/aarch64/aarch64-simd-builtins.def: Remove comment
13421 about aarch64_frecp<FRECP:frecp_suffix><mode>.
13423 2018-08-07 Martin Liska <mliska@suse.cz>
13425 PR middle-end/83023
13426 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
13427 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
13428 * predict.def (PRED_MALLOC_NONNULL): New predictor.
13429 * doc/extend.texi: Document that malloc attribute adds
13432 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
13435 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13436 Define to speculation_safe_value_not_needed.
13438 2018-08-06 Jeff Law <law@redhat.com>
13440 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
13441 the vr_values instance to cprop_into_stmt.
13442 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
13443 (cprop_operand): Also query EVRP to determine if OP is a constant.
13445 2018-08-06 Nathan Sidwell <nathan@acm.org>
13447 * diagnostic.c (diagnostic_report_current_module): Reroll
13448 included-at loop. Translate text.
13450 2018-08-06 David Malcolm <dmalcolm@redhat.com>
13452 * function-tests.c (selftest::test_expansion_to_rtl): Call
13453 free_after_compilation.
13455 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13457 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
13459 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
13461 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
13462 loops with memory block operations from getting unrolled.
13464 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
13467 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13468 Define to speculation_safe_value_not_needed.
13470 2018-08-06 Jeff Law <law@redhat.com>
13472 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
13475 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13478 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
13479 with all enabled __intN types.
13481 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
13483 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13485 * alias.c (record_set): Check for clobber high.
13486 * cfgexpand.c (expand_gimple_stmt): Likewise.
13487 * combine-stack-adj.c (single_set_for_csa): Likewise.
13488 * combine.c (find_single_use_1): Likewise.
13489 (set_nonzero_bits_and_sign_copies): Likewise.
13490 (get_combine_src_dest): Likewise.
13491 (is_parallel_of_n_reg_sets): Likewise.
13492 (try_combine): Likewise.
13493 (record_dead_and_set_regs_1): Likewise.
13494 (reg_dead_at_p_1): Likewise.
13495 (reg_dead_at_p): Likewise.
13496 * dce.c (deletable_insn_p): Likewise.
13497 (mark_nonreg_stores_1): Likewise.
13498 (mark_nonreg_stores_2): Likewise.
13499 * df-scan.c (df_find_hard_reg_defs): Likewise.
13500 (df_uses_record): Likewise.
13501 (df_get_call_refs): Likewise.
13502 * dwarf2out.c (mem_loc_descriptor): Likewise.
13503 * haifa-sched.c (haifa_classify_rtx): Likewise.
13504 * ira-build.c (create_insn_allocnos): Likewise.
13505 * ira-costs.c (scan_one_insn): Likewise.
13506 * ira.c (equiv_init_movable_p): Likewise.
13507 (rtx_moveable_p): Likewise.
13508 (interesting_dest_for_shprep): Likewise.
13509 * jump.c (mark_jump_label_1): Likewise.
13510 * postreload-gcse.c (record_opr_changes): Likewise.
13511 * postreload.c (reload_cse_simplify): Likewise.
13512 (struct reg_use): Add source expr.
13513 (reload_combine): Check for clobber high.
13514 (reload_combine_note_use): Likewise.
13515 (reload_cse_move2add): Likewise.
13516 (move2add_note_store): Likewise.
13517 * print-rtl.c (print_pattern): Likewise.
13518 * recog.c (decode_asm_operands): Likewise.
13519 (store_data_bypass_p): Likewise.
13520 (if_test_bypass_p): Likewise.
13521 * regcprop.c (kill_clobbered_value): Likewise.
13522 (kill_set_value): Likewise.
13523 * reginfo.c (reg_scan_mark_refs): Likewise.
13524 * reload1.c (maybe_fix_stack_asms): Likewise.
13525 (eliminate_regs_1): Likewise.
13526 (elimination_effects): Likewise.
13527 (mark_not_eliminable): Likewise.
13528 (scan_paradoxical_subregs): Likewise.
13529 (forget_old_reloads_1): Likewise.
13530 * reorg.c (find_end_label): Likewise.
13531 (try_merge_delay_insns): Likewise.
13532 (redundant_insn): Likewise.
13533 (own_thread_p): Likewise.
13534 (fill_simple_delay_slots): Likewise.
13535 (fill_slots_from_thread): Likewise.
13536 (dbr_schedule): Likewise.
13537 * resource.c (update_live_status): Likewise.
13538 (mark_referenced_resources): Likewise.
13539 (mark_set_resources): Likewise.
13540 * rtl.c (copy_rtx): Likewise.
13541 * rtlanal.c (reg_referenced_p): Likewise.
13542 (single_set_2): Likewise.
13543 (noop_move_p): Likewise.
13544 (note_stores): Likewise.
13545 * sched-deps.c (sched_analyze_reg): Likewise.
13546 (sched_analyze_insn): Likewise.
13548 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13550 * cse.c (invalidate_reg): New function extracted from...
13551 (invalidate): ...here.
13552 (canonicalize_insn): Check for clobber high.
13553 (invalidate_from_clobbers): invalidate clobber highs.
13554 (invalidate_from_sets_and_clobbers): Likewise.
13555 (count_reg_usage): Check for clobber high.
13556 (insn_live_p): Likewise.
13557 * cselib.c (cselib_expand_value_rtx_1):Likewise.
13558 (cselib_invalidate_regno): Check for clobber in setter.
13559 (cselib_invalidate_rtx): Pass through setter.
13560 (cselib_invalidate_rtx_note_stores):
13561 (cselib_process_insn): Check for clobber high.
13562 * cselib.h (cselib_invalidate_rtx): Add operand.
13564 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13566 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
13567 (mark_not_eliminable): Likewise.
13568 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
13569 * lra-lives.c (process_bb_lives): Check for clobber high.
13570 * lra.c (new_insn_reg): Remember clobber highs.
13571 (collect_non_operand_hard_regs): Check for clobber high.
13572 (lra_set_insn_recog_data): Likewise.
13573 (add_regs_to_insn_regno_info): Likewise.
13574 (lra_update_insn_regno_info): Likewise.
13576 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13578 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
13579 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
13581 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13583 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
13584 (copy_insn_1): Likewise.
13585 (gen_hard_reg_clobber_high): New gen function.
13586 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
13587 * genemit.c (gen_exp): Likewise.
13588 (gen_emit_seq): Pass through info.
13589 (gen_insn): Check for CLOBBER_HIGH.
13590 (gen_expand): Pass through info.
13591 (gen_split): Likewise.
13592 (output_add_clobbers): Likewise.
13593 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
13594 (remove_clobbers): Likewise.
13595 * rtl.h (gen_hard_reg_clobber_high): New declaration.
13597 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13599 * doc/rtl.texi (clobber_high): Add.
13600 (parallel): Add in clobber high
13601 * rtl.c (rtl_check_failed_code3): Add function.
13602 * rtl.def (CLOBBER_HIGH): Add expression.
13603 * rtl.h (RTL_CHECKC3): Add macro.
13604 (rtl_check_failed_code3): Add declaration.
13605 (XC3EXP): Add macro.
13607 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
13610 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
13611 cfun->machine->max_used_stack_alignment if needed.
13613 2018-08-04 Martin Sebor <msebor@redhat.com>
13615 PR tree-optimization/86571
13616 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
13619 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
13622 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13625 2018-08-03 Jeff Law <law@redhat.com>
13628 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13629 Define to speculation_safe_value_not_needed.
13631 2018-08-03 David Malcolm <dmalcolm@redhat.com>
13633 * doc/gcov.texi (-x): Remove duplicate "to".
13634 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
13635 (-Wif-not-aligned): Remove duplicate "is".
13636 (-flto): Remove duplicate "the".
13637 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
13638 duplicate "v5.00.b".
13639 (MSP430 Options): Remove duplicate "and" from the description
13640 of "-mgprel-sec=regexp".
13641 (x86 Options): Remove duplicate copies of "vmldLog102" and
13642 vmlsLog104 from description of "-mveclibabi=type".
13644 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
13646 * internal-fn.h (first_commutative_argument): Declare.
13647 * internal-fn.c (first_commutative_argument): New function.
13648 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
13649 restrictions for pattern statements. Use first_commutative_argument
13650 to look for commutative operands in calls to internal functions.
13652 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
13654 * Makefile.in (wide-int-range.o): New.
13655 * tree-vrp.c: Move all the wide_int_* functions to...
13656 * wide-int-range.cc: ...here.
13657 * tree-vrp.h: Move all the wide_int_* prototypes to...
13658 * wide-int-range.h: ...here.
13660 2018-08-03 Tom de Vries <tdevries@suse.de>
13662 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
13664 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
13665 * except.c (output_function_exception_table): Do early exit if
13666 targetm_common.except_unwind_info (&global_options) == UI_NONE.
13668 2018-08-03 Martin Liska <mliska@suse.cz>
13670 * predict.c (dump_prediction): Change to 2 digits
13673 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
13675 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
13677 (falkor_am_1_gtov_gtov): ... a new insn reservation.
13679 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
13681 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
13682 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
13683 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
13684 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
13685 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
13687 2018-08-02 David Malcolm <dmalcolm@redhat.com>
13689 * diagnostic-show-locus.c (diagnostic_show_locus): Use
13690 pp_take_prefix when saving the existing prefix.
13691 * diagnostic.c (diagnostic_append_note): Likewise.
13692 * langhooks.c (lhd_print_error_function): Likewise.
13693 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
13694 param's type. Free the existing prefix.
13695 (pp_take_prefix): New function.
13696 (pretty_printer::pretty_printer): Drop the prefix parameter.
13697 Rename the length parameter to match the comment.
13698 (pretty_printer::~pretty_printer): Free the prefix.
13699 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
13701 (struct pretty_printer): Drop the "const" from "prefix" field's
13702 type and clarify memory management.
13703 (pp_set_prefix): Drop the "const" from the 2nd param.
13704 (pp_take_prefix): New decl.
13706 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
13708 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
13709 for word_mode_ok here instead of passing as argument.
13710 (expand_block_compare): Change select_block_compare_mode() call.
13711 (expand_strncmp_gpr_sequence): New function.
13712 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
13714 2018-08-02 Jeff Law <law@redhat.com>
13717 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13718 Define to speculation_safe_value_not_needed.
13721 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13722 Define to speculation_safe_value_not_needed.
13724 2018-08-02 Tom de Vries <tdevries@suse.de>
13727 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
13728 function. Return UI_TARGET unconditionally.
13729 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
13730 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
13732 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
13734 * genemit.c (print_overload_test): Fix typo.
13736 2018-08-02 Richard Biener <rguenther@suse.de>
13738 PR tree-optimization/86816
13739 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
13740 which checks for value availability before querying it.
13741 (gvn_uses_equal): Use it.
13742 (same_succ_hash): Likewise.
13743 (gimple_equal_p): Likewise.
13745 2018-08-02 Nick Clifton <nickc@redhat.com>
13748 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13749 Define to speculation_safe_value_not_needed.
13752 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13753 Define to speculation_safe_value_not_needed.
13756 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13757 Define to speculation_safe_value_not_needed.
13760 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13761 Define to speculation_safe_value_not_needed.
13764 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13765 Define to speculation_safe_value_not_needed.
13768 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13769 Define to speculation_safe_value_not_needed.
13772 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13773 Define to speculation_safe_value_not_needed.
13776 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13777 Define to speculation_safe_value_not_needed.
13780 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
13781 speculation_safe_value_not_needed.
13784 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
13785 to speculation_safe_value_not_needed.
13787 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
13789 * doc/md.texi: Expand the documentation of instruction names
13790 to mention port-local uses. Document '@' in pattern names.
13791 * read-md.h (overloaded_instance, overloaded_name): New structs.
13792 (mapping): Declare.
13793 (md_reader::handle_overloaded_name): New member function.
13794 (md_reader::get_overloads): Likewise.
13795 (md_reader::m_first_overload): New member variable.
13796 (md_reader::m_next_overload_ptr): Likewise.
13797 (md_reader::m_overloads_htab): Likewise.
13798 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
13799 m_next_overload_ptr and m_overloads_htab.
13800 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
13801 (get_mode_token, get_code_token, get_int_token): New functions.
13802 (map_attr_string): Add an optional argument that passes back
13803 the associated iterator.
13804 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
13805 (md_reader::handle_overloaded_name, add_overload_instance): New
13807 (apply_iterators): Handle '@' names. Report an error if '@'
13808 is used without iterators.
13809 (initialize_iterators): Initialize the new iterator_group fields.
13810 * genopinit.c (handle_overloaded_code_for)
13811 (handle_overloaded_gen): New functions.
13812 (main): Use them to print declarations of maybe_code_for_* and
13813 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
13814 * genemit.c (print_overload_arguments, print_overload_test)
13815 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
13816 (main): Use it to print definitions of maybe_code_for_* and
13817 maybe_gen_* functions.
13818 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
13819 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
13820 instead of explicit mode checks.
13821 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
13822 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
13823 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
13824 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
13825 (aarch64_expand_compare_and_swap): Likewise
13826 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
13827 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
13828 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
13829 (aarch64_constant_pool_reload_icode): Delete.
13830 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
13831 instead of aarch64_constant_pool_reload_icode. Use
13832 code_for_aarch64_reload_mov instead of explicit mode checks.
13833 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
13834 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
13835 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
13836 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
13837 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
13838 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
13839 (aarch64_atomic_load_op_code): Delete.
13840 (aarch64_emit_atomic_load_op): Likewise.
13841 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
13842 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
13843 instead of aarch64_emit_atomic_load_op.
13844 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
13845 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
13846 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
13847 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
13848 character before the pattern name.
13849 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
13850 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
13851 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
13852 (aarch64_frecps<mode>): Likewise.
13853 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
13854 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
13855 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
13856 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
13857 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
13859 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
13861 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
13862 Allow HFmode constants if TARGET_FP_F16INST.
13864 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
13867 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
13868 No longer check last store for clobber of address register.
13870 2018-08-02 Martin Liska <mliska@suse.cz>
13872 PR gcov-profile/86817
13873 * gcov.c (process_all_functions): New function.
13875 (process_file): Move functions processing to
13876 process_all_functions.
13878 2018-08-02 David Malcolm <dmalcolm@redhat.com>
13880 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
13881 "const" to the "gimple *" and "rtx_insn *" parameters.
13882 * dumpfile.h (dump_user_location_t::dump_user_location_t):
13884 (dump_location_t::dump_location_t): Likewise.
13886 2018-08-01 Martin Sebor <msebor@redhat.com>
13888 PR tree-optimization/86650
13889 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
13890 rather than a "gcall *". Directly pass the data of interest
13891 to percent_K_format, rather than building a temporary CALL_EXPR
13893 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
13894 (gimple_fold_builtin_strncat): Adjust.
13895 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
13896 gcall* argument with gimple*.
13897 * gimple-ssa-warn-restrict.c (check_call): Same.
13898 (wrestrict_dom_walker::before_dom_children): Same.
13899 (builtin_access::builtin_access): Same.
13900 (check_bounds_or_overlap): Same
13901 (maybe_diag_overlap): Same.
13902 (maybe_diag_offset_bounds): Same.
13903 * tree-diagnostic.c (default_tree_printer): Move usage of
13904 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
13906 * tree-pretty-print.c (percent_K_format): Add argument.
13907 * tree-pretty-print.h: Add argument.
13908 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
13909 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
13910 (maybe_diag_stxncpy_trunc): Same.
13911 (handle_builtin_stxncpy): Same.
13912 (handle_builtin_strcat): Same.
13914 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13916 * match.pd: Optimise pointer range checks.
13918 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13920 PR tree-optimization/86758
13921 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
13922 to remove pattern statements.
13924 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13926 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
13927 result of dfs_enumerate_from when constructing stmt_vec_infos,
13928 instead of additionally calling get_loop_body.
13930 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13932 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
13934 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
13935 When creating an iv, assert that the step is not known to be zero.
13936 (vect_setup_realignment): Update call accordingly.
13937 * tree-vect-stmts.c (vectorizable_store): Likewise.
13938 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
13940 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13942 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
13943 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
13944 (vectorizable_reduction): Likewise.
13945 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13946 (vect_detect_hybrid_slp_stmts): Likewise.
13947 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13949 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
13951 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
13952 (wide_int_set_zero_nonzero_bits): ...this.
13953 (zero_nonzero_bits_from_vr): Rename to...
13954 (vrp_set_zero_nonzero_bits): ...this.
13955 (extract_range_from_multiplicative_op_1): Abstract wide int
13957 (wide_int_range_multiplicative_op): ...here.
13958 (extract_range_from_binary_expr_1): Extract wide int binary
13959 operations into their own functions.
13960 (wide_int_range_lshift): New.
13961 (wide_int_range_can_optimize_bit_op): New.
13962 (wide_int_range_shift_undefined_p): New.
13963 (wide_int_range_bit_xor): New.
13964 (wide_int_range_bit_ior): New.
13965 (wide_int_range_bit_and): New.
13966 (wide_int_range_trunc_mod): New.
13967 (extract_range_into_wide_ints): New.
13968 (vrp_shift_undefined_p): New.
13969 (extract_range_from_multiplicative_op): New.
13970 (vrp_can_optimize_bit_op): New.
13971 * tree-vrp.h (value_range::dump): New.
13972 (wide_int_range_multiplicative_op): New.
13973 (wide_int_range_lshift):New.
13974 (wide_int_range_shift_undefined_p): New.
13975 (wide_int_range_bit_xor): New.
13976 (wide_int_range_bit_ior): New.
13977 (wide_int_range_bit_and): New.
13978 (wide_int_range_trunc_mod): New.
13979 (zero_nonzero_bits_from_bounds): Rename to...
13980 (wide_int_set_zero_nonzero_bits): ...this.
13981 (zero_nonzero_bits_from_vr): Rename to...
13982 (vrp_set_zero_nonzero_bits): ...this.
13983 (range_easy_mask_min_max): Rename to...
13984 (wide_int_range_can_optimize_bit_op): this.
13985 * vr-values.c (simplify_bit_ops_using_ranges): Rename
13986 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
13988 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13990 * tree-vectorizer.h (vect_orig_stmt): New function.
13991 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
13992 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
13993 (vect_create_epilog_for_reduction): Likewise.
13994 (vectorizable_live_operation): Likewise.
13995 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
13996 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
13997 * tree-vect-stmts.c (vectorizable_call): Likewise.
13998 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
14000 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
14002 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
14004 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
14005 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
14006 (vect_transform_loop): Likewise.
14007 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
14009 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
14011 * tree-vectorizer.h (vect_schedule_slp): Return void.
14012 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
14013 (vect_schedule_slp): Likewise.
14015 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
14017 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
14019 (vect_transform_loop): Update calls accordingly. Schedule SLP
14020 instances before the main loop, if any exist.
14022 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
14024 PR tree-optimization/86749
14025 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
14026 If the lhs is used in a COND_EXPR, check that it is being used
14027 as the "then" or "else" value.
14029 2018-08-01 Tom de Vries <tdevries@suse.de>
14032 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
14033 speculation_safe_value_not_needed.
14035 2018-08-01 Richard Biener <rguenther@suse.de>
14037 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
14038 as base and offset.
14040 2018-08-01 Martin Liska <mliska@suse.cz>
14042 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
14043 format how successful transformation is dumped.
14044 (gimple_mod_pow2_value_transform): Likewise.
14045 (gimple_mod_subtract_transform): Likewise.
14046 (gimple_stringops_transform): Likewise.
14048 2018-08-01 Martin Liska <mliska@suse.cz>
14050 PR value-prof/35543
14051 * value-prof.c (interesting_stringop_to_profile_p):
14052 Simplify the code and add BUILT_IN_MEMMOVE.
14053 (gimple_stringops_transform): Likewise.
14055 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
14057 * config/aarch64/aarch64-simd.md
14058 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
14059 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
14060 use GPI iterator instead of SI mode.
14062 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
14064 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
14065 rs6000_speculation_barrier.
14066 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
14067 new barrier pattern name.
14069 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
14071 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
14072 (speculation_barrier): New insn.
14074 2018-08-01 Richard Biener <rguenther@suse.de>
14077 * graphite.h: Include isl/id.h and isl/space.h to allow build
14080 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
14083 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14084 mode for COFF targets.
14085 * defaults.h (TARGET_COFF): Define.
14086 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
14087 TARGET_COFF): Define.
14088 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
14089 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
14090 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
14092 2018-07-31 Alexandre Oliva <oliva@adacore.com>
14094 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
14095 * gimple-streamer-out.c (output_bb): Save it.
14096 * lto-streamer-in.c (input_struct_function_base): Restore
14097 instance discriminator if available. Create map on demand.
14098 * lto-streamer-out.c (output_struct_function_base): Save it if
14100 * final.c (decl_to_instance_map): Document LTO strategy.
14102 2018-07-31 Alexandre Oliva <oliva@adacore.com>
14103 Olivier Hainque <hainque@adacore.com>
14105 * debug.h (decl_to_instance_map_t): New type.
14106 (decl_to_instance_map): Declare.
14107 (maybe_create_decl_to_instance_map): New inline function.
14108 * final.c (bb_discriminator, last_bb_discriminator): New statics,
14109 to track basic block discriminators.
14110 (final_start_function_1): Initialize them.
14111 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
14113 (decl_to_instance_map): New variable.
14114 (map_decl_to_instance, maybe_set_discriminator): New functions.
14115 (notice_source_line): Set discriminator.
14117 2018-07-31 Ian Lance Taylor <iant@golang.org>
14119 * targhooks.c (default_have_speculation_safe_value): Add
14122 2018-07-31 David Malcolm <dmalcolm@redhat.com>
14124 * dump-context.h: Include "pretty-print.h".
14125 (dump_context::refresh_dumps_are_enabled): New decl.
14126 (dump_context::emit_item): New decl.
14127 (class dump_context): Add fields "m_test_pp" and
14129 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
14130 (temp_dump_context::get_dumped_text): New decl.
14131 (class temp_dump_context): Add field "m_pp".
14132 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
14133 (dump_context::refresh_dumps_are_enabled): ...and add a test for
14135 (set_dump_file): Update for above change.
14136 (set_alt_dump_file): Likewise.
14137 (dump_loc): New overload, taking a pretty_printer *.
14138 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
14139 to any test pretty-printer.
14140 (make_item_for_dump_gimple_stmt): New function, adapted from
14141 optinfo::add_gimple_stmt.
14142 (dump_context::dump_gimple_stmt): Call it, and use the result,
14143 eliminating the direct usage of dump_file and alt_dump_file in
14144 favor of indirectly using them via emit_item.
14145 (make_item_for_dump_gimple_expr): New function, adapted from
14146 optinfo::add_gimple_expr.
14147 (dump_context::dump_gimple_expr): Call it, and use the result,
14148 eliminating the direct usage of dump_file and alt_dump_file in
14149 favor of indirectly using them via emit_item.
14150 (make_item_for_dump_generic_expr): New function, adapted from
14152 (dump_context::dump_generic_expr): Call it, and use the result,
14153 eliminating the direct usage of dump_file and alt_dump_file in
14154 favor of indirectly using them via emit_item.
14155 (make_item_for_dump_printf_va): New function, adapted from
14156 optinfo::add_printf_va.
14157 (make_item_for_dump_printf): New function.
14158 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
14159 and use the result, eliminating the direct usage of dump_file and
14160 alt_dump_file in favor of indirectly using them via emit_item.
14161 (make_item_for_dump_dec): New function.
14162 (dump_context::dump_dec): Call it, and use the result,
14163 eliminating the direct usage of dump_file and alt_dump_file in
14164 favor of indirectly using them via emit_item.
14165 (make_item_for_dump_symtab_node): New function, adapted from
14166 optinfo::add_symtab_node.
14167 (dump_context::dump_symtab_node): Call it, and use the result,
14168 eliminating the direct usage of dump_file and alt_dump_file in
14169 favor of indirectly using them via emit_item.
14170 (dump_context::begin_scope): Reimplement, avoiding direct usage
14171 of dump_file and alt_dump_file in favor of indirectly using them
14173 (dump_context::emit_item): New member function.
14174 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
14175 Set up test pretty-printer on the underlying context. Call
14176 refresh_dumps_are_enabled.
14177 (temp_dump_context::~temp_dump_context): Call
14178 refresh_dumps_are_enabled.
14179 (temp_dump_context::get_dumped_text): New member function.
14180 (selftest::verify_dumped_text): New function.
14181 (ASSERT_DUMPED_TEXT_EQ): New macro.
14182 (selftest::test_capture_of_dump_calls): Run all tests twice, with
14183 and then without optinfo enabled. Add uses of
14184 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
14185 * dumpfile.h: Update comment for the dump_* API.
14186 * optinfo-emit-json.cc
14187 (selftest::test_building_json_from_dump_calls): Update for new
14188 param for temp_dump_context ctor.
14189 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
14190 and "m_owned" field.
14191 (optinfo_item::~optinfo_item): Likewise.
14192 (optinfo::add_item): New member function.
14193 (optinfo::emit): Update comment.
14194 (optinfo::add_string): Delete.
14195 (optinfo::add_printf): Delete.
14196 (optinfo::add_printf_va): Delete.
14197 (optinfo::add_gimple_stmt): Delete.
14198 (optinfo::add_gimple_expr): Delete.
14199 (optinfo::add_tree): Delete.
14200 (optinfo::add_symtab_node): Delete.
14201 (optinfo::add_dec): Delete.
14202 * optinfo.h (class dump_context): New forward decl.
14203 (optinfo::add_item): New decl.
14204 (optinfo::add_string): Delete.
14205 (optinfo::add_printf): Delete.
14206 (optinfo::add_printf_va): Delete.
14207 (optinfo::add_gimple_stmt): Delete.
14208 (optinfo::add_gimple_expr): Delete.
14209 (optinfo::add_tree): Delete.
14210 (optinfo::add_symtab_node): Delete.
14211 (optinfo::add_dec): Delete.
14212 (optinfo::add_poly_int): Delete.
14213 (optinfo_item::optinfo_item): Remove "owned" param.
14214 (class optinfo_item): Remove field "m_owned".
14216 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14218 PR middle-end/86705
14219 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
14220 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
14221 requested variable alignment.
14222 (expand_one_ssa_partition): Likewise.
14223 (expand_one_var): Likewise.
14225 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14227 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
14228 to speculation_safe_value_not_needed.
14230 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14232 * targhooks.h (speculation_safe_value_not_needed): New prototype.
14233 * targhooks.c (speculation_safe_value_not_needed): New function.
14234 * target.def (have_speculation_safe_value): Update documentation.
14235 * doc/tm.texi: Regenerated.
14237 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14239 * config/aarch64/iterators.md (ALLI_TI): New iterator.
14240 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
14242 (despeculate_copy<ALLI:mode>_insn): New insn.
14243 (despeculate_copyti_insn): New insn.
14244 (despeculate_simple<ALLI:mode>): New insn
14245 (despeculate_simpleti): New insn.
14246 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
14248 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
14249 aarch64_speculation_safe_value.
14250 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
14252 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14254 * config/aarch64/aarch64-speculation.cc: New file.
14255 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
14256 before pass_reorder_blocks.
14257 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
14259 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
14260 X14 and X15 when tracking speculation.
14261 * config/aarch64/aarch64.md (register name constants): Add
14262 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
14263 (unspec): Add UNSPEC_SPECULATION_TRACKER.
14264 (speculation_barrier): New insn attribute.
14265 (cmp<mode>): Allow SP in comparisons.
14266 (speculation_tracker): New insn.
14267 (speculation_barrier): Add speculation_barrier attribute.
14268 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
14269 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
14270 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
14272 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14274 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
14275 aarch64_track_speculation is true.
14276 (tb<optab><mode>1): Likewise.
14277 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
14278 generate CB[N]Z when tracking speculation.
14279 (aarch64_split_compare_and_swap): Likewise.
14280 (aarch64_split_atomic_op): Likewise.
14282 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14284 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
14286 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14288 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
14289 (speculation_barrier): New insn.
14291 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14293 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
14294 * config/arm/arm.md (speculation_barrier): New expand.
14295 (speculation_barrier_insn): New pattern.
14297 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14299 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
14300 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
14301 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
14302 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
14304 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
14305 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
14306 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
14307 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
14308 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
14309 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
14310 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
14311 * builtins.c (expand_speculation_safe_value): New function.
14312 (expand_builtin): Call it.
14313 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
14314 * doc/extend.texi: Document __builtin_speculation_safe_value.
14315 * doc/md.texi: Document "speculation_barrier" pattern.
14316 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
14317 TARGET_HAVE_SPECULATION_SAFE_VALUE.
14318 * doc/tm.texi: Regenerated.
14319 * target.def (have_speculation_safe_value, speculation_safe_value): New
14321 * targhooks.c (default_have_speculation_safe_value): New function.
14322 (default_speculation_safe_value): New function.
14323 * targhooks.h (default_have_speculation_safe_value): Add prototype.
14324 (default_speculation_safe_value): Add prototype.
14326 2018-07-31 David Malcolm <dmalcolm@redhat.com>
14328 * dump-context.h (dump_context::dump_loc): New decl.
14329 * dumpfile.c (dump_context::dump_loc): New member function.
14330 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
14331 and dump_gimple_stmt.
14332 (dump_context::dump_gimple_expr_loc): Likewise, using
14334 (dump_context::dump_generic_expr_loc): Likewise, using
14336 (dump_context::dump_printf_loc_va): Likewise, using
14338 (dump_context::begin_scope): Explicitly using the global function
14339 "dump_loc", rather than the member function.
14341 2018-07-31 Martin Sebor <msebor@redhat.com>
14343 PR tree-optimization/86741
14344 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
14346 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
14348 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
14349 depend on whether prefetch instructions will be emitted or not.
14350 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
14351 will be emitted or not.
14352 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
14353 (TARGET_SETMEM_PFD): New macros.
14355 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14357 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
14358 (NULL_STMT_VEC_INFO): Delete.
14359 (stmt_vec_info::operator*): Likewise.
14360 (stmt_vec_info::operator gimple *): Likewise.
14361 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
14362 of NULL_STMT_VEC_INFO.
14363 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14364 (vect_reassociating_reduction_p): Likewise.
14365 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
14366 (vectorizable_store): Likewise.
14367 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
14368 (vec_info::free_stmt_vec_infos): Likewise.
14370 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14372 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
14373 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
14375 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14377 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
14378 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
14379 (vec_info::free_stmt_vec_info): New private member functions.
14380 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
14381 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
14382 * tree-parloops.c (gather_scalar_reductions): Remove calls to
14383 set_stmt_vec_info_vec and free_stmt_vec_infos.
14384 * tree-vect-loop.c (_loop_vec_info): Remove call to
14385 set_stmt_vec_info_vec.
14386 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
14387 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
14388 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
14389 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
14390 (vec_info::free_stmt_vec_info): ...these new functions. Remove
14391 assignments in {vec_info::,}new_stmt_vec_info that are redundant
14392 with the clearing in the xcalloc.
14393 (stmt_vec_info_vec): Delete.
14394 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
14395 (vectorize_loops): Likewise.
14396 (vec_info::~vec_info): Remove argument from call to
14397 free_stmt_vec_infos.
14398 (vec_info::add_stmt): Remove vinfo argument from call to
14401 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14403 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
14404 rather than a gimple stmt.
14405 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
14406 information for pattern statements when passed the original
14407 statement; instead wait to be passed the pattern statement itself.
14408 Don't call set_vinfo_for_stmt here.
14409 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
14410 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
14411 stmt_vec_infos here.
14412 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
14413 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
14414 stmt_vec_infos entry.
14416 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14418 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
14419 * tree-vectorizer.c (vec_info::replace_stmt): New function.
14420 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
14421 * tree-vect-stmts.c (vectorizable_call): Likewise.
14422 (vectorizable_simd_clone_call): Likewise.
14424 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14426 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
14427 * tree-vectorizer.c (vec_info::remove_stmt): New function.
14428 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
14429 * tree-vect-loop.c (vect_transform_loop): Likewise.
14430 * tree-vect-slp.c (vect_schedule_slp): Likewise.
14431 * tree-vect-stmts.c (vect_remove_stores): Likewise.
14433 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14435 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
14436 (vect_dr_stmt): Delete.
14437 * tree-vectorizer.c (vec_info::lookup_dr): New function.
14438 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
14440 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14441 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
14442 (vect_verify_datarefs_alignment, vect_peeling_supportable)
14443 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
14444 (vect_analyze_data_refs): Likewise.
14445 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
14447 (vect_find_same_alignment_drs): Likewise.
14448 (vect_slp_analyze_node_dependences): Update calls accordingly.
14449 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
14450 instead of DR_VECT_AUX.
14451 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
14452 of a vector data references. Use vec_info::lookup_dr instead of
14454 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
14455 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
14456 instead of DR_VECT_AUX.
14458 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14460 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
14462 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
14464 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14465 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14466 (vect_gen_prolog_loop_niters): Likewise.
14468 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14470 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
14471 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
14472 (vect_known_alignment_in_bytes, vect_dr_behavior)
14473 (vect_get_scalar_dr_size): Take references as dr_vec_infos
14474 instead of data_references. Update calls to other routines for
14475 which the same change has been made.
14476 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
14477 dr_vec_infos instead of stmt_vec_infos.
14478 (vect_analyze_data_ref_dependence): Update call accordingly.
14479 (vect_slp_analyze_data_ref_dependence)
14480 (vect_record_base_alignments): Use DR_VECT_AUX.
14481 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
14482 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
14483 (vector_alignment_reachable_p, vect_get_data_access_cost)
14484 (vect_peeling_supportable, vect_analyze_group_access_1)
14485 (vect_analyze_group_access, vect_analyze_data_ref_access)
14486 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
14487 (vect_compile_time_alias, vect_small_gap_p)
14488 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
14489 (vect_supportable_dr_alignment): Take references as dr_vec_infos
14490 instead of data_references. Update calls to other routines for
14491 which the same change has been made.
14492 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
14493 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
14494 (vect_slp_analyze_and_verify_node_alignment)
14495 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
14496 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
14497 (vect_setup_realignment): Use dr_vec_infos. Update calls after
14499 (_vect_peel_info::dr): Replace with...
14500 (_vect_peel_info::dr_info): ...this new field.
14501 (vect_peeling_hash_get_most_frequent)
14502 (vect_peeling_hash_choose_best_peeling): Update accordingly.
14503 (vect_peeling_hash_get_lowest_cost):
14504 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
14505 routines for which the same change has been made.
14506 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
14508 * tree-vect-loop-manip.c (get_misalign_in_elems)
14509 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
14511 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14512 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
14513 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
14514 (get_group_load_store_type, get_negative_load_store_type)
14515 (vect_get_data_ptr_increment, vectorizable_store)
14516 (vectorizable_load): Likewise.
14517 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
14518 Update calls to other routines for which the same change has been made.
14520 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14522 * tree-vectorizer.h (vec_info::move_dr): New member function.
14523 (dataref_aux): Rename to...
14524 (dr_vec_info): ...this and add "dr" and "stmt" fields.
14525 (_stmt_vec_info::dr_aux): Update accordingly.
14526 (_stmt_vec_info::data_ref_info): Delete.
14527 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
14528 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
14529 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
14530 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
14532 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
14533 (STMT_VINFO_DR_INFO): New macro.
14534 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
14535 (set_dr_misalignment): Update after rename of dataref_aux.
14536 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
14537 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
14538 initialization of STMT_VINFO_DATA_REF.
14539 * tree-vectorizer.c (vec_info::move_dr): New function.
14540 * tree-vect-patterns.c (vect_recog_bool_pattern)
14541 (vect_recog_mask_conversion_pattern)
14542 (vect_recog_gather_scatter_pattern): Use it.
14543 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
14544 the "dr" and "stmt" fields of dr_vec_info instead of
14545 STMT_VINFO_DATA_REF.
14547 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14549 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
14550 (is_pattern_stmt_p): Use it.
14551 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
14552 on pattern statements.
14554 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14556 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
14557 original stmt as a stmt_vec_info rather than a gimple stmt.
14558 (vect_pattern_recog_1): Take the statement directly as a
14559 stmt_vec_info, rather than via a gimple_stmt_iterator.
14560 Update call to vect_mark_pattern_stmts.
14561 (vect_pattern_recog): Update calls accordingly.
14563 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14565 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
14566 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
14567 a vect_def_type for the first argument.
14568 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
14569 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
14570 operand if it isn't defined by a vectorized statement.
14571 (vect_build_gather_load_calls): Remove the mask_dt argument and
14572 update calls to vect_get_vec_def_for_stmt_copy.
14573 (vectorizable_bswap): Likewise the dt argument.
14574 (vectorizable_call): Update calls to vectorizable_bswap and
14575 vect_get_vec_def_for_stmt_copy.
14576 (vectorizable_simd_clone_call, vectorizable_assignment)
14577 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
14578 (vectorizable_comparison): Update calls to
14579 vect_get_vec_def_for_stmt_copy.
14580 (vectorizable_store): Likewise. Remove now-unnecessary calls to
14581 vect_is_simple_use.
14582 (vect_get_loop_based_defs): Remove dt argument and update call
14583 to vect_get_vec_def_for_stmt_copy.
14584 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
14585 and vect_get_vec_def_for_stmt_copy.
14586 (vectorizable_load): Update calls to vect_build_gather_load_calls
14587 and vect_get_vec_def_for_stmt_copy.
14588 * tree-vect-loop.c (vect_create_epilog_for_reduction)
14589 (vectorizable_reduction, vectorizable_live_operation): Update calls
14590 to vect_get_vec_def_for_stmt_copy.
14592 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14594 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
14595 and gimple stmt arguments with a stmt_vec_info.
14596 (vect_record_base_alignments): Update calls accordingly.
14597 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
14598 and gimple stmt arguments with a stmt_vec_info.
14599 (vect_build_slp_tree_1): Remove vinfo argument and update call
14600 to vect_record_max_nunits.
14601 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
14602 and vect_record_max_nunits.
14604 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14606 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
14607 file and take a stmt_vec_info instead of a gimple stmt.
14608 (supportable_widening_operation, vect_finish_replace_stmt)
14609 (vect_finish_stmt_generation, vect_get_store_rhs)
14610 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
14611 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
14612 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
14613 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
14614 (vect_create_data_ref_ptr, bump_vector_ptr)
14615 (vect_permute_store_chain, vect_setup_realignment)
14616 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
14617 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
14618 (vectorizable_reduction, vectorizable_induction)
14619 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
14620 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
14621 than gimple stmts as arguments.
14622 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
14623 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
14624 (can_group_stmts_p, vect_check_gather_scatter)
14625 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
14626 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
14627 (vect_permute_load_chain, vect_shift_permute_load_chain)
14628 (vect_transform_grouped_load)
14629 (vect_record_grouped_load_vectors): Likewise.
14630 * tree-vect-loop.c (vect_fixup_reduc_chain)
14631 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
14632 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
14633 (vectorizable_reduction, vectorizable_induction)
14634 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
14635 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
14636 (vect_get_load_store_mask): Likewise.
14637 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
14638 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
14639 * tree-vect-stmts.c (vect_mark_relevant)
14640 (is_simple_and_all_uses_invariant)
14641 (exist_non_indexing_operands_for_use_p, process_use)
14642 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
14643 (vect_get_vec_def_for_operand, vect_get_vec_defs)
14644 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
14645 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
14646 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
14647 (get_negative_load_store_type, get_load_store_type)
14648 (vect_check_load_store_mask, vect_check_store_rhs)
14649 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
14650 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14651 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14652 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14653 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
14654 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
14655 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
14656 (supportable_widening_operation): Likewise.
14658 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14660 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
14661 a stmt_vec_info instead of a gcall.
14662 (vect_check_gather_scatter): Update call accordingly.
14663 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
14665 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
14666 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
14667 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
14668 instead of a gimple stmt.
14669 (vect_transform_loop): Update calls accordingly.
14670 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
14671 stmt_vec_infos instead of gimple stmts.
14672 (vect_analyze_slp_instance): Update use accordingly.
14673 * tree-vect-stmts.c (read_vector_array, write_vector_array)
14674 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
14675 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
14676 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
14677 (vect_gen_widened_results_half, vect_get_loop_based_defs)
14678 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
14679 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
14680 down to subroutines.
14682 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14684 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
14685 of the worklist from a vector of gimple stmts to a vector of
14687 * tree-vect-stmts.c (vect_mark_relevant, process_use)
14688 (vect_mark_stmts_to_be_vectorized): Likewise
14690 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14692 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
14693 statement before passing it to vect_analyze_stmt.
14694 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
14695 the chain of phi vector definitions. Track the exit phi via its
14697 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
14698 STMT_VINFO_REDUC_DEF.
14699 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
14700 stmt_vec_infos to handle the statement chains.
14701 (vect_get_slp_defs): Record the first statement in the node
14702 using a stmt_vec_info.
14703 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
14704 statements here and pass their stmt_vec_info down to subroutines.
14705 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
14706 down to vect_finish_stmt_generation.
14707 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
14708 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
14709 stmt_vec_infos to subroutines.
14710 (vect_remove_stores): Use stmt_vec_infos to handle the statement
14713 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14715 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
14716 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
14717 (vect_permute_store_chain, vect_setup_realignment)
14718 (vect_permute_load_chain, vect_shift_permute_load_chain)
14719 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
14720 stmts internally, and when passing values to other vectorizer routines.
14721 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14722 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
14723 (vect_analyze_loop_operations, get_initial_def_for_reduction)
14724 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
14725 (vectorizable_reduction, vectorizable_induction)
14726 (vectorizable_live_operation, vect_transform_loop_stmt)
14727 (vect_transform_loop): Likewise.
14728 * tree-vect-patterns.c (vect_reassociating_reduction_p)
14729 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
14730 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
14731 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14732 (vect_slp_analyze_node_operations_1): Likewise.
14733 * tree-vect-stmts.c (vect_mark_relevant, process_use)
14734 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
14735 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
14736 (vect_finish_stmt_generation_1, get_group_load_store_type)
14737 (get_load_store_type, vect_build_gather_load_calls)
14738 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14739 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14740 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14741 (vectorizable_store, vectorizable_load, vectorizable_condition)
14742 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
14743 (supportable_widening_operation): Likewise.
14744 (vect_get_vector_types_for_stmt): Likewise.
14745 * tree-vectorizer.h (vect_dr_behavior): Likewise.
14747 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14749 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14750 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
14751 (vect_permute_store_chain, vect_permute_load_chain)
14752 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
14753 repeated stmt_vec_info lookups.
14754 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14755 (vect_update_ivs_after_vectorizer): Likewise.
14756 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
14757 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14758 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
14759 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14760 (vect_bb_slp_scalar_cost): Likewise.
14761 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
14763 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14765 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
14766 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
14767 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
14768 to get gassigns and gcalls, rather than operating on generc gimple
14770 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
14771 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
14772 (vectorizable_load, vect_analyze_stmt): Likewise.
14773 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
14775 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14777 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
14778 return stmt_vec_infos rather than gimple stmts. Do not accept
14780 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
14782 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
14783 Update use of get_later_stmt.
14784 (vect_get_constant_vectors): Update call accordingly.
14785 (vect_schedule_slp_instance): Likewise
14786 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
14787 (vect_slp_analyze_instance_dependence): Likewise.
14788 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
14790 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14792 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
14793 (stmt_info_for_cost::stmt_info): ...this new field.
14794 (add_stmt_costs): Update accordingly.
14795 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
14796 (vect_get_known_peeling_cost): Likewise.
14797 (vect_estimate_min_profitable_iters): Likewise.
14798 * tree-vect-stmts.c (record_stmt_cost): Likewise.
14800 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14802 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
14803 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14804 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
14806 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
14808 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14810 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
14811 a gimple stmt to a stmt_vec_info.
14812 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
14814 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14816 * tree-vectorizer.h (vec_info::grouped_stores): Change from
14817 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14818 (_loop_vec_info::reduction_chains): Likewise.
14819 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
14821 * tree-vect-slp.c (vect_analyze_slp): Likewise.
14823 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14825 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
14826 a gimple stmt to a stmt_vec_info.
14827 (_stmt_vec_info::next_element): Likewise.
14828 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
14829 (vect_slp_analyze_and_verify_node_alignment)
14830 (vect_analyze_group_access_1, vect_analyze_group_access)
14831 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
14832 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
14833 (vect_supportable_dr_alignment): Update accordingly.
14834 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
14835 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
14836 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
14837 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14838 * tree-vect-slp.c (vect_build_slp_tree_1)
14839 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
14840 (vect_split_slp_store_group, vect_analyze_slp_instance)
14841 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
14842 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
14843 (get_group_load_store_type, get_load_store_type)
14844 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
14845 (vect_transform_stmt, vect_remove_stores): Likewise.
14847 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14849 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
14850 than a gimple stmt.
14851 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14852 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
14853 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
14854 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
14855 (vector_alignment_reachable_p, vect_get_data_access_cost)
14856 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
14857 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
14858 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
14859 (vect_analyze_group_access_1, vect_analyze_group_access)
14860 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
14861 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
14862 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
14863 result of vect_dr_stmt and use the stmt_vec_info instead of
14864 the associated gimple stmt.
14865 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14866 (vect_gen_prolog_loop_niters): Likewise.
14867 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14869 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14871 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
14872 to a vec<stmt_vec_info>.
14873 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
14874 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
14875 vec<stmt_vec_info>.
14876 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
14877 to a vec<stmt_vec_info>.
14878 (bst_traits::value_type, bst_traits::value_type): Likewise.
14879 (bst_traits::hash): Update accordingly.
14880 (vect_get_and_check_slp_defs): Change the stmts parameter from
14881 a vec<gimple *> to a vec<stmt_vec_info>.
14882 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
14883 (vect_build_slp_tree): Likewise.
14884 (vect_build_slp_tree_2): Likewise. Update uses of
14885 SLP_TREE_SCALAR_STMTS.
14886 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
14887 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
14888 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
14889 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
14890 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
14891 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
14892 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
14893 (vect_get_constant_vectors, vect_get_slp_defs)
14894 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
14895 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
14896 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
14897 instead of gimple stmts.
14898 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
14899 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
14900 (vect_slp_analyze_instance_dependence): Update uses of
14901 SLP_TREE_SCALAR_STMTS.
14902 (vect_slp_analyze_and_verify_node_alignment): Likewise.
14903 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
14904 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
14905 (get_initial_defs_for_reduction): Likewise.
14906 (vect_create_epilog_for_reduction): Likewise.
14907 (vectorize_fold_left_reduction): Likewise.
14908 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
14909 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
14910 (can_vectorize_live_stmts): Likewise.
14912 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14914 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
14915 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14916 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
14918 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
14920 (gather_scalar_reductions): Update after above interface changes.
14921 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
14922 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
14924 (vect_force_simple_reduction): Likewise.
14925 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
14926 LOOP_VINFO_REDUCTIONS.
14927 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14929 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14931 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
14932 a gimple stmt to a stmt_vec_info.
14933 * tree-vect-loop.c (vect_active_double_reduction_p)
14934 (vect_force_simple_reduction, vectorizable_reduction): Update
14937 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14939 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
14940 vec<gimple *> to a vec<stmt_vec_info>.
14941 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
14942 the reduction_phis argument from a vec<gimple *> to a
14943 vec<stmt_vec_info>.
14944 (vectorizable_reduction): Likewise the phis local variable that
14945 is passed to vect_create_epilog_for_reduction. Update for new type
14946 of SLP_TREE_VEC_STMTS.
14947 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
14948 (vectorizable_live_operation): Likewise.
14949 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
14950 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
14952 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14954 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
14955 a gimple stmt to a stmt_vec_info.
14956 (vectorizable_condition, vectorizable_live_operation)
14957 (vectorizable_reduction, vectorizable_induction): Pass back the
14958 vectorized statement as a stmt_vec_info.
14959 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
14960 use of STMT_VINFO_VEC_STMT.
14961 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
14962 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
14963 as stmt_vec_infos rather than gimple stmts.
14964 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
14965 to a stmt_vec_info.
14966 (vectorizable_live_operation): Likewise.
14967 (vectorizable_reduction, vectorizable_induction): Likewise,
14968 updating use of STMT_VINFO_VEC_STMT.
14969 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
14970 of STMT_VINFO_VEC_STMT.
14971 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
14972 (vectorizable_simd_clone_call, vectorizable_conversion)
14973 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14974 (vectorizable_store, vectorizable_load, vectorizable_condition)
14975 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
14976 from a gimple stmt to a stmt_vec_info.
14977 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
14978 pointer to a stmt_vec_info to the vectorizable_* routines.
14980 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14982 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
14983 a gimple stmt to a stmt_vec_info.
14984 (is_pattern_stmt_p): Update accordingly.
14985 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
14986 (vect_record_grouped_load_vectors): Likewise.
14987 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
14988 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
14989 (vect_model_reduction_cost): Likewise.
14990 (vect_create_epilog_for_reduction): Likewise.
14991 (vectorizable_reduction, vectorizable_induction): Likewise.
14992 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14993 Return the stmt_vec_info for the pattern statement.
14994 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
14995 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
14996 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
14997 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
14998 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
14999 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
15000 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
15001 (free_stmt_vec_info, vect_is_simple_use): Likewise.
15003 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15005 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
15006 (vect_finish_stmt_generation): Likewise.
15007 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
15008 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
15009 (vect_build_gather_load_calls): Use the return value of the above
15010 functions instead of a separate call to vinfo_for_stmt. Use narrow
15011 scopes for the input gimple stmt and wider scopes for the associated
15012 stmt_vec_info. Use vec_info::lookup_def when setting these
15013 stmt_vec_infos from an SSA_NAME definition.
15014 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
15015 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
15016 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
15017 (vectorizable_store, vectorizable_load, vectorizable_condition)
15018 (vectorizable_comparison): Likewise.
15019 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
15020 (vectorizable_reduction): Likewise.
15022 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15024 * tree-vectorizer.h (vect_is_simple_use): Add an optional
15025 stmt_vec_info * parameter before the optional gimple **.
15026 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
15027 (process_use, vect_get_vec_def_for_operand_1): Update callers.
15028 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
15029 * tree-vect-loop.c (vectorizable_reduction): Likewise.
15030 (vectorizable_live_operation): Likewise.
15031 * tree-vect-patterns.c (type_conversion_p): Likewise.
15032 (vect_look_through_possible_promotion): Likewise.
15033 (vect_recog_rotate_pattern): Likewise.
15034 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
15036 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15038 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
15039 a typedef to a wrapper class.
15040 (NULL_STMT_VEC_INFO): New macro.
15041 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
15042 (stmt_vec_info::operator*): New function.
15043 (stmt_vec_info::operator gimple *): Likewise.
15044 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
15045 (add_stmt_costs): Likewise.
15046 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
15047 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
15048 (vect_get_known_peeling_cost): Likewise.
15049 (vect_estimate_min_profitable_iters): Likewise.
15050 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
15051 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
15052 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
15053 (vectorizable_store, free_stmt_vec_infos): Likewise.
15054 (new_stmt_vec_info): Change return type of xcalloc to
15057 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15059 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
15060 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
15061 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
15062 a single_imm_use-based sequence.
15063 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
15065 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15067 * tree-vectorizer.h (vec_info::lookup_def): Declare.
15068 * tree-vectorizer.c (vec_info::lookup_def): New function.
15069 * tree-vect-patterns.c (vect_get_internal_def): Use it.
15070 (vect_widened_op_tree): Likewise.
15071 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
15072 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
15073 (vectorizable_reduction): Likewise.
15074 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
15076 (vect_is_slp_reduction): Update calls accordingly. Use
15077 vec_info::lookup_def.
15078 (vect_is_simple_reduction): Likewise
15079 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
15081 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15083 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
15084 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
15085 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
15087 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
15088 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
15089 (vect_update_vf_for_slp, vect_analyze_loop_operations)
15090 (vect_is_slp_reduction, vectorizable_induction)
15091 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
15092 * tree-vect-patterns.c (vect_init_pattern_stmt):
15093 (vect_determine_min_output_precision_1, vect_determine_precisions)
15094 (vect_pattern_recog): Likewise.
15095 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
15096 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
15097 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
15098 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
15099 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
15100 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
15101 info field from a loop to a loop_vec_info.
15103 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15105 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
15106 (vec_info::add_stmt): Declare.
15107 * tree-vectorizer.c (vec_info::add_stmt): New function.
15108 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
15109 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
15110 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
15111 (vectorizable_induction): Likewise.
15112 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
15113 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
15114 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
15115 (vectorizable_load): Likewise.
15116 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
15117 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
15118 (vect_recog_gather_scatter_pattern): Likewise.
15119 (append_pattern_def_seq): Likewise. Remove a check that is
15120 performed by add_stmt itself.
15122 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15124 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
15125 which make_ssa_name was called with new_stmt before new_stmt
15128 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15130 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
15132 (vect_is_slp_reduction): ...here...
15133 (vect_is_simple_reduction): ...and here. Remove repetition of tests
15134 that are already known to be false.
15136 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15138 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
15139 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
15140 STMT_VINFO_NUM_SLP_USES when it's true.
15141 (vect_free_slp_instance): Add a final_p parameter and pass it to
15142 vect_free_slp_tree.
15143 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
15144 (vect_analyze_slp_instance): Likewise.
15145 (vect_slp_analyze_operations): Likewise.
15146 (vect_slp_analyze_bb_1): Likewise.
15147 * tree-vectorizer.c (vec_info): Likewise.
15148 * tree-vect-loop.c (vect_transform_loop): Likewise.
15150 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15152 * tree-vect-loop.c (vectorizable_reduction): Assert that the
15153 function is not called for second and subsequent members of
15156 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15158 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
15159 cases for nested loops from here to ...
15160 (vect_create_epilog_for_reduction): ...here. Only call
15161 vect_is_simple_use for inner-loop reductions.
15163 2018-07-31 Martin Liska <mliska@suse.cz>
15165 PR gcov-profile/85338
15166 PR gcov-profile/85350
15167 PR gcov-profile/85372
15168 * profile.c (struct location_triplet): New.
15169 (struct location_triplet_hash): Likewise.
15170 (output_location): Do not output a BB that
15171 is already recorded for a line.
15172 (branch_prob): Use streamed_locations.
15174 2018-07-31 Martin Liska <mliska@suse.cz>
15176 PR gcov-profile/85370
15177 * coverage.c (coverage_begin_function): Do not mark target
15178 clones as artificial functions.
15180 2018-07-31 Martin Liska <mliska@suse.cz>
15182 PR gcov-profile/83813
15183 PR gcov-profile/84758
15184 PR gcov-profile/85217
15185 PR gcov-profile/85332
15186 * profile.c (branch_prob): Do not record GOTO expressions
15187 for GIMPLE statements which locations are already streamed.
15189 2018-07-31 Olivier Hainque <hainque@adacore.com>
15191 * gcc.c (handle_spec_function): Accept a soft_matched_part
15192 argument, as do_spec_1. Pass it down to ...
15193 (eval_spec_function): Accept a soft_matched_part argument,
15194 and pass it down to ...
15195 (do_spec_2): Accept a soft_matched_part argument, and pass
15196 it down to do_spec_1.
15197 (do_spec_1): Pass soft_matched_part to handle_spec_function.
15198 (handle_braces): Update call to handle_spec_function.
15199 (driver::set_up_specs): Update calls to do_spec_2.
15200 (compare_debug_dump_opt_spec_function): Likewise.
15201 (compare_debug_self_opt_spec_function): Likewise.
15203 2018-07-31 Olivier Hainque <hainque@adacore.com>
15205 * common.opt (nolibc): New option.
15206 * doc/invoke.texi (Link Options): Document it.
15207 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
15208 * config/alpha/linux.h: Likewise.
15209 * config/arc/elf.h: Likewise.
15210 * config/arm/uclinux-elf.h: Likewise.
15211 * config/arm/unknown-elf.h: Likewise.
15212 * config/avr/avrlibc.h: Likewise.
15213 * config/bfin/bfin.h: Likewise.
15214 * config/bfin/linux.h: Likewise.
15215 * config/bfin/uclinux.h: Likewise.
15216 * config/darwin.h: Likewise.
15217 * config/darwin10.h: Likewise.
15218 * config/darwin12.h: Likewise.
15219 * config/gnu-user.h: Likewise.
15220 * config/lm32/uclinux-elf.h: Likewise.
15221 * config/pa/pa-hpux11.h: Likewise.
15222 * config/pa/pa64-hpux.h: Likewise.
15223 * config/sparc/sparc.h: Likewise.
15225 2018-07-31 Olivier Hainque <hainque@adacore.com>
15227 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
15228 undefined variables.
15230 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
15233 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
15234 instead of GEN_INT.
15236 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
15238 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
15239 terminated string literal.
15241 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
15243 PR rtl-optimization/85160
15244 * combine.c (is_just_move): New function.
15245 (try_combine): Allow combining two instructions into two if neither of
15246 the original instructions was a move.
15248 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
15251 * doc/extend.texi (Global Register Variables): Discourage use of type
15253 (Local Register Variables): Likewise.
15255 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
15257 PR tree-optimization/86506
15258 * hwint.h (ceil_log2): Resync with hwint.c implementation.
15260 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
15263 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
15264 hard_regno, make sure no insn between `from` and `to` clobbers it.
15266 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
15267 Tom de Vries <tdevries@suse.de>
15269 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
15270 (PTX_DEFAULT_RUNTIME_DIM): ... this.
15271 (nvptx_goacc_validate_dims): Set default worker and gang dims to
15272 PTX_DEFAULT_RUNTIME_DIM.
15273 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
15275 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
15277 * config/pa/pa.c (pa_output_addr_vec): Align address table.
15278 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
15279 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
15281 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
15283 * config/rs6000/constraints.md (wG constraint): Delete, no longer
15285 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
15286 predicate to reflect toc fusion has been deleted.
15287 (toc_fusion_mem_raw): Delete, no longer used.
15288 (toc_fusion_mem_wrapped): Likewise.
15289 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
15291 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
15292 Delete, no longer used.
15293 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
15294 meant to be used for toc fusion.
15295 (rs6000_debug_print_mode): Delete toc fusion debugging.
15296 (rs6000_debug_reg_global): Likewise.
15297 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
15298 fusion and secondary reload support that were never used.
15299 (rs6000_option_override_internal): Delete TOC fusion, that was only
15300 partially defined, and it did not work unless you also used the
15302 (rs6000_legitimate_address_p): Delete TOC fusion support.
15303 (rs6000_opt_masks): Likewise.
15304 (fusion_wrap_memory_address): Delete function, no longer used.
15305 (fusion_split_address); Delete TOC fusion support.
15306 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
15307 longer used with toc fusion being deleted.
15308 (TARGET_TOC_FUSION_FP): Likewise.
15309 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
15311 (toc fusion spliter): Delete TOC fusion support.
15312 (toc_fusionload_<mode>): Likewise.
15313 (toc_fusionload_di): Likewise.
15314 (fusion_gpr_load_<mode>): Delete generator function, this insn no
15315 longer needs to be named. Rename predicate to delete TOC fusion.
15316 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
15317 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
15318 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
15319 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
15320 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
15322 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
15324 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
15325 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
15326 __int128 in built-in function prototypes.
15327 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
15328 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
15330 2018-07-27 Martin Sebor <msebor@redhat.com>
15332 PR tree-optimization/86696
15333 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
15334 types, including enums.
15335 (handle_char_store): Be prepared for the above function to fail.
15337 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
15339 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
15340 when optimization level is lower than 2 or optimize for size.
15342 2018-07-26 Martin Sebor <msebor@redhat.com>
15344 PR tree-optimization/86043
15345 PR tree-optimization/86042
15346 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
15347 (get_string_cst_length): Rename...
15348 (get_min_string_length): ...to this. Add argument.
15349 (handle_char_store): Extend to handle multi-character stores by
15351 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
15352 * tree.h (initializer_zerop): Add argument.
15354 2018-07-26 Jakub Jelinek <jakub@redhat.com>
15356 PR middle-end/86660
15357 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
15358 declare target to variables if they have always,{to,from,tofrom} map
15361 2018-07-26 Martin Liska <mliska@suse.cz>
15364 * lto-wrapper.c: Add linker_output as prefix
15365 for ltrans_output_file.
15367 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
15369 PR rtl-optimization/85805
15370 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
15371 value for hard registers if that was written in the same mode.
15373 2018-07-26 Martin Liska <mliska@suse.cz>
15375 PR gcov-profile/86536
15376 * gcov.c (format_gcov): Use printf format %.*f directly
15377 and do not handle special values.
15379 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15381 * common/config/arc/arc-common.c (arc_option_optimization_table):
15382 Update default optimizations for size.
15384 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15386 * config/arc/arc.md (movsf_insn): Add short instruction selection.
15387 * config/arc/constraints.md (CfZ): New constraint.
15388 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
15389 (subsf3_fpu): Likewise.
15390 (cmpsf_fpu): Likewise.
15391 (cmpsf_fpu_uneq): Likewise.
15393 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15395 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
15397 (arc_is_uncached_mem_p): uncached applies to both the variable and
15400 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15402 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
15405 2018-07-25 David Malcolm <dmalcolm@redhat.com>
15407 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
15408 field "m_scopes" from vec to auto_vec.
15410 2018-07-25 Martin Liska <mliska@suse.cz>
15412 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
15415 2018-07-25 Richard Biener <rguenther@suse.de>
15418 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
15419 special wrt context_die late.
15420 (gen_subprogram_die): Re-use DIEs in local scope.
15422 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
15424 PR tree-optimization/86644
15425 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
15427 2018-07-25 Martin Liska <mliska@suse.cz>
15429 PR middle-end/86645
15430 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
15431 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
15433 2018-07-25 Martin Liska <mliska@suse.cz>
15436 * params.def: Explain ASan abbreviation and provide
15437 a documentation link.
15439 2018-07-24 Martin Sebor <msebor@redhat.com>
15441 PR tree-optimization/86622
15442 PR tree-optimization/86532
15443 * builtins.h (string_length): Declare.
15444 * builtins.c (c_strlen): Correct handling of non-constant offsets.
15445 (check_access): Be prepared for non-constant length ranges.
15446 (string_length): Make extern.
15447 * expr.c (string_constant): Only handle the minor non-constant
15448 array index. Use string_constant to compute the length of
15449 a generic string constant.
15451 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
15453 PR tree-optimization/86618
15454 * tree-vect-stmts.c (vectorizable_call): Don't take the address
15455 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
15457 2018-07-24 David Malcolm <dmalcolm@redhat.com>
15459 PR tree-optimization/86636
15460 * json.cc (json::object::set): Fix comment. Add assertions.
15461 (json::array::append): Move here from json.h. Add comment and an
15463 (json::string::string): Likewise.
15464 * json.h (json::array::append): Move to json.cc.
15465 (json::string::string): Likewise.
15466 * optinfo-emit-json.cc
15467 (optrecord_json_writer::impl_location_to_json): Assert that we
15468 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
15469 wrapper around it. Expand the location once, rather than three
15471 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
15472 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
15474 (optrecord_json_writer::optinfo_to_json): Likewise, in four
15475 places. Fix some overlong lines.
15477 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
15479 * config/aarch64/aarch64-simd.md
15480 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
15481 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
15482 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
15483 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
15484 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
15485 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
15486 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
15487 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
15488 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
15490 2018-07-24 Jakub Jelinek <jakub@redhat.com>
15492 PR middle-end/86627
15493 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
15494 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
15495 and abs_d == d, do the power of two handling if profitable.
15497 2018-07-24 Richard Biener <rguenther@suse.de>
15499 * match.pd: Add BIT_FIELD_REF canonicalizations.
15501 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
15504 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
15506 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
15508 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
15509 terminated STRING_CST object.
15511 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
15513 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
15515 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
15517 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
15518 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
15519 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
15520 the elements into a register.
15521 (rs6000_split_v4si_init_di_reg): Delete.
15522 (rs6000_split_v4si_init): Delete.
15523 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
15524 (vsx_init_v4si): Rewrite as a define_expand.
15526 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
15528 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
15529 zero_extend argument from memory): New.
15531 2018-07-22 Martin Sebor <msebor@redhat.com>
15534 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
15535 diagnosing calls with unknown arguments unless -Walloca-larger-than
15536 is restricted to less than PTRDIFF_MAX bytes.
15538 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
15540 * doc/gcov.texi (Invoking Gcov): Editorial changes.
15542 2018-07-20 David Malcolm <dmalcolm@redhat.com>
15544 * pretty-print.c (text_info::set_location): Remove redundant
15545 "line_table" parameter from call to rich_location::set_range.
15547 2018-07-20 Martin Sebor <msebor@redhat.com>
15549 PR middle-end/82063
15550 * builtins.c (expand_builtin_alloca): Adjust.
15551 * calls.c (alloc_max_size): Simplify.
15552 * cgraphunit.c (cgraph_node::expand): Adjust.
15553 * common.opt (larger_than_size, warn_frame_larger_than): Remove
15555 (frame_larger_than_size): Same.
15556 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
15557 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
15559 * doc/invoke.texi (GCC Command Options): Document option arguments.
15560 Explain byte-size arguments and suffixes.
15561 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
15562 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
15563 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
15564 * doc/options.texi (UInteger): Expand.
15565 (Host_Wide_Int, ByteSize): Document new properties.
15566 * final.c (final_start_function_1): Include sizes in an error message.
15567 * function.c (frame_offset_overflow): Same.
15568 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
15569 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
15570 Diagnose unbounded alloca calls only for limits of less than
15572 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
15573 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
15575 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
15576 only for limits of less than PTRDIFF_MAX.
15577 * langhooks-def.h (lhd_handle_option): Change function argument
15579 * langhooks.c (lhd_handle_option): Same.
15580 * langhooks.h (handle_option): Same.
15581 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
15583 (var_type, var_type_struct): Same.
15584 (var_set): Handle ByteSize flag.
15585 * optc-gen.awk: Add comments to output to ease debugging. Make
15586 use of HOST_WIDE_INT where appropriate.
15587 * opts-gen-save.awk: Use %lx to format unsigned long.
15588 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
15589 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
15590 arguments. Parse bytes-size suffixes.
15591 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
15592 (enum_value_to_arg): Same.
15593 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
15594 (handle_option): Adjust.
15595 (generate_option): Change function argument to HOST_WIDE_INT.
15596 (cmdline_handle_error): Adjust.
15597 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
15598 (set_option): Change function argument to HOST_WIDE_INT.
15599 (option_enabled): Handle cl_host_wide_int.
15600 (get_option_state): Handle CLVC_SIZE.
15601 (control_warning_option): Same.
15602 * opts.c (common_handle_option): Change function argument to
15603 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
15604 OPT_Wvla_larger_than_.
15605 * opts.h (enum cl_var_type): Add an enumerator.
15606 * stor-layout.c (layout_decl): Print a more meaningful warning.
15607 * toplev.c (output_stack_usage): Adjust.
15609 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
15611 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
15612 call to inline_expand_builtin_string_cmp.
15613 (expand_builtin_strcmp): Likewise.
15614 (expand_builtin_strncmp): Likewise.
15615 (inline_string_cmp): Delete the last parameter, change char_type_node
15616 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
15618 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
15619 the inlining expansion on target where the type of the call has same or
15620 narrower precision than unsigned char.
15622 2018-07-20 David Malcolm <dmalcolm@redhat.com>
15624 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
15625 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
15626 * common.opt (fsave-optimization-record): New option.
15627 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
15628 * doc/invoke.texi (-fsave-optimization-record): New option.
15629 * dumpfile.c: Include "optinfo-emit-json.h".
15630 (struct kv_pair): Move to coretypes.h.
15631 (optgroup_options): Make non-static.
15632 (dump_context::end_scope): Call
15633 optimization_records_maybe_pop_dump_scope.
15634 * dumpfile.h (optgroup_options): New decl.
15635 * json.cc: New file.
15636 * json.h: New file.
15637 * optinfo-emit-json.cc: New file.
15638 * optinfo-emit-json.h: New file.
15639 * optinfo.cc: Include "optinfo-emit-json.h".
15640 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
15641 (optinfo_enabled_p): Check optimization_records_enabled_p.
15642 (optinfo_wants_inlining_info_p): Likewise.
15643 * optinfo.h: Update comment.
15644 * profile-count.c (profile_quality_as_string): New function.
15645 * profile-count.h (profile_quality_as_string): New decl.
15646 (profile_count::quality): New accessor.
15647 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
15648 and optinfo_emit_json_cc_tests.
15649 * selftest.h (selftest::json_cc_tests): New decl.
15650 (selftest::optinfo_emit_json_cc_tests): New decl.
15651 * toplev.c: Include "optinfo-emit-json.h".
15652 (compile_file): Call optimization_records_finish.
15653 (do_compile): Call optimization_records_start.
15654 * tree-ssa-live.c: Include optinfo.h.
15655 (remove_unused_scope_block_p): Retain inlining information if
15656 optinfo_wants_inlining_info_p returns true.
15658 2018-07-20 Richard Biener <rguenther@suse.de>
15661 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
15662 to cover -flto-partition=none.
15664 2018-07-20 Martin Liska <mliska@suse.cz>
15666 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
15667 (get_decl_source_range): Remove unused function.
15669 2018-07-20 Richard Biener <rguenther@suse.de>
15671 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
15672 (struct vn_phi_s): Likewise.
15673 (struct vn_reference_s): Likewise.
15674 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
15675 for searching the slot of an entry known to be in the hash itself.
15676 (vn_phi_hasher::equal): Likewise.
15677 (vn_reference_hasher::equal): Likewise.
15678 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
15680 (optimistic_info, current_info): Remove, keeping only valid_info.
15681 (vn_reference_lookup_1): Remove fallback lookup.
15682 (vn_reference_lookup_2): Likewise.
15683 (vn_nary_op_lookup_1): Likewise.
15684 (vn_phi_lookup): Likewise.
15685 (vn_nary_build_or_lookup_1): Make sure to not chain the built
15687 (vn_reference_insert): Adjust, chain the inserted hash element
15688 at last_inserted_ref.
15689 (vn_reference_insert_pieces): Likewise.
15690 (visit_reference_op_call): Likewise.
15691 (vn_nary_op_insert_into): Chain the inserted hash element at
15692 last_inserted_nary.
15693 (vn_nary_op_insert_pieces): Adjust.
15694 (vn_nary_op_insert): Likewise.
15695 (vn_nary_op_insert_stmt): Likewise.
15696 (vn_phi_insert): Adjust, chain the inserted hash element at
15698 (process_scc): Remove clearing and copying the optimistic
15699 table. Instead remove elements inserted during an optimistic
15700 iteration from the single table we maintain.
15701 (init_scc_vn): Adjust.
15702 (free_scc_vn): Likewise.
15703 (sccvn_dom_walker::record_cond): Likewise.
15704 (sccvn_dom_walker::after_dom_children): Likewise.
15706 2018-07-19 Martin Sebor <msebor@redhat.com>
15708 PR tree-optimization/84047
15709 PR tree-optimization/83776
15710 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
15711 (check_array_bounds): Call it.
15713 2018-07-19 Martin Sebor <msebor@redhat.com>
15715 * align.h (align_flags): Use member initialization.
15717 2018-07-19 David Malcolm <dmalcolm@redhat.com>
15719 * Makefile.in (OBJS): Add optinfo.o.
15720 * coretypes.h (class symtab_node): New forward decl.
15721 (struct cgraph_node): New forward decl.
15722 (class varpool_node): New forward decl.
15723 * dump-context.h: New file.
15724 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
15726 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
15727 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
15728 (set_alt_dump_file): Likewise.
15729 (dump_context::~dump_context): New dtor.
15730 (dump_gimple_stmt): Move implementation to...
15731 (dump_context::dump_gimple_stmt): ...this new member function.
15732 Add the stmt to any pending optinfo, creating one if need be.
15733 (dump_gimple_stmt_loc): Move implementation to...
15734 (dump_context::dump_gimple_stmt_loc): ...this new member function.
15735 Start a new optinfo and add the stmt to it.
15736 (dump_gimple_expr): Move implementation to...
15737 (dump_context::dump_gimple_expr): ...this new member function.
15738 Add the stmt to any pending optinfo, creating one if need be.
15739 (dump_gimple_expr_loc): Move implementation to...
15740 (dump_context::dump_gimple_expr_loc): ...this new member function.
15741 Start a new optinfo and add the stmt to it.
15742 (dump_generic_expr): Move implementation to...
15743 (dump_context::dump_generic_expr): ...this new member function.
15744 Add the tree to any pending optinfo, creating one if need be.
15745 (dump_generic_expr_loc): Move implementation to...
15746 (dump_context::dump_generic_expr_loc): ...this new member
15747 function. Add the tree to any pending optinfo, creating one if
15749 (dump_printf): Move implementation to...
15750 (dump_context::dump_printf_va): ...this new member function. Add
15751 the text to any pending optinfo, creating one if need be.
15752 (dump_printf_loc): Move implementation to...
15753 (dump_context::dump_printf_loc_va): ...this new member function.
15754 Start a new optinfo and add the stmt to it.
15755 (dump_dec): Move implementation to...
15756 (dump_context::dump_dec): ...this new member function. Add the
15757 value to any pending optinfo, creating one if need be.
15758 (dump_context::dump_symtab_node): New member function.
15759 (dump_context::get_scope_depth): New member function.
15760 (dump_context::begin_scope): New member function.
15761 (dump_context::end_scope): New member function.
15762 (dump_context::ensure_pending_optinfo): New member function.
15763 (dump_context::begin_next_optinfo): New member function.
15764 (dump_context::end_any_optinfo): New member function.
15765 (dump_context::s_current): New global.
15766 (dump_context::s_default): New global.
15767 (dump_scope_depth): Delete global.
15768 (dumpfile_ensure_any_optinfo_are_flushed): New function.
15769 (dump_symtab_node): New function.
15770 (get_dump_scope_depth): Reimplement in terms of dump_context.
15771 (dump_begin_scope): Likewise.
15772 (dump_end_scope): Likewise.
15773 (selftest::temp_dump_context::temp_dump_context): New ctor.
15774 (selftest::temp_dump_context::~temp_dump_context): New dtor.
15775 (selftest::verify_item): New function.
15776 (ASSERT_IS_TEXT): New macro.
15777 (ASSERT_IS_TREE): New macro.
15778 (ASSERT_IS_GIMPLE): New macro.
15779 (selftest::test_capture_of_dump_calls): New test.
15780 (selftest::dumpfile_c_tests): Call it.
15781 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
15782 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
15783 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
15784 descriptive comment.
15785 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
15786 (dump_node, dump_bb): Move these unrelated decls.
15787 (class dump_manager): Add leading comment.
15788 * optinfo.cc: New file.
15789 * optinfo.h: New file.
15791 2018-07-19 Michael Collison <michael.collison@arm.com>
15792 Richard Henderson <rth@redhat.com>
15794 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
15795 (subti): Handle op1 zero.
15796 (subvti4, usub4ti4): New.
15797 (*sub<GPI>3_compare1_imm): New.
15798 (sub<GPI>3_carryinCV): New.
15799 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
15800 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
15802 2018-07-19 Michael Collison <michael.collison@arm.com>
15803 Richard Henderson <rth@redhat.com>
15805 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
15806 (addti3): Create simpler code if low part is already known to be 0.
15807 (addvti4, uaddvti4): New.
15808 (*add<GPI>3_compareC_cconly_imm): New.
15809 (*add<GPI>3_compareC_cconly): New.
15810 (*add<GPI>3_compareC_imm): New.
15811 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
15812 handle constants within this pattern..
15813 (*add<GPI>3_compareV_cconly_imm): New.
15814 (*add<GPI>3_compareV_cconly): New.
15815 (*add<GPI>3_compareV_imm): New.
15816 (add<GPI>3_compareV): New.
15817 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
15818 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
15819 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
15820 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
15821 with 'comparison' operator.
15822 (*add<GPI>3_compareV_cconly_imm): Ditto.
15823 (*add<GPI>3_compareV_cconly): Ditto.
15824 (*add<GPI>3_compareV_imm): Ditto.
15825 (add<GPI>3_compareV): Ditto.
15826 (add<mode>3_carryinC): Ditto.
15827 (*add<mode>3_carryinC_zero): Ditto.
15828 (*add<mode>3_carryinC): Ditto.
15829 (add<mode>3_carryinV): Ditto.
15830 (*add<mode>3_carryinV_zero): Ditto.
15831 (*add<mode>3_carryinV): Ditto.
15833 2018-07-19 Michael Collison <michael.collison@arm.com>
15834 Richard Henderson <rth@redhat.com>
15836 * config/aarch64/aarch64-modes.def (CC_V): New.
15837 * config/aarch64/aarch64-protos.h
15838 (aarch64_addti_scratch_regs): Declare
15839 (aarch64_subvti_scratch_regs): Declare.
15840 (aarch64_expand_subvti): Declare.
15841 (aarch64_gen_unlikely_cbranch): Declare
15842 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
15843 for signed overflow using CC_Vmode.
15844 (aarch64_get_condition_code_1): Handle CC_Vmode.
15845 (aarch64_gen_unlikely_cbranch): New function.
15846 (aarch64_addti_scratch_regs): New function.
15847 (aarch64_subvti_scratch_regs): New function.
15848 (aarch64_expand_subvti): New function.
15850 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15852 * config/aarch64/aarch64-option-extensions.def: New entry for profile
15854 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
15855 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
15858 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15861 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15862 address check not strict.
15864 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15866 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
15868 (store_pair_lanes<mode>): Likewise.
15869 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
15870 enum value 'ADDR_QUERY_LDP_STP_N'.
15871 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
15872 (aarch64_print_address_internal): Add declaration.
15873 (aarch64_print_ldpstp_address): Remove.
15874 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
15875 (aarch64_print_operand): Change printing of 'y'.
15876 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
15877 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
15878 'true' rather than '1'.
15879 * config/aarch64/constraints.md (Uml): Likewise.
15880 (Uml): Rename to Umn.
15883 2018-07-19 Richard Biener <rguenther@suse.de>
15885 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
15887 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
15888 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
15889 (vn_reference_hasher): Likewise.
15890 (struct vn_tables_s): Remove obstack and alloc-pool members.
15891 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
15892 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
15893 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
15894 (vn_reference_insert_pieces): Likewise.
15895 (alloc_vn_nary_op_noinit): Adjust.
15896 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
15897 (vn_phi_eq): Adjust.
15898 (shared_lookup_phiargs): Remove.
15899 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
15900 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
15901 (visit_reference_op_call): Likewise.
15902 (copy_nary, copy_phi, copy_reference): Remove.
15903 (process_scc): Rewind the obstack when iterating. Do not
15904 copy the elements to valid_info but just move them from one
15905 hashtable to the other.
15906 (allocate_vn_table): Adjust.
15907 (free_vn_table): Likewise.
15908 (init_scc_vn): Likewise.
15909 (free_scc_vn): Likewise.
15911 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
15914 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
15915 indirect_return as function type attribute.
15916 (ix86_attribute_table): Change indirect_return to function
15918 * doc/extend.texi: Update indirect_return attribute.
15920 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
15922 * wide-int.h (widest2_int): New.
15923 * gimple-fold.c (arith_overflowed_p): Use it.
15924 * tree.h (widest2_int_cst): New.
15925 * tree-vrp.c (wide_int_binop_overflow): Rename from
15926 vrp_int_const_binop.
15927 Rewrite to work on trees.
15928 (extract_range_from_multiplicative_op_1): Abstract code to...
15929 (wide_int_range_min_max): ...here.
15930 (wide_int_range_cross_product): ...and here.
15931 (extract_range_from_binary_expr_1): Abstract overflow code to...
15932 (wide_int_range_mult_wrapping): ...here.
15933 * tree-vrp.h (wide_int_range_cross_product): New.
15934 (wide_int_range_mult_wrapping): New.
15936 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
15937 Julia Koval <julia.koval@intel.com>
15939 * config/i386/x86-tune-costs.h (skylake_memcpy,
15940 skylake_memset): Replace rep_prefix with unrolling for size 512.
15942 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
15944 PR middle-end/86544
15945 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
15946 comparision with EQ_EXPR in last stmt.
15948 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
15950 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
15951 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
15952 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
15953 previously known as "PowerPC AltiVec Built-in Functions". Move
15954 some material to new subsubsections "PowerPC AltiVec Built-in
15955 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
15957 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
15958 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
15959 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
15960 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
15962 2018-07-18 Richard Biener <rguenther@suse.de>
15964 PR tree-optimization/86557
15965 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
15968 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
15970 * config/s390/s390.c (s390_function_profiler): Generate CFI.
15972 2018-07-17 Jeff Law <law@redhat.com>
15974 * config/arm/arm.c (get_label_padding): Update for recent
15975 changes to label_to_alignment.
15977 PR tree-optimization/86010
15978 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
15980 * config/mips/mips.c (vr4130_align_insns): Update for recent
15981 changes to label_to_alignment.
15983 * config/frv/frv.c (frv_label_align): Update for recent changes
15984 to label_to_alignment.
15986 * config/nios2/nios2.c (nios2_label_align): Update for recent
15987 changes which dropped ALIGN_LABELS_LOG.
15989 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
15991 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
15992 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
15994 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
15996 * config/arc/arc.c (arc_label_align): Use align_labels instead of
15997 deprecated align_labels_log.
15999 2018-07-17 Richard Biener <rguenther@suse.de>
16002 * dwarf2out.c (init_sections_and_labels): Always generate
16003 a debug_line_str_section for early LTO debug.
16004 (dwarf2out_finish): Reset debug_line_str_hash output early.
16005 Bump counter for extra dwarf5 .debug_loc labels to not conflict
16007 (dwarf2out_early_finish): Output debug_line_str.
16009 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
16011 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
16012 index register on z196 or later.
16014 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
16016 * config/s390/s390.c (s390_default_align): Set default function
16018 (s390_override_options_after_change): Call s390_default align.
16019 (s390_option_override_internal): Call s390_default_align.
16020 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
16022 2018-07-17 Jakub Jelinek <jakub@redhat.com>
16024 PR middle-end/86542
16025 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
16026 to _looptemp_ clauses, other than the first two.
16028 2018-07-17 Martin Liska <mliska@suse.cz>
16030 * opts.c: Do not enable OPT_falign_* for -Os.
16032 2018-07-17 Martin Liska <mliska@suse.cz>
16034 * align.h (MAX_CODE_ALIGN): New.
16035 (MAX_CODE_ALIGN_VALUE): New.
16036 * common/config/i386/i386-common.c (ix86_handle_option):
16037 (MAX_CODE_ALIGN): Moved to align.h.
16038 * final.c (MAX_CODE_ALIGN): Likewise.
16039 * opts.c (parse_and_check_align_values):
16040 (MAX_CODE_ALIGN): Likewise.
16041 (MAX_CODE_ALIGN_VALUE): Likewise.
16043 2018-07-17 Martin Liska <mliska@suse.cz>
16045 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
16046 in order to fulfil coding style.
16047 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
16048 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16049 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
16050 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
16051 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
16052 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
16053 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
16054 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
16055 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
16057 2018-07-17 Martin Liska <mliska@suse.cz>
16059 * align.h: New file.
16060 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
16062 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
16063 align_flags of label_to_alignment.
16064 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
16066 * config/m68k/m68k.c: Do not use removed align_labels_value and
16068 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
16069 (LOOP_ALIGN): Likewise.
16070 (LABEL_ALIGN): Likewise.
16071 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
16072 Remove not used macro.
16073 (rs6000_loop_align): Change return type to align_flags.
16074 (rs6000_loop_align_max_skip): Remove.
16075 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
16076 Change return type to align_flags.
16077 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
16078 Remove not used macro.
16079 (rs6000_loop_align): Change return type to align_flags.
16080 (rs6000_loop_align_max_skip): Remove.
16081 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
16082 * config/rx/rx-protos.h (rx_align_for_label): Make it
16084 * config/rx/rx.c (rx_align_for_label): Change return type
16086 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
16088 into align_flags class.
16089 (LABEL_ALIGN): Likewise.
16090 (LOOP_ALIGN): Likewise.
16091 * config/s390/s390.c (s390_label_align): Use align_flags
16093 (s390_asm_output_function_label): Likewise.
16094 * config/sh/sh.c (sh_override_options_after_change):
16095 Use align_flags class directly without macros.
16096 (find_barrier): Likewise.
16097 (barrier_align): Likewise.
16098 (sh_loop_align): Likewise.
16099 * config/spu/spu.c (spu_option_override):
16100 Use align_flags_tuple::get_value instead of removed macros.
16101 (spu_sched_init): Likewise.
16102 * config/spu/spu.h (GTY): Likewise.
16103 * config/visium/visium.c (visium_option_override):
16104 Set "8" as default secondary alignment.
16105 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
16106 in order to guarantee secondary alignment of 8.
16107 * coretypes.h: Include align.h header file.
16108 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
16109 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
16110 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
16111 * doc/tm.texi.in: Likewise.
16112 * final.c (struct label_alignment): Remove not used structure.
16113 (LABEL_ALIGN): Change type to align_flags.
16114 (LOOP_ALIGN): Likewise.
16115 (JUMP_ALIGN): Likewise.
16116 (default_loop_align_max_skip): Remove.
16117 (default_label_align_max_skip): Likewise.
16118 (default_jump_align_max_skip): Likewise.
16119 (default_label_align_after_barrier_max_skip):
16120 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
16121 (LABEL_TO_MAX_SKIP): Remove.
16122 (label_to_alignment): Return align_flags type instead of integer.
16123 (label_to_max_skip): Remove.
16124 (align_fuzz): Use align_flags type.
16125 (compute_alignments): Use align_flags type and use align_flags::max
16126 to combine multiple alignments.
16127 (grow_label_align): Grow vec instead of C array.
16128 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
16129 (shorten_branches): Use align_flags type and use align_flags::max
16130 to combine multiple alignments.
16131 (final_scan_insn_1): Remove usage of secondary alignment that comes
16132 from label alignment, but instead use proper secondary alignment
16133 which is computed in grow_label_align.
16134 * flags.h (struct align_flags_tuple): Move to align.h.
16135 (struct align_flags): Likewise.
16136 (state_align_loops): Rename to align_loops.
16137 (state_align_jumps): Rename to align_jumps.
16138 (state_align_labels): Rename to align_labels.
16139 (state_align_functions): Rename to align_functions.
16140 (align_loops_log): Remove.
16141 (align_jumps_log): Remove.
16142 (align_labels_log): Remove.
16143 (align_functions_log): Remove.
16144 (align_loops_max_skip): Remove.
16145 (align_jumps_max_skip): Remove.
16146 (align_labels_max_skip): Remove.
16147 (align_functions_max_skip): Remove.
16148 (align_loops_value): Remove.
16149 (align_jumps_value): Remove.
16150 (align_labels_value): Remove.
16151 (align_functions_value): Remove.
16152 * output.h (label_to_alignment): Change return type to align_flags.
16153 (label_to_max_skip): Remove.
16154 * target.def: Remove loop_align_max_skip, label_align_max_skip,
16155 jump_align_max_skip macros.
16156 * targhooks.h (default_loop_align_max_skip): Remove.
16157 (default_label_align_max_skip): Likewise.
16158 (default_jump_align_max_skip): Likewise.
16159 (default_label_align_after_barrier_max_skip): Remove.
16160 * toplev.c (read_log_maxskip): Use ::normalize function.
16161 (parse_N_M): Remove not used argument and also call ::normalize.
16162 (parse_alignment_opts): Do not pass unused arguments.
16163 * varasm.c (assemble_start_function): Use directly align_functions
16164 instead of removed macros.
16165 * system.h: Do not poison removed macros.
16167 2018-07-17 Jakub Jelinek <jakub@redhat.com>
16169 PR middle-end/86539
16170 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
16171 and cond temporaries don't have reference type if iterator has
16172 pointer type. For init use &for_pre_body instead of pre_p if
16173 for_pre_body is non-empty.
16175 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
16177 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
16178 double-double modes to SFmode directly directly.
16179 (trunc<mode>sf2_fprs): Delete.
16181 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
16183 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
16184 for conversions between IFmode and the decimal floating point modes.
16185 (init_float128_ieee): Use the correct names for conversions between
16186 KFmode and the decimal floating point modes.
16188 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
16190 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
16191 for the conversions between TDmode and IFmode.
16192 (init_float128_ieee): Use more correct names for the conversions
16193 between TDmode and KFmode.
16195 2018-07-16 Jakub Jelinek <jakub@redhat.com>
16197 PR tree-optimization/86526
16198 * builtins.c (expand_builtin_memcmp): Formatting fixes.
16199 (inline_expand_builtin_string_cmp): Likewise.
16200 (inline_string_cmp): Likewise. Use c_readstr instead of
16201 builtin_memcpy_read_str. Add unit_mode temporary.
16203 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
16205 PR middle-end/86528
16206 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
16207 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
16209 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16211 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
16212 Alphabetize prototypes of built-in functions, separating out
16213 built-in functions that are listed in this section but should be
16214 described elsewhere.
16216 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
16219 * expmed.c (emit_store_flag): Do not emit setcc followed by a
16220 conditional move when trapping comparison was split to a
16221 non-trapping one (and vice versa).
16223 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
16225 * config/s390/s390.c (s390_function_profiler): Generate nops
16226 instead of profiler call sequences.
16227 * config/s390/s390.opt: Add the new option.
16229 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
16231 * config/s390/s390.c (s390_function_profiler): Generate
16232 __mcount_loc section.
16233 * config/s390/s390.opt: Add the new option.
16235 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
16237 * common.opt: Add the new warning.
16238 * config/s390/s390.c (s390_function_profiler): Emit "brasl
16239 %r0,__fentry__" when -mfentry is specified.
16240 (s390_option_override_internal): Disallow -mfentry for 31-bit
16242 * config/s390/s390.opt: Add the new option.
16244 2018-07-16 Richard Biener <rguenther@suse.de>
16247 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
16248 for function-local FUNCTION_DECL and RESULT_DECL immediately.
16250 2018-07-16 Martin Liska <mliska@suse.cz>
16253 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
16256 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
16258 * config/arc/arcHS.md: Update ARCHS scheduling rules.
16260 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
16262 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
16264 * config/arc/arc-cpus.def (hs4x): New cpu.
16266 * config/arc/arc-tables.opt: Regenerate.
16267 * config/arc/arc.c (arc_sched_issue_rate): New function.
16268 (TARGET_SCHED_ISSUE_RATE): Define.
16269 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
16270 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
16272 (attr tune): Add ARCHS4x tune values.
16273 (attr tune_dspmpy): Define.
16274 (*tst): Correct instruction type.
16275 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
16276 * config/arc/arcHS4x.md: New file.
16277 * config/arc/fpu.md: Update instruction type attributes.
16278 * config/arc/t-multilib: Regenerate.
16280 2018-07-16 Tom de Vries <tdevries@suse.de>
16283 * var-tracking.c (vt_initialize): Fix pre_dec handling.
16285 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16287 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
16290 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
16292 PR tree-optimization/86514
16293 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
16294 conversion to a boolean type from a type with greater precision.
16296 2018-07-16 Tom de Vries <tdevries@suse.de>
16298 * var-tracking.c (vt_initialize): Print adjusted insn slim if
16299 dump_flags request TDF_SLIM.
16301 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
16303 * fold-const.c (int_const_binop_1): Abstract...
16304 (wide_int_binop): ...wide int code here.
16305 (poly_int_binop): ...poly int code here.
16306 Abstract the rest of int_const_binop_1 into int_const_binop.
16307 * fold-const.h (wide_int_binop): New.
16308 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
16309 Remove useless PLUS/MINUS_EXPR case.
16310 (zero_nonzero_bits_from_vr): Move wide int code...
16311 (zero_nonzero_bits_from_bounds): ...here.
16312 (extract_range_from_binary_expr_1): Move mask optimization code...
16313 (range_easy_mask_min_max): ...here.
16314 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
16315 (range_easy_mask_min_max): New.
16317 2018-07-15 Jeff Law <law@redhat.com>
16320 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
16323 2018-07-14 Jim Wilson <jimw@sifive.com>
16325 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
16327 2018-07-14 Paul Koning <ni1d@arrl.net>
16329 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
16331 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
16333 * lto-streamer-out.c (copy_function_or_variable): Dump info about
16336 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
16337 Steve Munroe <munroesj52@gmail.com>
16339 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
16340 (_mm_andnot_si128): Likewise.
16341 (_mm_or_si128): Likewise.
16342 (_mm_xor_si128): Likewise.
16344 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
16346 PR middle-end/78809
16347 * builtins.c (expand_builtin_memcmp): Inline the calls first
16348 when result_eq is false.
16349 (expand_builtin_strcmp): Inline the calls first.
16350 (expand_builtin_strncmp): Likewise.
16351 (inline_string_cmp): New routine. Expand a string compare
16352 call by using a sequence of char comparison.
16353 (inline_expand_builtin_string_cmp): New routine. Inline expansion
16354 a call to str(n)cmp/memcmp.
16355 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
16357 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
16359 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
16361 * config/arm/driver-arm.c: Include arm-native.h.
16362 (host_detect_local_cpu): Use auto-generated data tables.
16363 (vendors, arm_cpu_table): Delete. Move part information to ...
16364 * config/arm/arm-cpus.in: ... here.
16365 * config/arm/parsecpu.awk (gen_native): New function.
16366 (vendor, part): New CPU fields.
16367 (END): Add support for building the native CPU detection tables.
16368 * config/arm/t-arm (arm-native.h): Add build rule.
16369 (driver-arm.o): Add dependency on arm-native.h.
16371 2018-07-13 Richard Biener <rguenther@suse.de>
16373 PR middle-end/85974
16374 * match.pd (addr1 - addr2): Allow either of the operand to
16377 2018-07-13 Tom de Vries <tdevries@suse.de>
16379 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
16382 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
16384 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
16385 arrays instead of numbered variables.
16387 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
16389 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
16390 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
16392 2018-07-13 Richard Biener <rguenther@suse.de>
16395 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
16396 instead of get_context_die.
16398 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
16399 Richard Biener <rguenther@suse.de>
16401 PR middle-end/86489
16402 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
16403 that the loop latch destination where phi is defined.
16405 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
16407 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
16408 (riscv_expand_epilogue): Add assertion to check interrupt mode.
16409 (riscv_set_current_function): Extract getting interrupt type to new
16411 (riscv_get_interrupt_type): New function.
16412 (riscv_merge_decl_attributes): New function, checking interrupt type is
16414 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
16416 2018-07-12 Paul Koning <ni1d@arrl.net>
16418 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
16421 2018-07-12 Paul Koning <ni1d@arrl.net>
16423 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
16424 zero reference, add doloop_end instead.
16425 * doc/md.texi (decrement_and_branch_until_zero): Remove.
16426 (Looping patterns): Remove decrement_and_branch_until_zero. Add
16427 detail for doloop_end.
16429 2018-07-12 Martin Sebor <msebor@redhat.com>
16432 * attribs.c (decl_attributes): Reject conflicting attributes before
16433 calling attribute handlers.
16435 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
16437 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
16439 (gcc::dump_manager::get_dump_file_name): likewise.
16440 (dump_begin): Likewise.
16441 * dumpfile.h (dump_begin): Update prototype.
16442 (gcc::dump_manager::get_dump_file_name,
16443 gcc::dump_manager::get_dump_file_name): Update prototype.
16445 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16447 * internal-fn.h (vectorizable_internal_fn_p): New function.
16448 * tree-vect-slp.c (compatible_calls_p): Likewise.
16449 (vect_build_slp_tree_1): Remove nops argument. Handle calls
16450 to internal functions.
16451 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
16453 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16455 * fold-const.h (inverse_conditions_p): Declare.
16456 * fold-const.c (inverse_conditions_p): New function.
16457 * match.pd: Use inverse_conditions_p. Add folds of view_converts
16458 that test the inverse condition of a conditional internal function.
16459 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
16460 * internal-fn.c (internal_fn_mask_index): Handle conditional
16461 internal functions.
16462 (vectorized_internal_fn_supported_p): New function.
16463 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
16464 (any_pred_load_store): Replace with...
16465 (need_to_predicate): ...this new variable.
16466 (redundant_ssa_names): New variable.
16467 (ifcvt_can_use_mask_load_store): Move initial checks to...
16468 (ifcvt_can_predicate): ...this new function. Handle tree codes
16469 for which a conditional internal function exists.
16470 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
16471 instead of ifcvt_can_use_mask_load_store. Update after variable
16473 (predicate_load_or_store): New function, split out from
16474 predicate_mem_writes.
16475 (check_redundant_cond_expr): New function.
16476 (value_available_p): Likewise.
16477 (predicate_rhs_code): Likewise.
16478 (predicate_mem_writes): Rename to...
16479 (predicate_statements): ...this. Use predicate_load_or_store
16480 and predicate_rhs_code.
16481 (combine_blocks, tree_if_conversion): Update after above name changes.
16482 (ifcvt_local_dce): Handle redundant_ssa_names.
16483 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
16484 general conditional functions.
16485 * tree-vect-stmts.c (vectorizable_call): Likewise.
16487 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16488 Alan Hayward <alan.hayward@arm.com>
16489 David Sherwood <david.sherwood@arm.com>
16491 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
16492 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
16493 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
16494 plus and minus and convert them into IFN_COND_FMA-based sequences.
16495 (convert_mult_to_fma): Handle conditional plus and minus.
16497 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16499 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
16500 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
16501 (cond_fnms_optab): New optabs.
16502 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
16503 internal functions.
16504 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
16505 * internal-fn.h (get_conditional_internal_fn): Declare.
16506 (get_unconditional_internal_fn): Likewise.
16507 * internal-fn.c (cond_ternary_direct): New macro.
16508 (expand_cond_ternary_optab_fn): Likewise.
16509 (direct_cond_ternary_optab_supported_p): Likewise.
16510 (FOR_EACH_COND_FN_PAIR): Likewise.
16511 (get_conditional_internal_fn): New function.
16512 (get_unconditional_internal_fn): Likewise.
16513 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
16514 (gimple_match_op::gimple_match_op): Add a new overload for 5
16516 (gimple_match_op::set_op): Likewise.
16517 (gimple_resimplify5): Declare.
16518 * genmatch.c (decision_tree::gen): Generate simplifications for
16520 * gimple-match-head.c (gimple_simplify): Define an overload for
16521 5 operands. Handle calls with 5 arguments in the top-level overload.
16522 (convert_conditional_op): Handle conversions from unconditional
16523 internal functions to conditional ones.
16524 (gimple_resimplify5): New function.
16525 (build_call_internal): Pass a fifth operand.
16526 (maybe_push_res_to_seq): Likewise.
16527 (try_conditional_simplification): Try converting conditional
16528 internal functions to unconditional internal functions.
16529 Handle 3-operand unconditional forms.
16530 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
16531 Define ternary equivalents of the current rules for binary conditional
16532 internal functions.
16533 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
16534 ternary operations.
16535 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
16536 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
16537 (optab): Handle them.
16538 (SVE_COND_FP_TERNARY): New int iterator.
16539 (sve_fmla_op, sve_fmad_op): New int attributes.
16540 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
16541 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
16542 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
16544 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16546 * target.def (preferred_else_value): New target hook.
16547 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
16548 * doc/tm.texi: Regenerate.
16549 * targhooks.h (default_preferred_else_value): Declare.
16550 * targhooks.c (default_preferred_else_value): New function.
16551 * internal-fn.h (conditional_internal_fn_code): Declare.
16552 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
16553 (get_conditional_internal_fn): Use it.
16554 (conditional_internal_fn_code): New function.
16555 * gimple-match.h (gimple_match_cond): New struct.
16556 (gimple_match_op): Add a cond member function.
16557 (gimple_match_op::gimple_match_op): Update all forms to take a
16559 * genmatch.c (expr::gen_transform): Use the same condition as res_op
16560 for the suboperation, but don't specify a particular else_value.
16561 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
16562 (visit_nary_op, visit_reference_op_load): Pass
16563 gimple_match_cond::UNCOND to the gimple_match_op constructor.
16564 * gimple-match-head.c: Include tree-eh.h
16565 (convert_conditional_op): New function.
16566 (maybe_resimplify_conditional_op): Likewise.
16567 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
16568 (gimple_resimplify2): Likewise.
16569 (gimple_resimplify3): Likewise.
16570 (gimple_resimplify4): Likewise.
16571 (maybe_push_res_to_seq): Return null for conditional operations.
16572 (try_conditional_simplification): New function.
16573 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
16575 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
16577 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
16579 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
16581 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
16583 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
16585 (hash_tree): Do not hash DECL_FCONTEXT
16586 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
16587 Do not stream DECL_FCONTEXT.
16588 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
16589 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
16591 2018-07-12 Richard Biener <rguenther@suse.de>
16594 * dwarf2out.c (gen_block_die): Only output blocks when they have
16595 at least one !DECL_IGNORED_P variable.
16597 2018-07-12 Richard Biener <rguenther@suse.de>
16600 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
16601 Remove -mieee-fp handling.
16603 2018-07-12 Richard Biener <rguenther@suse.de>
16605 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
16606 left-over from last patch.
16608 2018-07-12 Jakub Jelinek <jakub@redhat.com>
16610 PR tree-optimization/86492
16611 * gimple-ssa-store-merging.c
16612 (imm_store_chain_info::coalesce_immediate_stores): Call
16613 check_no_overlap even for the merge_overlapping case. Formatting fix.
16615 2018-07-12 Richard Biener <rguenther@suse.de>
16617 PR middle-end/86479
16618 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
16619 move possibly trapping operations into the conditional.
16621 2018-07-12 Richard Biener <rguenther@suse.de>
16623 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
16624 (vn_lookup_simplify_result): Remove recursion limit applied
16626 (vn_nary_build_or_lookup_1): Adjust.
16627 (try_to_simplify): Likewise.
16628 * gimple-match-head.c (gimple_resimplify1): Instead apply one
16630 (gimple_resimplify2): Likewise.
16631 (gimple_resimplify3): Likewise.
16632 (gimple_resimplify4): Likewise.
16634 2018-07-11 Jakub Jelinek <jakub@redhat.com>
16636 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
16637 Use __mmask64 type instead of __mmask8 for __M argument.
16638 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
16639 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
16641 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
16642 __mmask16 for __M argument.
16643 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
16644 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
16645 to __mmask16 instead of __mmask8.
16646 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
16647 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
16648 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
16649 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
16650 instead of __mmask16 for __U argument.
16651 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
16652 __mmask16 instead of __mmask8 for __U argument.
16653 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
16655 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
16657 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
16659 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
16661 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
16663 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
16665 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
16667 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
16669 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
16671 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
16673 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
16674 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
16675 return type as well as __M argument type and all casts from __mmask8
16677 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
16678 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
16679 return type as well as __M argument type and all casts from __mmask8
16681 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
16682 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
16683 return type as well as __M argument type and all casts from __mmask8
16685 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
16686 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
16687 return type as well as __M argument type and all casts from __mmask8
16689 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
16690 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
16693 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
16695 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
16696 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
16699 2018-07-11 Paul Koning <ni1d@arrl.net>
16701 * doc/md.texi (define_subst): Document how multiple occurrences of
16702 the same argument in the replacement pattern are handled.
16704 2018-07-11 Paul Koning <ni1d@arrl.net>
16706 * doc/extend.texi (Common Variable Attributes): Move "mode" into
16707 alphabetical order.
16708 (Common Type Attributes): Add "mode" attribute.
16710 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
16712 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
16713 stream DECL_ORIGINAL_TYPE.
16714 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
16715 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
16716 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
16717 Do not walk original type.
16718 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
16720 (write_ts_decl_non_common_tree_pointers): Do not stream
16722 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
16723 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
16725 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
16727 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
16728 thread twice from the same starting edge.
16730 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
16732 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
16733 * gimple.c (gimple_call_nonnull_result_p): ...here...
16734 (gimple_call_nonnull_arg): ...and here.
16735 * gimple.h (gimple_call_nonnull_result_p): New.
16736 (gimple_call_nonnull_arg): New.
16738 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
16740 * config/arm/arm-cpus.in: Move information from fpu field of each
16741 cpu definition to the isa field.
16742 * config/arm/parsecpu.awk (fpu): Delete match rule.
16743 (gen_comm_data): Don't add bits from the CPU's FPU entry.
16745 2018-07-11 Richard Biener <rguenther@suse.de>
16748 * dwarf2out.c (init_sections_and_labels): Use
16749 output_asm_line_debug_info consistently.
16750 (dwarf2out_early_finish): Likewise.
16751 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
16754 2018-07-11 Richard Biener <rguenther@suse.de>
16756 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
16757 Rework father_bb setting in a way to avoid propagating constants
16758 multiple times on a loop body.
16760 2018-07-10 Mark Wielaard <mark@klomp.org>
16763 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
16765 2018-07-10 Richard Biener <rguenther@suse.de>
16767 * hash-map.h (hash_map::iterator::operator*): Return
16768 references to key and value.
16770 2018-07-10 Jakub Jelinek <jakub@redhat.com>
16773 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
16774 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
16775 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
16776 (gimplify_omp_for): For composite loops, move outer
16777 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
16778 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
16779 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
16780 TREE_LIST for both the original class iterator and the "last" helper
16781 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
16782 loop, remember has_decl_expr from outer composite loops for the
16783 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
16785 2018-07-09 Martin Sebor <msebor@redhat.com>
16787 PR middle-end/77357
16788 PR middle-end/86428
16789 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
16790 accessing implicitly initialized array elements.
16791 * expr.c (string_constant): Handle string initializers of
16792 character arrays within aggregates.
16793 * gimple-fold.c (fold_array_ctor_reference): Add argument.
16794 Store element offset. As a special case, handle zero size.
16795 (fold_nonarray_ctor_reference): Same.
16796 (fold_ctor_reference): Add argument. Store subobject offset.
16797 * gimple-fold.h (fold_ctor_reference): Add argument.
16799 2018-07-09 Paul Koning <ni1d@arrl.net>
16801 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
16802 (pdp11_insn_cost): New function.
16803 (pdp11_md_asm_adjust): New function.
16804 (TARGET_INVALID_WITHIN_DOLOOP): Define.
16805 (pdp11_rtx_costs): Update to match machine better.
16806 (output_addr_const_pdp11): Correct format mismatch warnings.
16807 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
16808 * config/pdp11/pdp11.md: General change to add base_cost and/or
16809 length attributes for use by new pdp11_insn_cost function.
16810 (MIN_BRANCH): Correct definition.
16812 (doloop_end): Use standard pattern name for looping pattern.
16813 (doloop_end_nocc): New.
16814 (movsf): Add another constraint alternative.
16815 (zero_extendqihi2): Add constraint alternatives for not in place
16817 (zero_extendhisi2): Remove.
16818 (shift patterns): Add CC handling variants.
16822 (define_peephole2): New peephole to recognize mov that sets CC for
16825 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16827 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
16828 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
16831 2018-07-09 Jakub Jelinek <jakub@redhat.com>
16834 * real.c (real_nextafter): Return true if result is denormal.
16836 2018-07-09 Martin Liska <mliska@suse.cz>
16838 * common.opt: Add back wrongly removed attribute.
16840 2018-07-09 Richard Biener <rguenther@suse.de>
16843 * dwarf2out.c (gen_block_die): For an early generated DIE
16844 always output high/low PC attributes.
16846 2018-07-09 Tom de Vries <tdevries@suse.de>
16848 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
16849 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
16850 onto VAR_DECL with abstract origin.
16852 2018-07-07 Jim Wilson <jimw@sifive.com>
16854 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
16856 2018-07-07 Tom de Vries <tdevries@suse.de>
16858 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
16860 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
16862 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
16864 (combine_bound): Use wide-int overflow calculation instead of
16866 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
16868 * fold-const.c (int_const_binop_2): Same.
16869 (extract_muldiv_1): Same.
16870 (fold_div_compare): Same.
16871 (fold_abs_const): Same.
16873 * poly-int.h (add): Same.
16877 * predict.c (predict_iv_comparison): Same.
16878 * profile-count.c (slow_safe_scale_64bit): Same.
16879 * simplify-rtx.c (simplify_const_binary_operation): Same.
16880 * tree-chrec.c (tree_fold_binomial): Same.
16881 * tree-data-ref.c (split_constant_offset_1): Same.
16882 * tree-if-conv.c (idx_within_array_bound): Same.
16883 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
16884 * tree-ssa-phiopt.c (minmax_replacement): Same.
16885 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
16886 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
16887 * vr-values.c (vr_values::adjust_range_with_scev): Same.
16888 * wide-int.cc (wi::add_large): Same.
16889 (wi::mul_internal): Same.
16890 (wi::sub_large): Same.
16891 (wi::divmod_internal): Same.
16892 * wide-int.h: Change overflow type to overflow_type for neg, add,
16893 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
16894 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
16895 mul_internal, divmod_internal.
16896 (overflow_type): New enum.
16897 (accumulate_overflow): New.
16899 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
16901 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
16902 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
16904 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
16906 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
16907 argument is checked for zero before entering loop, avoid checking again.
16909 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
16911 * gimplify.h (generic_expr_could_trap_p): Set as global function.
16912 * gimplify.h (generic_expr_could_trap_p): Likwise.
16913 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
16915 2018-07-06 Jakub Jelinek <jakub@redhat.com>
16917 PR tree-optimization/86401
16918 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
16919 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
16920 (fold_bit_and_mask): ... here. New helper function for match.pd.
16921 * fold-const.h (fold_bit_and_mask): Declare.
16922 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
16924 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
16927 * target.def (translate_mode_attribute): New hook.
16928 * targhooks.h (default_translate_mode_attribute): Declare.
16929 * targhooks.c (default_translate_mode_attribute): New function.
16930 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
16931 * doc/tm.texi: Regenerate.
16932 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
16933 (rs6000_translate_mode_attribute): New function.
16935 2018-07-06 Paul Koning <ni1d@arrl.net>
16937 * doc/md.texi (define_split): Document DONE and FAIL.
16938 (define_peephole2): Ditto.
16940 2018-07-05 Jeff Law <law@redhat.com>
16942 PR tree-optimization/86010
16943 * tree-ssa-dse.c (compute_trims): More aggressively trim at
16944 both the head and tail of mem* and str* calls.
16946 2018-07-05 Jim Wilson <jimw@sifive.com>
16948 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
16950 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
16952 * config/aarch64/aarch64-simd.md: correct flags text for
16953 MIN_EXPR replacement.
16955 2018-07-05 James Clarke <jrtc27@jrtc27.com>
16957 * configure: Regenerated.
16959 2018-07-05 Carl Love <cel@us.ibm.com>
16961 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
16962 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
16963 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
16964 VSX_BUILTIN_DOUBLEL_V4SF.
16966 2018-07-05 Martin Sebor <msebor@redhat.com>
16969 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
16970 than its domain to compute its the upper bound of a char array.
16972 2018-07-05 Nathan Sidwell <nathan@acm.org>
16974 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
16975 * doc/cpp.texi: Update comment.
16976 * doc/tm.texi: Rebuilt.
16977 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
16978 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
16979 * doc/extend.texi (Backwards Compatibility): Clarify it is system
16980 headers affected by extern "C".
16981 * system.h: Poison NO_IMPLICIT_EXTERN_C.
16982 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
16983 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
16984 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
16985 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
16986 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
16987 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
16988 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
16989 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
16990 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
16991 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
16992 config/sparc/sp64-elf.h, config/spu/spu.h,
16993 config/stormy16/stormy16.h, config/v850/v850.h,
16994 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
16995 define NO_IMPLICIT_EXTERN_C.
16996 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
16998 2018-07-05 Tamar Christina <tamar.christina@arm.com>
17001 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17002 instead of GET_MODE_SIZE when comparing Units.
17004 2018-07-05 Tamar Christina <tamar.christina@arm.com>
17007 * rtlanal.c (set_noop_p): Constrain on mode change,
17008 include hard-reg-set.h
17010 2018-07-05 Tamar Christina <tamar.christina@arm.com>
17012 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
17014 2018-07-05 Jakub Jelinek <jakub@redhat.com>
17017 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
17020 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
17022 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
17024 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
17027 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
17029 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
17031 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
17033 PR middle-end/86380
17034 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
17036 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
17038 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
17039 neg_*_op* variables.
17041 2018-07-04 Martin Liska <mliska@suse.cz>
17043 * tree-switch-conversion.c: Define
17044 max_ratio_for_speed and max_ratio_for_size constants.
17046 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
17047 Martin Liska <mliska@suse.cz>
17049 PR middle-end/66240
17052 * common.opt: Rename align options with 'str_' prefix.
17053 * common/config/i386/i386-common.c (set_malign_value): New
17055 (ix86_handle_option): Use it to set -falign-* options/
17056 * config/aarch64/aarch64-protos.h (struct tune_params): Change
17057 type from int to string.
17058 * config/aarch64/aarch64.c: Update default values from int
17060 * config/alpha/alpha.c (alpha_override_options_after_change):
17062 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
17063 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17064 max skip conditionally.
17065 * config/i386/freebsd.h (SUBALIGN_LOG): New.
17066 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17067 max skip conditionally.
17068 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17069 max skip conditionally.
17070 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
17071 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17072 max skip conditionally.
17073 * config/i386/i386.c (struct ptt): Change type from int to
17075 (ix86_default_align): Set default values.
17076 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
17077 max skip conditionally.
17078 * config/i386/iamcu.h (SUBALIGN_LOG): New.
17079 (ASM_OUTPUT_MAX_SKIP_ALIGN):
17080 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
17081 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17082 max skip conditionally.
17083 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
17084 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
17085 * config/i386/x86-64.h (SUBALIGN_LOG): New.
17086 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
17087 max skip conditionally.
17088 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17089 * config/ia64/ia64.c (ia64_option_override): Set default values
17090 for alignment options.
17091 * config/m68k/m68k.c: Handle new str_align_* options.
17092 * config/mips/mips.c (mips_set_compression_mode): Change
17094 (mips_option_override): Set default values for options.
17095 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
17097 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17099 * config/rx/rx.c (rx_option_override): Likewise.
17100 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
17101 (LABEL_ALIGN): Use align_labels_log.
17102 (LOOP_ALIGN): Use align_loops_align.
17103 * config/s390/s390.c (s390_asm_output_function_label): Use new
17105 * config/sh/sh.c (sh_override_options_after_change):
17106 Change type of constants.
17107 * config/spu/spu.c (spu_sched_init): Likewise.
17108 * config/sparc/sparc.c (sparc_option_override): Set default
17109 values for options.
17110 * config/visium/visium.c (visium_option_override): Likewise.
17111 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
17112 emit p2align format with last argument if it's not needed.
17113 * doc/invoke.texi: Document extended format of -falign-*.
17114 * final.c: Use align_labels alignment.
17115 * flags.h (struct target_flag_state): Change type to use
17117 (struct align_flags_tuple): New.
17118 (struct align_flags): Likewise.
17119 (align_loops_log): Redefine macro to use new types.
17120 (align_loops_max_skip): Redefine macro to use new types.
17121 (align_jumps_log): Redefine macro to use new types.
17122 (align_jumps_max_skip): Redefine macro to use new types.
17123 (align_labels_log): Redefine macro to use new types.
17124 (align_labels_max_skip): Redefine macro to use new types.
17125 (align_functions_log): Redefine macro to use new types.
17126 (align_loops): Redefine macro to use new types.
17127 (align_jumps): Redefine macro to use new types.
17128 (align_labels): Redefine macro to use new types.
17129 (align_functions): Redefine macro to use new types.
17130 (align_functions_max_skip): Redefine macro to use new types.
17131 (align_loops_value): New macro.
17132 (align_jumps_value): New macro.
17133 (align_labels_value): New macro.
17134 (align_functions_value): New macro.
17135 * function.c (invoke_set_current_function_hook): Propagate
17136 alignment values from flags to global variables default in
17138 * ipa-icf.c (sem_function::equals_wpa): Use
17139 cl_optimization_option_eq instead of memcmp.
17140 * lto-streamer.h (cl_optimization_stream_out): Support streaming
17142 (cl_optimization_stream_in): Likewise.
17143 * optc-save-gen.awk: Support strings in cl_optimization.
17144 * opth-gen.awk: Likewise.
17145 * opts.c (finish_options): Remove error checking of invalid
17147 (MAX_CODE_ALIGN): Remove.
17148 (MAX_CODE_ALIGN_VALUE): Likewise.
17149 (parse_and_check_align_values): New function.
17150 (check_alignment_argument): Likewise.
17151 (common_handle_option): Use check_alignment_argument.
17152 * opts.h (parse_and_check_align_values): Declare.
17153 * toplev.c (init_alignments): Remove.
17154 (read_log_maxskip): New.
17155 (parse_N_M): Likewise.
17156 (parse_alignment_opts): Likewise.
17157 (backend_init_target): Remove usage of init_alignments.
17158 * toplev.h (parse_alignment_opts): Declare.
17159 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
17161 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
17162 * tree.c (cl_option_hasher::equal): New.
17163 * varasm.c: Use new global macros.
17165 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
17167 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
17168 Use a simpler align directive also if MAXSKIP = ALIGN-1.
17169 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17170 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17171 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17172 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17173 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
17174 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
17175 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
17176 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17177 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17178 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17179 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17181 2018-07-04 Martin Liska <mliska@suse.cz>
17182 Jonathan Wakely <jwakely@redhat.com>
17184 * coverage.c: Use correct type.
17185 * doc/invoke.texi: Language correction.
17187 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
17190 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
17191 ENDBRANCH for non-tail call which may return via indirect branch.
17192 * doc/extend.texi: Document indirect_return attribute.
17194 2018-07-03 Martin Sebor <msebor@redhat.com>
17196 PR tree-optimization/86274
17197 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
17199 (format_floating): Correct handling of infinities and NaNs.
17201 2018-07-03 Martin Sebor <msebor@redhat.com>
17203 * print-tree.c (print_real_cst): New function.
17204 (print_node_brief): Call it.
17205 (print_node): Ditto.
17207 2018-07-03 Jeff Law <law@redhat.com>
17209 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
17210 into a single pattern.
17212 * config/h8300/h8300.md (ors code_iterator): New.
17213 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
17214 a single pattern and single splitter.
17215 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
17216 (iorqi3_1, xorqi3_1): Likewise.
17217 (iorqi3, xorqi3 expanders): Similarly.
17219 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
17220 (movmd_internal) into a single pattern using the P mode iterator.
17221 (movmd splitters): Similarly.
17222 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
17223 (movsd splitters): Similarly.
17225 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
17226 ADDB, ADDW and ADDL into a single ADD attribute which selects the
17227 right table based on the size of the operand.
17228 * config/h8300/h8300.md (length_table): Corresponding changes. All
17229 references to "addb", "addw" and "addl" changed to "add".
17230 (btst patterns): Merge two variants into a single pattern.
17231 (tstqi, tsthi): Likewise.
17232 (addhi3_incdec, addsi3_incdec): Likewise.
17233 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
17234 (mulhi3, mulsi3): Likewise.
17235 (udivhi3, udivsi3): Likewise.
17236 (divhi3, divsi3): Likewise.
17237 (andorqi3, andorhi3, andorsi3): Likewise.
17239 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
17242 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
17243 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
17245 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17247 PR tree-optimization/85694
17248 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
17249 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
17251 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
17252 (<u>avg<mode>3_ceil): New patterns.
17254 2018-07-03 David Malcolm <dmalcolm@redhat.com>
17256 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
17257 scan-tree-dump directive.
17258 * gcc.dg/vect/slp-perm-2.c: Likewise.
17259 * gcc.dg/vect/slp-perm-3.c: Likewise.
17260 * gcc.dg/vect/slp-perm-5.c: Likewise.
17261 * gcc.dg/vect/slp-perm-6.c: Likewise.
17262 * gcc.dg/vect/slp-perm-7.c: Likewise.
17263 * gcc.dg/vect/slp-perm-8.c: Likewise.
17265 2018-07-03 Marek Polacek <polacek@redhat.com>
17267 PR middle-end/86202
17268 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
17270 2018-07-03 Richard Biener <rguenther@suse.de>
17273 * tree-ssa-structalias.c (find_func_clobbers): Properly
17274 handle indirect calls.
17276 2018-07-03 Jeff Law <law@redhat.com>
17278 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
17279 (shifts): New code iterator.
17280 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
17281 expander. Fix HImode handling on H8/SX.
17282 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
17283 (subqi3, subhi3, subsi3 expanders): Likewise.
17284 (andqi3, andhi3, andsi3 expanders): Likewise.
17285 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
17286 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
17287 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
17288 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
17289 (zero_extendqihi2, zero_extendqisi2): Likewise.
17290 (extendqihi2, extendqisi2): Likewise.
17291 (rotlqi3, rotlhi3, rotlsi3): Likewise.
17292 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
17293 (rotlqi3_1, rotlhi3_1): Likewise.
17294 (logicalhi3_sn, logicalsi3_sn): Likewise.
17295 (logicalhi3, logicalsi3): Likewise.
17297 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17299 * tree-vect-patterns.c (vect_recog_rotate_pattern)
17300 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
17301 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
17302 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
17303 type to append_pattern_def_seq instead of creating a stmt_vec_info
17305 (build_mask_conversion): Likewise. Remove vinfo argument.
17306 (vect_add_conversion_to_patterm): Likewise, renaming to...
17307 (vect_add_conversion_to_pattern): ...this.
17308 (vect_recog_mask_conversion_pattern): Update call to
17309 build_mask_conversion. Pass the vector type to
17310 append_pattern_def_seq here too.
17311 (vect_recog_gather_scatter_pattern): Update call to
17312 vect_add_conversion_to_pattern.
17314 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17316 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
17317 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17318 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
17319 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
17320 STMT_VINFO_PATTERN_DEF_SEQ to null here.
17321 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
17322 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
17323 append_pattern_def_seq instead of new_pattern_def_seq.
17324 (vect_recog_divmod_pattern): Do both of the above.
17325 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
17328 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17330 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
17331 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
17332 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
17333 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
17334 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
17335 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
17336 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
17337 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
17338 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
17339 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
17340 parameter with a single stmt_vec_info.
17341 (vect_recog_func_ptr): Likewise.
17342 (vect_recog_gather_scatter_pattern): Likewise, folding in...
17343 (vect_try_gather_scatter_pattern): ...this.
17344 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
17345 the stmt_vec_info of the statement to be matched. Don't clear
17346 STMT_VINFO_RELATED_STMT.
17347 (vect_pattern_recog): Update call accordingly.
17349 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17351 PR tree-optimization/85694
17352 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
17353 (uavgM3_ceil): Document new optabs.
17354 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
17355 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
17357 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
17358 (savg_ceil_optab): New optabs.
17359 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
17360 (vect_vect_recog_func_ptrs): Add it.
17361 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
17362 constant directly from the associated lhs.
17364 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17366 * tree-vect-patterns.c (vect_split_statement): New function.
17367 (vect_convert_input): Use it to try to split an existing cast.
17369 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17371 * poly-int.h (print_hex): New function.
17372 * dumpfile.h (dump_dec, dump_hex): Declare.
17373 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
17374 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
17375 min_input_precision, operation_precision and operation_sign.
17376 * tree-vect-patterns.c (vect_get_range_info): New function.
17377 (vect_same_loop_or_bb_p, vect_single_imm_use)
17378 (vect_operation_fits_smaller_type): Delete.
17379 (vect_look_through_possible_promotion): Add an optional
17380 single_use_p parameter.
17381 (vect_recog_over_widening_pattern): Rewrite to use new
17382 stmt_vec_info infomration. Handle one operation at a time.
17383 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
17384 (vect_truncatable_operation_p, vect_set_operation_type)
17385 (vect_set_min_input_precision): New functions.
17386 (vect_determine_min_output_precision_1): Likewise.
17387 (vect_determine_min_output_precision): Likewise.
17388 (vect_determine_precisions_from_range): Likewise.
17389 (vect_determine_precisions_from_users): Likewise.
17390 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
17391 (vect_vect_recog_func_ptrs): Put over_widening first.
17393 (vect_pattern_recog): Call vect_determine_precisions.
17395 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17397 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
17398 statements that have been replaced by further pattern statements.
17399 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
17401 2018-07-03 Richard Biener <rguenther@suse.de>
17403 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
17404 always set *dt. Dump vectype in vectype overload.
17405 * dumpfile.h (dump_gimple_expr): New function.
17406 (dump_gimple_expr_loc): Likewise.
17407 * dumpfile.c (dump_gimple_expr): New function.
17408 (dump_gimple_expr_loc): Likewise.
17410 2018-07-02 Jeff Law <law@redhat.com>
17412 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
17413 the H8/300, H8/300H and H8/S variants into a single pattern.
17414 (movhi_h8300, movqi_h8300hs): Similarly.
17415 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
17416 (QHI mode iterator): New.
17418 * config/h8300/h8300.md: Remove trailing whitespace.
17420 2018-07-02 Jim Wilson <jimw@sifive.com>
17422 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
17423 instead of emit_insn for interrupt returns.
17424 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
17425 (riscv_sret, riscv_uret): Likewise.
17427 2018-07-02 David Malcolm <dmalcolm@redhat.com>
17429 * pretty-print.c (selftest::test_pp_format): Move save and restore
17430 of quotes to class auto_fix_quotes, and add an instance.
17431 * selftest.c: Include "intl.h".
17432 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
17433 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
17434 * selftest.h (selftest::auto_fix_quotes): New class.
17436 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
17438 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
17439 (aarch64_sve_prepare_conditional_op): Remove.
17440 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
17441 Allow aarch64_simd_reg_or_zero as select operand; remove
17442 the aarch64_sve_prepare_conditional_op call.
17443 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
17444 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
17445 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
17446 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
17447 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
17448 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
17449 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
17450 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
17451 and a splitters to match all of the *_any patterns.
17452 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
17454 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
17455 (SVE_COND_FP_BINARY_REV): Remove.
17456 (sve_int_op_rev, sve_fp_op_rev): New.
17457 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
17458 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
17459 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
17460 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
17461 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
17462 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
17463 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
17464 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
17465 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
17467 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
17468 Remove match_dup 1 from the inner unspec.
17469 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
17471 * config/aarch64/aarch64.md (movprfx): New attr.
17472 (length): Default movprfx to 8.
17473 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
17474 (*madd<SVE_I>, *msub<SVE_I): Likewise.
17475 (*<su>mul<SVE_I>3_highpart): Likewise.
17476 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
17477 (*v<ASHIFT><SVE_I>3): Likewise.
17478 (*<su><MAXMIN><SVE_I>3): Likewise.
17479 (*<su><MAXMIN><SVE_F>3): Likewise.
17480 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
17481 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
17482 (*div<SVE_F>4): Likewise.
17484 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
17486 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
17489 2018-07-02 Richard Biener <rguenther@suse.de>
17491 PR tree-optimization/86363
17492 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
17493 memset argument refers to a non-variable address.
17495 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
17497 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
17498 {PLUS,MINUS}_EXPR code to...
17499 (adjust_symbolic_bound): ...here,
17500 (combine_bound): ...here,
17501 (set_value_range_with_overflow): ...and here.
17503 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
17505 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
17507 (extract_range_from_abs_expr): ...here.
17509 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
17511 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
17512 -fno-omit-frame-pointer when not optimizing.
17514 2018-07-02 Martin Liska <mliska@suse.cz>
17517 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
17518 (propagate_nothrow): Likewise.
17520 2018-07-02 Martin Liska <mliska@suse.cz>
17523 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
17525 2018-07-02 David Malcolm <dmalcolm@redhat.com>
17527 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
17528 function in r262149, changing "loc" param from source_location to
17529 const dump_location_t &.
17530 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
17531 declaration, as above.
17533 2018-07-01 Paul Koning <ni1d@arrl.net>
17535 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
17536 -munit-asm, -mgnu-asm, -mdec-asm.
17537 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
17538 (pdp11_output_labelref): New.
17539 (pdp11_output_def): New.
17540 (pdp11_output_addr_vec_elt): New.
17541 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
17542 %# and %@ format codes.
17543 (pdp11_option_override): New.
17544 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
17545 (pdp11_output_ident): New.
17546 (pdp11_asm_named_section): New.
17547 (pdp11_asm_init_sections): New.
17548 (pdp11_file_start): New.
17549 (pdp11_file_end): New.
17550 (output_ascii): Use .ascii/.asciz for -mdec-asm.
17551 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
17552 %o, like %c but octal.
17553 (pdp11_option_override): New.
17554 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
17556 (DATA_SECTION_ASM_OP): Ditto.
17557 (READONLY_DATA_SECTION_ASM_OP): New.
17558 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
17559 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
17560 (ASM_OUTPUT_LABELREF): Ditto.
17561 (ASM_OUTPUT_DEF): Ditto.
17562 (ASM_OUTPUT_EXTERNAL): New.
17563 (ASM_OUTPUT_SOURCE_FILENAME): New.
17564 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
17565 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
17566 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
17567 %# and %@ format codes.
17568 * config/pdp11/pdp11.opt (mgnu-asm): New.
17569 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
17570 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
17571 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
17573 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
17575 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
17576 dereferencing path[] beyond its length.
17578 (debug_all_paths): New.
17579 (rewire_first_differing_edge): New.
17580 (adjust_paths_after_duplication): New.
17581 (duplicate_thread_path): Call adjust_paths_after_duplication.
17583 (thread_through_all_blocks): Add new argument to
17584 duplicate_thread_path.
17586 2018-06-30 Jim Wilson <jimw@sifive.com>
17588 * config/riscv/predicates.md (p2m1_shift_operand): New.
17589 (high_mask_shift_operand): New.
17590 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
17591 pattern using p2m1_shift_operand.
17592 (lshsi3_zero_extend_3+2): New combiner pattern using
17593 high_mask_shift_operand.
17595 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17597 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
17599 (vect_recog_rotate_pattern): ...here.
17600 (vect_convert_input): Try to insert casts of invariants in the
17602 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
17603 preheader to be empty.
17605 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17607 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
17608 vector type. If given, install it in the new statement's
17609 STMT_VINFO_VECTYPE.
17610 (vect_element_precision): New function.
17611 (vect_unpromoted_value): New struct.
17612 (vect_unpromoted_value::vect_unpromoted_value): New function.
17613 (vect_unpromoted_value::set_op): Likewise.
17614 (vect_look_through_possible_promotion): Likewise.
17615 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
17616 (vect_widened_op_tree, vect_convert_input): Likewise.
17617 (vect_convert_inputs, vect_convert_output): Likewise.
17618 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
17619 to handle the optional cast of the multiplication result and
17620 vect_widened_op_tree to detect the widened multiplication itself.
17621 Do not require the input and output of promotion casts to have
17622 the same sign, but base the signedness of the operation on the
17623 input rather than the result. If the pattern includes two
17624 promotions, check that those promotions have the same sign.
17625 Do not restrict the MULT_EXPR handling to a double-width result;
17626 handle quadruple-width results and wider. Use vect_convert_inputs
17627 to convert the inputs to the common type.
17628 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
17629 to handle the optional cast of the ABS result. Also allow a sign
17630 change or a sign extension between the ABS and MINUS.
17631 Use vect_widened_op_tree to detect the widened subtraction and use
17632 vect_convert_inputs to convert the inputs to the common type.
17633 (vect_handle_widen_op_by_const): Delete.
17634 (vect_recog_widen_op_pattern): New function.
17635 (vect_recog_widen_mult_pattern): Use it.
17636 (vect_recog_widen_shift_pattern): Likewise.
17637 (vect_recog_widen_sum_pattern): Use
17638 vect_look_through_possible_promotion to handle the promoted
17641 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17643 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
17644 the containing gimple_seq *.
17645 * gimple-iterator.h (gsi_for_stmt): Declare it.
17646 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
17647 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
17648 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
17649 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
17650 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
17652 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
17654 (vect_mark_pattern_stmts): ...here. Handle cases in which the
17655 statement being replaced is part of an existing pattern
17656 definition sequence, inserting the new pattern statements before
17658 (vect_pattern_recog_1): Don't return a bool. If the statement
17659 is already part of a pattern, instead apply pattern matching
17660 to the pattern definition statements. Don't clear the
17661 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
17662 (vect_pattern_recog): Don't break after the first match;
17663 continue processing the pattern definition statements instead.
17664 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
17666 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17668 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17669 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17670 (vect_recog_widen_sum_pattern): Use it.
17672 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17674 * tree-vect-loop.c (vectorizable_reduction): Assert that the
17675 phi is not a pattern statement and has not been replaced by
17676 a pattern statement.
17677 * tree-vect-patterns.c (type_conversion_p): Don't check
17678 STMT_VINFO_IN_PATTERN_P.
17679 (vect_recog_vector_vector_shift_pattern): Likewise.
17680 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
17681 the pattern statement rather than the original statement; check
17682 directly for a WIDEN_MULT_EXPR here.
17683 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
17684 vect_is_simple_use to return the pattern statement rather
17685 than the original statement; use is_pattern_stmt_p to check
17686 for such a pattern statement.
17687 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
17688 to return the pattern statement rather than the original statement;
17689 don't do the same transformation here.
17690 (vect_is_simple_use): If the defining statement has been replaced
17691 by a pattern statement, return the pattern statement instead.
17692 Remove the corresponding (local) transformation from the vectype
17695 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17697 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
17698 end and default to null.
17699 * tree-vect-loop.c (vect_create_epilog_for_reduction)
17700 (vectorizable_reduction): Update calls accordingly, dropping the
17701 gimple ** argument if the passed-back statement isn't needed.
17702 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
17703 (vect_recog_rotate_pattern): Likewise.
17704 (vect_recog_mask_conversion_pattern): Likewise.
17705 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
17706 (vect_mask_constant_operand_p): Likewise.
17707 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
17708 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
17709 (get_group_load_store_type, get_load_store_type): Likewise.
17710 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
17711 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
17712 (vectorizable_conversion, vectorizable_assignment): Likewise.
17713 (vectorizable_shift, vectorizable_operation): Likewise.
17714 (vectorizable_store, vect_is_simple_cond): Likewise.
17715 (vectorizable_condition, vectorizable_comparison): Likewise.
17716 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
17717 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
17718 and move it to the end. Cope with null def_stmt_outs.
17720 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
17722 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
17724 2018-06-29 Jeff Law <law@redhat.com>
17726 * config/v850/v850.c (v850_legitimate_address_p): Handle large
17727 displacements for TARGET_V850E2V3 and newer.
17728 (TARGET_LRA_P): Remove. Defaults to LRA now.
17729 * config/v850/v850.md (sign23byte_load): Remove.
17730 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
17731 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
17733 2018-06-29 Martin Liska <mliska@suse.cz>
17736 * coverage.c (coverage_init): Mangle full path name.
17737 * doc/invoke.texi: Document the change.
17738 * gcov-io.c (mangle_path): New.
17739 * gcov-io.h (mangle_path): Likewise.
17740 * gcov.c (mangle_name): Use mangle_path for path mangling.
17742 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17744 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
17745 if starting source register is not even.
17747 2018-06-29 Martin Liska <mliska@suse.cz>
17749 PR tree-optimization/86263
17750 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
17751 Make edge redirection.
17753 2018-06-29 David Malcolm <dmalcolm@redhat.com>
17755 * dumpfile.c (dump_loc): Add indentation based on scope depth.
17756 (dump_scope_depth): New variable.
17757 (get_dump_scope_depth): New function.
17758 (dump_begin_scope): New function.
17759 (dump_end_scope): New function.
17760 * dumpfile.h (get_dump_scope_depth): New declaration.
17761 (dump_begin_scope): New declaration.
17762 (dump_end_scope): New declaration.
17763 (class auto_dump_scope): New class.
17764 (AUTO_DUMP_SCOPE): New macro.
17765 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
17768 2018-06-29 Richard Biener <rguenther@suse.de>
17770 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
17771 compute_all_dependences succeeds.
17772 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
17773 exceed --param loop-max-datarefs-for-datadeps.
17775 2018-06-29 Jakub Jelinek <jakub@redhat.com>
17777 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
17779 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
17782 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
17783 alternative 0 in preferred_for_speed attribute.
17785 2018-06-28 Paul Koning <ni1d@arrl.net>
17787 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
17788 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
17789 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
17790 * config/pdp11/pdp11.md: Correct "length" attribute calculation
17791 for shift insn patterns.
17793 2018-06-28 David Malcolm <dmalcolm@redhat.com>
17795 * cgraph.c (cgraph_node::get_body): Replace assignments to
17796 "dump_file" with calls to set_dump_file.
17797 * dumpfile.c (alt_dump_file): Make static, and group with...
17798 (alt_flags): ...this definition.
17799 (dumps_are_enabled): New variable.
17800 (refresh_dumps_are_enabled): New function.
17801 (set_dump_file): New function.
17802 (set_alt_dump_file): New function.
17803 (gcc::dump_manager::dump_start): Replace assignments to
17804 "dump_file" and "alt_dump_file" with calls to set_dump_file and
17806 (gcc::dump_manager::dump_finish): Likewise.
17807 * dumpfile.h (alt_dump_file): Delete decl.
17808 (dumps_are_enabled): New variable decl.
17809 (set_dump_file): New function decl.
17810 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
17812 * tree-nested.c (lower_nested_functions): Replace assignments to
17813 "dump_file" with calls to set_dump_file.
17815 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
17817 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
17818 goto_locus of each outgoing edge of each basic block.
17820 2018-06-28 Richard Biener <rguenther@suse.de>
17822 * dwarf2out.c (decl_scope_table): Remove.
17823 (push_decl_scope): Likewise.
17824 (pop_decl_scope): Likewise.
17825 (gen_type_die_for_member): Do not call push/pop_decl_scope.
17826 (gen_struct_or_union_type_die): Likewise.
17827 (gen_tagged_type_die): Likewise.
17828 (dwarf2out_init): Do not initialize decl_scope_table.
17829 (dwarf2out_c_finalize): Do not free it.
17831 2018-06-28 Richard Biener <rguenther@suse.de>
17833 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
17834 deciding whether to not re-use a DIE.
17836 2018-06-28 Richard Biener <rguenther@suse.de>
17838 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
17839 DW_AT_abstract_origin attribute.
17841 2018-06-28 Martin Liska <mliska@suse.cz>
17843 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
17844 Use newly introduced constants.
17845 * tree-switch-conversion.h (struct jump_table_cluster):
17846 Define max_ratio_for_size and max_ratio_for_speed.
17848 2018-06-28 Martin Liska <mliska@suse.cz>
17850 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17851 Add new checking assert to catch invalid state.
17852 (jump_table_cluster::can_be_handled): Handle single case
17854 (jump_table_cluster::is_beneficial): Bail out for such case.
17855 (bit_test_cluster::find_bit_tests):
17856 Add new checking assert to catch invalid state.
17857 (bit_test_cluster::can_be_handled): Handle single case
17859 (bit_test_cluster::is_beneficial): Bail out for such case.
17860 (switch_decision_tree::analyze_switch_statement):
17863 2018-06-28 Martin Liska <mliska@suse.cz>
17865 * common.opt: Introduce -completion option.
17866 * gcc.c (driver_handle_option): Handle it.
17867 (driver::main): Print completions if completion
17869 * opt-suggestions.c (option_proposer::get_completions):
17871 (option_proposer::suggest_completion): Likewise.
17872 (option_proposer::find_param_completions): Likewise.
17873 (verify_autocompletions): Likewise.
17874 (test_completion_valid_options): Likewise.
17875 (test_completion_valid_params): Likewise.
17876 (in_completion_p): Likewise.
17877 (empty_completion_p): Likewise.
17878 (test_completion_partial_match): Likewise.
17879 (test_completion_garbage): Likewise.
17880 (opt_proposer_c_tests): Likewise.
17881 * opt-suggestions.h: Declare new functions.
17882 * opts.c (common_handle_option): Handle OPT__completion_.
17883 * selftest-run-tests.c (selftest::run_tests): Add
17884 opt_proposer_c_tests.
17885 * selftest.c (assert_str_startswith): New.
17886 * selftest.h (assert_str_startswith): Likewise.
17887 (opt_proposer_c_tests): New.
17888 (ASSERT_STR_STARTSWITH): Likewise.
17890 2018-06-28 Martin Liska <mliska@suse.cz>
17892 * Makefile.in: Add opt-suggestions.o.
17893 * gcc-main.c: Include opt-suggestions.h.
17894 * gcc.c (driver::driver): Likewise.
17895 (driver::~driver): Remove m_option_suggestions.
17896 (driver::build_option_suggestions): Moved to option_proposer.
17897 (driver::suggest_option): Likewise.
17898 (driver::handle_unrecognized_options): Use option_proposer.
17899 * gcc.h (class driver): Add new memver m_option_proposer.
17900 * opt-suggestions.c: New file.
17901 * opt-suggestions.h: New file.
17903 2018-06-28 Martin Liska <mliska@suse.cz>
17905 * vec.h (class auto_string_vec): New (moved from auto_argvec).
17906 (auto_string_vec::~auto_string_vec): Likewise.
17908 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
17910 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
17911 prevent_decl_creation_for_types fields up and add reset_location field.
17912 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
17913 statement if id->reset_location is true.
17914 (copy_edges_for_bb): Do not set goto_locus on the new edges if
17915 id->reset_location is true.
17916 (copy_phis_for_bb): Force input_location on the arguments if
17917 id->reset_location is true.
17918 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
17919 is set on the function to be inlined.
17921 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
17923 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
17925 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
17927 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
17928 registers for Pmode.
17929 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
17930 hard registers for the clobbered pseudo.
17932 2018-06-27 Paul Koning <ni1d@arrl.net>
17934 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
17935 mutually exclusive options.
17936 * config/pdp11/constraints.md (h): New constraint.
17937 (O): Update definition to match shift code generation.
17938 (D): New constraint.
17939 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
17941 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
17943 (output_jump): Change arguments.
17944 (pdp11_fixed_cc_regs): New function.
17945 (pdp11_cc_mode): Ditto.
17946 (pdp11_expand_shift): Ditto.
17947 (pdp11_assemble_shift): Ditto.
17948 (pdp11_small_shift): Ditto.
17949 (pdp11_branch_cost): Remove.
17950 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
17952 (pdp11_register_move_cost): Update for CC registers.
17953 (pdp11_rtx_costs): Add case for LSHIFTRT.
17954 (pdp11_output_jump): Add CCNZ mode conditional branches.
17955 (notice_update_cc_on_set): Remove.
17956 (pdp11_cc_mode): New function.
17957 (simple_memory_operand): Correct pre/post decrement case.
17958 (no_side_effect_operand): New function.
17959 (pdp11_regno_reg_class): Add CC_REGS class.
17960 (pdp11_fixed_cc_regs): New function.
17961 (pdp11_small_shift): New function.
17962 (pdp11_expand_shift): New function to expand shift insns.
17963 (pdp11_assemble_shift): New function to output shifts.
17964 (pdp11_branch_cost): Remove.
17965 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
17966 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
17967 (WCHAR_TYPE): Ditto.
17968 (PTRDIFF_TYPE): Ditto.
17969 (ADJUST_INSN_LENGTH): New macro.
17970 (FIXED_REGISTERS): Add CC registers.
17971 (CALL_USED_REGISTERS): Ditto.
17972 (reg_class): Ditto.
17973 (REG_CLASS_NAMES): Ditto.
17974 (REG_CLASS_CONTENTS): Ditto.
17975 (SELECT_CC_MODE): Use new function.
17976 (TARGET_FLAGS_REGNUM): New macro.
17977 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
17978 (cc0_reg_rtx): Remove.
17979 (CC_STATUS_MDEP): Remove.
17980 (CC_STATUS_MDEFP_INIT): Remove.
17981 (CC_IN_FPU): Remove.
17982 (NOTICE_UPDATE_CC): Remove.
17983 (REGISTER_NAMES): Add CC registers.
17984 (BRANCH_COST): Change to constant 1.
17985 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
17987 * config/pdp11/pdp11.opt (mbcopy): Remove.
17988 (mbcopy-builtin): Remove.
17989 (mbranch-cheap): Remove.
17990 (mbranch-expensive): Remove.
17991 * config/pdp11/predicates.md (expand_shift_operand): Update to
17992 match shift code generation.
17993 (ccnz_operator): New predicate.
17994 * doc/invoke.texi (PDP-11 Options): Remove deleted options
17995 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
17996 Remove non-existent option -mabshi, -mno-abshi. Document mutually
17998 * doc/md.texi (PDP-11): Document new D and h constraints. Update
17999 description of O constraint.
18001 2018-06-27 Jeff Law <law@redhat.com>
18002 Austin Law <austinklaw@gmail.com>
18004 * config/v850/v850.md (addsi3_set_flags): New pattern.
18005 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
18006 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
18007 (zero_extendhisi2_v850_set_flags): Likewise.
18008 (zero_extendqisi2_v850_set_flags): Likewise.
18009 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
18010 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
18011 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
18013 * config/v850/v850-protos.h (notice_update_cc): Remove.
18014 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
18015 (v850_print_operand): Handle 'D' and "d".
18016 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
18017 Add handling of arithmetic/logical operations compared against zero.
18018 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
18019 Do not look at v850_compare_op, instead get mode from last argument.
18020 (v850_gen_compare): Remove
18021 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
18022 after reload for prologue insns.
18023 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
18025 (construct_save_jarl): Likewise.
18026 (TARGET_FLAGS_REGNUM): Define.
18027 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
18028 (NOTICE_UPDATE_CC): Remove.
18029 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
18030 than cc0. Conditionalize on reload_completed.
18031 (cmpsi_insn, setfcc_insn): Likewise.
18032 (tst1 splitter): Turn into define_and_split which sets the flags
18034 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
18035 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
18036 (cstoresf4, cstoredf4): Clobber the flags.
18037 (cmpsi, cmpsf, cmpdf): Remove expanders.
18038 (setf_insn): Remove pattern.
18039 (addsi3): Turn into define_and_split which clobbers the flags after
18040 reload and a suitable pattern (addsi3_clobber_flags) for use after
18042 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
18043 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
18044 (ashrsi3, ashrsi3_v850e2): Likewise.
18045 (bins): Clobber the flags.
18046 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
18047 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
18048 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
18049 (call_value_internal_short, call_value_internal_long): Likewise.
18050 (callt_save_interrupt, callt_return_interrupt): Likewise.
18051 (save_interrupt, return_interrupt): Likewise.
18052 (callt_save_all_interrupt, save_all_interrupt): Likewise.
18053 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
18054 (restore_all_interrupt, _restore_all_interrupt): Likewise.
18055 (All FP comparisons): Only allow after reload has completed.
18057 (divh, divhu): Tweak output template.
18058 (branch_z_normal, branch_z_invert): Remove
18059 (branch_nz_normal, branch_nz_invert): Likewise.
18060 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
18062 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
18063 * config/v850/v850.c (notice_update_cc): Remove.
18064 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
18065 (CC_NO_CARRY): Likewise.
18066 (NOTICE_UPDATE_CC): Define to nothing.
18067 * config/v850/v850.md: Remove block comment on cc0 handling
18068 Remove "cc" attribute from all patterns. Remove cc_status handling
18069 from all patterns. Minor formatting fixes.
18071 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18073 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
18074 (cortex-a76.cortex-a55): Likewise.
18075 * config/aarch64/aarch64-tune.md: Regenerate.
18076 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
18077 cortex-a76.cortex-a55.
18079 2018-06-27 Jeff Law <law@redhat.com>
18081 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
18082 (MULTILIB_DIRNAMES): Similarly.
18084 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
18086 * gimple.h (gimple_return_retbnd): Delete.
18087 (gimple_return_set_retbnd): Likewise.
18088 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
18089 gimple_return_set_retbnd.
18090 * gimple-pretty-print.c (dump_gimple_return): Remove call to
18091 gimple_return_retbnd and adjust.
18092 * tree-inline.h (struct copy_body_data): Remove retbnd field.
18093 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
18094 Explicitly return NULL in a couple more cases. Move assertion
18095 on debug statements and remove unreachable code.
18096 (reset_debug_binding): Do not test id->retbnd.
18097 (expand_call_inline): Do not set it.
18099 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
18101 * configure.ac: Add --disable-gcov option.
18102 * configure: Regenerate.
18103 * Makefile.in: Honour @enable_gcov@.
18104 * doc/install.texi: Document --disable-gcov.
18106 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18108 * config/arm/arm-cpus.in (cortex-a76): New entry.
18109 (cortex-a76.cortex-a55): Likewise.
18110 * config/arm/arm-tables.opt: Regenerate.
18111 * config/arm/arm-tune.md: Likewise.
18112 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
18113 * doc/invoke.texi (ARM Options): Document cortex-a76 and
18114 cortex-a76.cortex-a55.
18116 2018-06-27 Tamar Christina <tamar.christina@arm.com>
18119 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
18121 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
18123 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
18125 (EPILOGUE_USES): Likewise.
18127 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
18129 * tree-inline.c (remap_location): New function extracted from...
18130 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
18131 (copy_phis_for_bb): ...here. Call remap_location.
18132 (copy_cfg_body): Adjust call to copy_edges_for_bb.
18134 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
18136 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
18137 unaligned vsx for 16B memset.
18139 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
18142 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
18143 ieee128_float_type_node to long_double_type_node unless
18144 TARGET_LONG_DOUBLE_128 is set.
18146 2018-06-26 David Malcolm <dmalcolm@redhat.com>
18148 * cfgloop.c (get_loop_location): Convert return type from
18149 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
18150 by implicit construction from rtx_insn *, and using
18151 dump_user_location_t::from_function_decl for the fallback case.
18152 * cfgloop.h (get_loop_location): Convert return type from
18153 location_t to dump_user_location_t.
18154 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
18155 dump_printf_loc to pass in a dump_location_t rather than a
18156 location_t, via the gimple stmt.
18157 * coverage.c (get_coverage_counts): Update calls to
18158 dump_printf_loc to pass in dump_location_t rather than a
18160 * doc/optinfo.texi (Dump types): Convert example of
18161 dump_printf_loc from taking "locus" to taking "insn". Update
18162 description of the "_loc" calls to cover dump_location_t.
18163 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
18165 (dump_user_location_t::dump_user_location_t): New constructors,
18166 from gimple *stmt and rtx_insn *.
18167 (dump_user_location_t::from_function_decl): New function.
18168 (dump_loc): Make static.
18169 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
18170 const dump_location_t &.
18171 (dump_generic_expr_loc): Delete.
18172 (dump_printf_loc): Convert param "loc" from location_t to
18173 const dump_location_t &.
18174 (selftest::test_impl_location): New function.
18175 (selftest::dumpfile_c_tests): New function.
18176 * dumpfile.h: Include "profile-count.h".
18177 (class dump_user_location_t): New class.
18178 (struct dump_impl_location_t): New struct.
18179 (class dump_location_t): New class.
18180 (dump_printf_loc): Convert 2nd param from source_location to
18181 const dump_location_t &.
18182 (dump_generic_expr_loc): Delete.
18183 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
18184 const dump_location_t &.
18185 * gimple-fold.c (fold_gimple_assign): Update call to
18186 dump_printf_loc to pass in a dump_location_t rather than a
18187 location_t, via the gimple stmt.
18188 (gimple_fold_call): Likewise.
18189 * gimple-loop-interchange.cc
18190 (loop_cand::analyze_iloop_reduction_var): Update for change to
18191 check_reduction_path.
18192 (tree_loop_interchange::interchange): Update for change to
18193 find_loop_location.
18194 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
18195 change in return-type of find_loop_location.
18196 (graphite_regenerate_ast_isl): Likewise.
18197 * graphite-optimize-isl.c (optimize_isl): Likewise.
18198 * graphite.c (graphite_transform_loops): Likewise.
18199 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
18200 pass in a dump_location_t rather than a location_t, via the
18202 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
18203 * ipa.c (walk_polymorphic_call_targets): Likewise.
18204 * loop-unroll.c (report_unroll): Convert "locus" param from
18205 location_t to dump_location_t.
18206 (decide_unrolling): Update for change to get_loop_location's
18208 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
18209 location_t to dump_user_location_t.
18210 (grid_find_single_omp_among_assignments_1): Updates calls to
18211 dump_printf_loc to pass in a dump_location_t rather than a
18212 location_t, via the gimple stmt.
18213 (grid_parallel_clauses_gridifiable): Convert "tloc" from
18214 location_t to dump_location_t. Updates calls to dump_printf_loc
18215 to pass in a dump_location_t rather than a location_t, via the
18217 (grid_inner_loop_gridifiable_p): Likewise.
18218 (grid_dist_follows_simple_pattern): Likewise.
18219 (grid_gfor_follows_tiling_pattern): Likewise.
18220 (grid_target_follows_gridifiable_pattern): Likewise.
18221 (grid_attempt_target_gridification): Convert initialization
18222 of local "grid" from memset to zero-initialization; FIXME: does
18223 this require C++11? Update call to dump_printf_loc to pass in a
18224 optinfo_location rather than a location_t, via the gimple stmt.
18225 * profile.c (read_profile_edge_counts): Updates call to
18226 dump_printf_loc to pass in a dump_location_t rather than a
18228 (compute_branch_probabilities): Likewise.
18229 * selftest-run-tests.c (selftest::run_tests): Call
18231 * selftest.h (dumpfile_c_tests): New decl.
18232 * tree-loop-distribution.c (pass_loop_distribution::execute):
18233 Update for change in return type of find_loop_location.
18234 * tree-parloops.c (parallelize_loops): Likewise.
18235 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
18236 "locus" from location_t to dump_user_location_t.
18237 (canonicalize_loop_induction_variables): Likewise.
18238 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
18239 for change in return type of find_loop_location.
18240 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
18241 to dump_printf_loc to pass in a dump_location_t rather than a
18242 location_t, via the stmt.
18243 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
18245 * tree-vect-loop-manip.c (find_loop_location): Convert return
18246 type from source_location to dump_user_location_t.
18247 (vect_do_peeling): Update for above change.
18248 (vect_loop_versioning): Update for change in type of
18250 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
18251 from location_t to dump_user_location_t.
18252 (vect_estimate_min_profitable_iters): Update for change in type
18254 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
18255 location_t to dump_location_t.
18256 (vect_slp_bb): Update for change in type of vect_location.
18257 * tree-vectorizer.c (vect_location): Convert from source_location
18258 to dump_user_location_t.
18259 (try_vectorize_loop_1): Update for change in vect_location's type.
18260 (vectorize_loops): Likewise.
18261 (increase_alignment): Likewise.
18262 * tree-vectorizer.h (vect_location): Convert from source_location
18263 to dump_user_location_t.
18264 (find_loop_location): Convert return type from source_location to
18265 dump_user_location_t.
18266 (check_reduction_path): Convert 1st param from location_t to
18267 dump_user_location_t.
18268 * value-prof.c (check_counter): Update call to dump_printf_loc to
18269 pass in a dump_user_location_t rather than a location_t; update
18270 call to error_at for change in type of "locus".
18271 (check_ic_target): Update call to dump_printf_loc to
18272 pass in a dump_user_location_t rather than a location_t, via the
18275 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
18277 * config/s390/s390.h (enum processor_flags): Do not use
18278 default tune parameter when -march was specified.
18280 2018-06-26 Jakub Jelinek <jakub@redhat.com>
18283 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
18284 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
18287 2018-06-26 Richard Biener <rguenther@suse.de>
18289 PR tree-optimization/86287
18291 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
18292 (vect_analyze_loop): Initialize n_stmts.
18294 2018-06-26 Richard Biener <rguenther@suse.de>
18296 PR middle-end/86271
18297 * fold-const.c (fold_convertible_p): Pointer extension
18300 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
18303 * dwarf2out.c (loc_list_has_views): Adjust comments.
18304 (dw_loc_list): Split single cross-partition range with
18307 2018-06-25 Jeff Law <law@redhat.com>
18309 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
18310 on -mbig-switch by default.
18312 * config/v850/predicates.md (const_float_1_operand): Fix match_code
18314 (const_float_0_operand): Remove unused predicate.
18315 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
18316 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
18317 (recipsf2): New expander. Original pattern now called
18319 (recipdf2, recipdf2_insn): Similarly.
18320 (rsqrtsf2, rsqrtsf2_insn): Similarly
18321 (rsqrtdf2, rsqrtdf2_insn): Similarly
18323 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
18325 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
18326 Simplify logic for FreeBSD (twice).
18328 2018-06-25 Martin Sebor <msebor@redhat.com>
18330 PR tree-optimization/86204
18331 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
18332 a strnlen result if it's less than the length of the string.
18334 2018-06-25 Martin Sebor <msebor@redhat.com>
18336 PR tree-optimization/85700
18337 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
18338 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
18339 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
18341 2018-06-25 Martin Sebor <msebor@redhat.com>
18343 * doc/extend.texi (Zero-length arrays): Update and clarify.
18345 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
18347 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
18348 added IEEE/IBM long double multilib support on PowerPC little
18349 endian Linux systems.
18350 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
18351 (MULTILIB_DEFAULTS): Likewise.
18352 * config/rs6000/rs6000.c (rs6000_option_override_internal):
18354 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
18355 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
18356 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
18358 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
18360 PR middle-end/86311
18361 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
18362 (REORDER_45): Likewise.
18364 2018-06-25 Jeff Law <law@redhat.com>
18366 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
18367 dividend to 32 bits. Adjust length.
18368 (udivmodhi4): Cleanup output template. Fix length.
18370 2018-06-25 Carl Love <cel@us.ibm.com>
18372 * config/rs6000/vsx.md: Change word selector to prefered location.
18374 2018-06-25 Richard Biener <rguenther@suse.de>
18376 PR tree-optimization/86304
18377 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
18378 epilogue-if-converted loops as well.
18380 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
18382 * lto-section-out.c (lto_begin_section): Do not print section
18383 name for noaddr and unnumbered dumps.
18385 2018-06-25 Richard Biener <rguenther@suse.de>
18387 * tree-vectorizer.h (struct vec_info_shared): New structure
18388 with parts split out from struct vec_info and loop_nest from
18389 struct _loop_vec_info.
18390 (struct vec_info): Adjust accordingly.
18391 (struct _loop_vec_info): Likewise.
18392 (LOOP_VINFO_LOOP_NEST): Adjust.
18393 (LOOP_VINFO_DATAREFS): Likewise.
18394 (LOOP_VINFO_DDRS): Likewise.
18395 (struct _bb_vec_info): Likewise.
18396 (BB_VINFO_DATAREFS): Likewise.
18397 (BB_VINFO_DDRS): Likewise.
18398 (struct _stmt_vec_info): Add dr_aux member.
18399 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
18400 (DR_MISALIGNMENT_UNINITIALIZED): New.
18401 (set_dr_misalignment): Adjust.
18402 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
18403 (vect_analyze_loop): Adjust prototype.
18404 (vect_analyze_loop_form): Likewise.
18405 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
18406 Compute dependences lazily.
18407 (vect_record_base_alignments): Use shared datarefs/ddrs.
18408 (vect_verify_datarefs_alignment): Likewise.
18409 (vect_analyze_data_refs_alignment): Likewise.
18410 (vect_analyze_data_ref_accesses): Likewise.
18411 (vect_analyze_data_refs): Likewise.
18412 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
18413 constructor parameter for shared part.
18414 (vect_analyze_loop_form): Pass in shared part and adjust.
18415 (vect_analyze_loop_2): Pass in storage for the number of
18416 stmts. Move loop nest finding to the caller. Compute
18418 (vect_analyze_loop): Pass in shared part.
18419 (vect_transform_loop): Verify shared datarefs are unchanged.
18420 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
18421 constructor parameter for shared part.
18422 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
18423 (vect_slp_bb): Verify shared datarefs are unchanged before
18425 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
18427 (new_stmt_vec_info): Initialize DR_AUX misalignment to
18428 DR_MISALIGNMENT_UNINITIALIZED.
18429 * tree-vectorizer.c (vec_info::vec_info): Add constructor
18430 parameter for shared part.
18431 (vec_info::~vec_info): Adjust.
18432 (vec_info_shared::vec_info_shared): New.
18433 (vec_info_shared::~vec_info_shared): Likewise.
18434 (vec_info_shared::save_datarefs): Likewise.
18435 (vec_info_shared::check_datarefs): Likewise.
18436 (try_vectorize_loop_1): Construct shared part live for analyses
18437 of a single loop for multiple vector sizes.
18438 * tree-parloops.c (gather_scalar_reductions): Adjust.
18440 2018-06-25 Richard Biener <rguenther@suse.de>
18442 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
18443 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
18444 (vect_analyze_data_refs): Remove similar code from here and
18445 simplify accordingly.
18447 2018-06-25 Richard Biener <rguenther@suse.de>
18449 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
18450 for reverse storage order accesses rather than asserting
18451 they cannot happen here.
18453 2018-06-25 Tom de Vries <tdevries@suse.de>
18456 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
18457 Use data16 instead of .byte for insn prefix.
18459 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
18462 * parser.c (make_char_string_pack): Pass this literal chars
18463 through cpp_interpret_string.
18464 (cp_parser_userdef_numeric_literal): Check the result of
18465 make_char_string_pack.
18467 2018-06-24 Maya Rashish <coypu@sdf.org>
18469 * ginclude/stddef.h: Simplify conditions around avoiding
18470 re-definition of __size_t.
18472 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
18474 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
18475 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
18477 2018-06-22 Maya Rashish <coypu@sdf.org>
18479 * doc/invoke.texi (mno-fancy-math-387): Update for changes
18480 made to OpenBSD and NetBSD through the years.
18482 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18484 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
18485 behavior of vec_pack (vector double, vector double) to match
18486 behavior of vec_float2 (vector double, vector double).
18488 2018-06-22 Olivier Hainque <hainque@adacore.com>
18490 * gimplify.c (gimplify_function_tree): Prevent creation
18491 of a trampoline for the address of the current function
18492 passed to entry/exit instrumentation hooks.
18494 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
18497 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
18500 2018-06-22 Martin Liska <mliska@suse.cz>
18502 PR tree-optimization/86263
18503 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18504 Bail out if is_enabled is false.
18505 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
18507 (jump_table_cluster::is_enabled): New function.
18509 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
18511 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
18512 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
18513 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
18514 (lto_input_ts_binfo_tree_pointers): Likewise.
18515 * tree-streamer-out.c (streamer_write_tree_bitfields,
18516 write_ts_binfo_tree_pointers): Likewise.
18517 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
18519 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
18521 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
18523 2018-06-22 Martin Liska <mliska@suse.cz>
18525 * symbol-summary.h (get): Make it pure and inline move
18526 functionality from ::get function.
18527 (get): Remove and inline into ::get and ::get_create.
18528 (get_create): Move code from ::get function.
18530 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18533 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
18534 -x assembler-with-cpp.
18536 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18538 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
18539 _FILE_OFFSET_BITS=64 for C++.
18541 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
18543 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
18544 conversion insn that shows up when pr85657-3.c is compiled using
18545 IEEE 128-bit long double.
18546 (neg<mode>2_internal): Use the correct mode to check whether the
18547 mode is IBM extended.
18548 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
18549 multiply and divide external functions from being created more
18552 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18554 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
18556 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
18557 the edge can be forwarded.
18558 (cfg_layout_merge_blocks): Likewise.
18560 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18562 * except.c (finish_eh_generation): Commit edge insertions only after
18563 the EH edges have been redirected from post-landing to landing pads.
18565 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18567 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
18568 create_tmp_var_for to create the FRAME decl.
18569 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
18571 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18573 * tree-inline.c (copy_edges_for_bb): Minor tweak.
18574 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
18575 debug statement when resetting its value.
18576 (expand_call_inline): Copy the locus of the call onto the assignment
18577 of the return value, if any. Use local variable in more cases.
18579 2018-06-21 Martin Liska <mliska@suse.cz>
18581 * ipa-pure-const.c (propagate_nothrow): Use
18582 funct_state_summaries->get.
18583 (dump_malloc_lattice): Likewise.
18584 (propagate_malloc): Likewise.
18586 2018-06-21 Richard Biener <rguenther@suse.de>
18588 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
18589 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
18590 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
18592 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
18593 BLOCK_ABSTRACT_ORIGIN unconditionally.
18595 2018-06-21 David Malcolm <dmalcolm@redhat.com>
18597 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
18599 * ipa-reference.c (ipa_reference_c_finalize): Delete
18600 ipa_ref_opt_sum_summaries and set it to NULL.
18602 2018-06-21 Tom de Vries <tdevries@suse.de>
18604 PR tree-optimization/85859
18605 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
18606 test with comment from bb_no_side_effects_p.
18608 2018-06-21 Richard Biener <rguenther@suse.de>
18610 PR tree-optimization/86232
18611 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
18612 max for constant niter.
18614 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
18616 * config/aarch64/aarch64-simd.md
18617 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
18619 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
18621 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
18622 Make opernads of the unspec commutative.
18624 2018-06-21 Richard Biener <rguenther@suse.de>
18626 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
18627 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
18628 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
18629 (vect_analyze_data_ref_dependence): Re-order checks to deal with
18631 (vect_record_base_alignments): Do not record base alignment
18632 for gathers or scatters.
18633 (vect_compute_data_ref_alignment): Drop return value that is always
18634 true. Bail out early for gathers or scatters.
18635 (vect_enhance_data_refs_alignment): Bail out early for gathers
18637 (vect_find_same_alignment_drs): Likewise.
18638 (vect_analyze_data_refs_alignment): Remove dead code.
18639 (vect_slp_analyze_and_verify_node_alignment): Likewise.
18640 (vect_analyze_data_refs): For possible gathers or scatters do
18641 not create an alternate DR, just check their possible validity
18642 and mark them. Adjust DECL_NONALIASED handling to not rely
18643 on DR_BASE_ADDRESS.
18644 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
18645 update inits of gathers or scatters.
18646 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
18647 Also copy gather/scatter flag to pattern vinfo.
18649 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
18651 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
18652 behavior of vec_packsu (vector unsigned long long, vector unsigned
18653 long long) to match behavior of vec_packs with same signature.
18655 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
18656 Thomas Schwinge <thomas@codesourcery.com>
18657 Cesar Philippidis <cesar@codesourcery.com>
18659 * gimplify.c (gimplify_scan_omp_clauses): Add support for
18660 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18661 (gimplify_adjust_omp_clauses): Likewise.
18662 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
18663 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
18664 (gimplify_omp_target_update): Update handling of acc update and
18666 * omp-low.c (install_var_field): Remove unused parameter
18667 base_pointers_restrict.
18668 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
18669 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
18671 (omp_target_base_pointers_restrict_p): Delete.
18672 (scan_omp_target): Update call to scan_sharing_clauses.
18673 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
18675 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
18676 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18677 (convert_local_omp_clauses): Likewise.
18678 * tree-pretty-print.c (dump_omp_clause): Likewise.
18679 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
18681 (omp_clause_code_name): Likewise.
18683 2018-06-20 Jakub Jelinek <jakub@redhat.com>
18686 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
18689 PR tree-optimization/86231
18690 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
18691 anti-range don't overwrite *vr0min before using it to compute *vr0max.
18693 2018-06-20 Tom de Vries <tdevries@suse.de>
18695 PR tree-optimization/86097
18696 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
18697 iv type if signedness of iv type is not the same as that of *nit.
18699 2018-06-20 Jakub Jelinek <jakub@redhat.com>
18701 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
18702 EDGE_EH edges, verify they are all EDGE_EH.
18704 2018-06-20 Maya Rashish <coypu@sdf.org>
18706 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
18708 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18710 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
18711 * config/aarch64/aarch64.c (xgene1_tunings): Add
18712 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
18713 (aarch64_mode_valid_for_sched_fusion_p):
18714 Allow 16-byte modes.
18715 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
18716 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
18718 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
18720 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
18721 * config/aarch64/iterators.md (VQ2): New mode iterator.
18723 2018-06-20 Martin Liska <mliska@suse.cz>
18725 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
18726 Change default ratio from 10 to 8.
18728 2018-06-20 Martin Liska <mliska@suse.cz>
18730 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18732 (bit_test_cluster::find_bit_tests): Likewise.
18733 (switch_decision_tree::analyze_switch_statement): Find clusters.
18734 * tree-switch-conversion.h (struct jump_table_cluster): Document
18737 2018-06-20 Martin Liska <mliska@suse.cz>
18739 * tree-switch-conversion.c (switch_conversion::collect):
18740 Record m_uniq property.
18741 (switch_conversion::expand): Bail out for special conditions.
18742 (group_cluster::~group_cluster): New.
18743 (group_cluster::group_cluster): Likewise.
18744 (group_cluster::dump): Likewise.
18745 (jump_table_cluster::emit): New.
18746 (switch_decision_tree::fix_phi_operands_for_edges): New.
18747 (struct case_node): Remove struct.
18748 (jump_table_cluster::can_be_handled): New.
18749 (case_values_threshold): Moved to header.
18750 (reset_out_edges_aux): Likewise.
18751 (jump_table_cluster::is_beneficial): New.
18752 (bit_test_cluster::can_be_handled): Likewise.
18753 (add_case_node): Remove.
18754 (bit_test_cluster::is_beneficial): New.
18755 (case_bit_test::cmp): New.
18756 (bit_test_cluster::emit): New.
18757 (expand_switch_as_decision_tree_p): Remove.
18758 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
18759 (fix_phi_operands_for_edge): Likewise.
18760 (switch_decision_tree::analyze_switch_statement): New.
18761 (compute_cases_per_edge): Move ...
18762 (switch_decision_tree::compute_cases_per_edge): ... here.
18763 (try_switch_expansion): Likewise.
18764 (switch_decision_tree::try_switch_expansion): Likewise.
18765 (record_phi_operand_mapping): Likewise.
18766 (switch_decision_tree::record_phi_operand_mapping): Likewise.
18767 (emit_case_decision_tree): Likewise.
18768 (switch_decision_tree::emit): Likewise.
18769 (balance_case_nodes): Likewise.
18770 (switch_decision_tree::balance_case_nodes): Likewise.
18771 (dump_case_nodes): Likewise.
18772 (switch_decision_tree::dump_case_nodes): Likewise.
18773 (emit_jump): Likewise.
18774 (switch_decision_tree::emit_jump): Likewise.
18775 (emit_cmp_and_jump_insns): Likewise.
18776 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
18777 (emit_case_nodes): Likewise.
18778 (switch_decision_tree::emit_case_nodes): Likewise.
18779 (conditional_probability): Remove.
18780 * tree-switch-conversion.h (enum cluster_type): New.
18782 (struct cluster): Likewise.
18783 (cluster::cluster): Likewise.
18784 (struct simple_cluster): Likewise.
18785 (simple_cluster::simple_cluster): Likewise.
18786 (struct group_cluster): Likewise.
18787 (struct jump_table_cluster): Likewise.
18788 (struct bit_test_cluster): Likewise.
18789 (struct min_cluster_item): Likewise.
18790 (struct case_tree_node): Likewise.
18791 (case_tree_node::case_tree_node): Likewise.
18792 (jump_table_cluster::case_values_threshold): Likewise.
18793 (struct case_bit_test): Likewise.
18794 (struct switch_decision_tree): Likewise.
18795 (struct switch_conversion): Likewise.
18796 (switch_decision_tree::reset_out_edges_aux): Likewise.
18798 2018-06-20 Martin Liska <mliska@suse.cz>
18800 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
18801 (hoist_edge_and_branch_if_true): Likewise.
18802 (expand_switch_using_bit_tests_p): Likewise.
18803 (struct case_bit_test): Likewise.
18804 (case_bit_test_cmp): Likewise.
18805 (emit_case_bit_tests): Likewise.
18806 (switch_conversion::switch_conversion): New class.
18807 (struct switch_conv_info): Remove old struct.
18808 (collect_switch_conv_info): More to ...
18809 (switch_conversion::collect): ... this.
18810 (check_range): Likewise.
18811 (switch_conversion::check_range): Likewise.
18812 (check_all_empty_except_final): Likewise.
18813 (switch_conversion::check_all_empty_except_final): Likewise.
18814 (check_final_bb): Likewise.
18815 (switch_conversion::check_final_bb): Likewise.
18816 (create_temp_arrays): Likewise.
18817 (switch_conversion::create_temp_arrays): Likewise.
18818 (free_temp_arrays): Likewise.
18819 (gather_default_values): Likewise.
18820 (switch_conversion::gather_default_values): Likewise.
18821 (build_constructors): Likewise.
18822 (switch_conversion::build_constructors): Likewise.
18823 (constructor_contains_same_values_p): Likewise.
18824 (switch_conversion::contains_same_values_p): Likewise.
18825 (array_value_type): Likewise.
18826 (switch_conversion::array_value_type): Likewise.
18827 (build_one_array): Likewise.
18828 (switch_conversion::build_one_array): Likewise.
18829 (build_arrays): Likewise.
18830 (switch_conversion::build_arrays): Likewise.
18831 (gen_def_assigns): Likewise.
18832 (switch_conversion::gen_def_assigns): Likewise.
18833 (prune_bbs): Likewise.
18834 (switch_conversion::prune_bbs): Likewise.
18835 (fix_phi_nodes): Likewise.
18836 (switch_conversion::fix_phi_nodes): Likewise.
18837 (gen_inbound_check): Likewise.
18838 (switch_conversion::gen_inbound_check): Likewise.
18839 (process_switch): Use the newly created class.
18840 (switch_conversion::expand): New.
18841 (switch_conversion::~switch_conversion): New.
18842 * tree-switch-conversion.h: New file.
18844 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18846 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
18847 tree-vect-patterns.c.
18848 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
18849 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
18850 (vect_recog_sad_pattern): Likewise.
18851 (vect_recog_widen_sum_pattern): Likewise.
18852 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
18853 (vect_recog_widen_shift_pattern): Remove the type_in argument.
18854 (vect_recog_rotate_pattern): Likewise.
18855 (vect_recog_mult_pattern): Likewise.
18856 (vect_recog_vector_vector_shift_pattern): Likewise.
18857 (vect_recog_divmod_pattern): Likewise.
18858 (vect_recog_mixed_size_cond_pattern): Likewise.
18859 (vect_recog_bool_pattern): Likewise.
18860 (vect_recog_mask_conversion_pattern): Likewise.
18861 (vect_try_gather_scatter_pattern): Likewise.
18862 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
18863 (vect_recog_over_widening_pattern): Likewise.
18864 (vect_recog_gather_scatter_pattern): Likewise.
18865 (vect_recog_func_ptr): Move from tree-vectorizer.h
18866 (vect_vect_recog_func_ptrs): Move further down the file.
18867 (vect_recog_func): Likewise. Remove the third argument.
18868 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
18869 (vect_pattern_recog_1): Expect the pattern function to do any
18870 necessary target tests. Also expect it to provide a vector type.
18871 Remove the type_in handling.
18873 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18875 * tree-vect-patterns.c (vect_pattern_detected): New function.
18876 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
18877 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
18878 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
18879 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
18880 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
18881 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
18882 (vect_recog_mask_conversion_pattern)
18883 (vect_try_gather_scatter_pattern): Likewise.
18885 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18887 * tree-vect-patterns.c (vect_get_internal_def): New function.
18888 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
18889 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
18890 (search_type_for_mask_1): Use it.
18892 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18894 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
18895 redundant WIDEN_SUM_EXPR handling.
18896 (vect_recog_sad_pattern): Likewise.
18898 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18900 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
18901 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
18902 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
18903 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
18904 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
18906 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18908 * tree-vect-stmts.c (vectorizable_call): Make sure that we
18909 use the stmt_vec_info of the original bb statement for the
18910 new zero assignment, even if the call is part of a pattern.
18912 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18914 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
18915 that the sequence is attached to the original statement rather
18916 than the pattern statement.
18917 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
18918 PATTERN_DEF_SEQ from the original statement rather than
18919 the main pattern statement.
18920 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
18921 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
18922 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
18924 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18926 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
18927 definition statements before the early exit for statements that aren't
18929 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
18931 (vect_transform_loop): ...here. Process pattern definition
18932 statements without first checking whether the main pattern
18933 statement is live or relevant.
18935 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
18937 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
18938 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
18940 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
18942 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
18943 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
18944 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
18945 (expand_block_compare): Change select_block_compare_mode call.
18946 (expand_strncmp_align_check): Use new functions, fix comment.
18947 (emit_final_str_compare_gpr): New function.
18948 (expand_strn_compare): Refactor and clean up code.
18949 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
18951 2018-06-19 Tony Reix <tony.reix@atos.com>
18952 Damien Bergamini <damien.bergamini@atos.com>
18953 David Edelsohn <dje.gcc@gmail.com>
18955 * collect2.c (static_obj): New variable.
18956 (static_libs): New variable.
18957 (is_in_list): Uncomment declaration.
18958 (main): Track AIX libraries linked statically.
18959 (is_in_list): Uncomment definition.
18960 (scan_prog_file): Don't add AIX shared libraries initializer
18961 to constructor list if linking statically.
18963 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
18965 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
18967 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
18969 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
18971 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
18974 2018-06-19 Martin Liska <mliska@suse.cz>
18976 * config/i386/i386.c (ix86_can_inline_p): Do not use
18977 ipa_fn_summaries::get_create.
18978 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
18980 (devirtualization_time_bonus): Likewise.
18981 (ipcp_propagate_stage): Likewise.
18982 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
18983 (edge_set_predicate): Likewise.
18984 (evaluate_conditions_for_known_args): Likewise.
18985 (evaluate_properties_for_edge): Likewise.
18986 (ipa_call_summary::reset): Tranform to ...
18987 (ipa_call_summary::~ipa_call_summary): ... this.
18988 (ipa_fn_summary::reset): Transform to ...
18989 (ipa_fn_summary::~ipa_fn_summary): ... this.
18990 (ipa_fn_summary_t::remove): Rename to ...
18991 (ipa_fn_summary_t::remove_callees): ... this.
18992 (ipa_fn_summary_t::duplicate): Use placement new
18993 instead of memory copy.
18994 (ipa_call_summary_t::duplicate): Likewise.
18995 (ipa_call_summary_t::remove): Remove.
18996 (dump_ipa_call_summary): Change get_create to get.
18997 (ipa_dump_fn_summary): Dump only when summary exists.
18998 (analyze_function_body): Use symbol_summary::get instead
19000 (compute_fn_summary): Likewise.
19001 (estimate_edge_devirt_benefit): Likewise.
19002 (estimate_edge_size_and_time): Likewise.
19003 (inline_update_callee_summaries): Likewise.
19004 (remap_edge_change_prob): Likewise.
19005 (remap_edge_summaries): Likewise.
19006 (ipa_merge_fn_summary_after_inlining): Likewise.
19007 (write_ipa_call_summary): Likewise.
19008 (ipa_fn_summary_write): Likewise.
19009 (ipa_free_fn_summary): Likewise.
19010 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
19011 (struct ipa_call_summary): Likewise.
19012 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
19014 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
19015 (estimate_size_after_inlining): Likewise.
19016 (estimate_growth): Likewise.
19017 (growth_likely_positive): Likewise.
19018 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
19019 (inline_call): Likewise.
19020 * ipa-inline.c (caller_growth_limits): Likewise.
19021 (can_inline_edge_p): Likewise.
19022 (can_inline_edge_by_limits_p): Likewise.
19023 (compute_uninlined_call_time): Likewise.
19024 (compute_inlined_call_time): Likewise.
19025 (want_inline_small_function_p): Likewise.
19026 (edge_badness): Likewise.
19027 (update_caller_keys): Likewise.
19028 (update_callee_keys): Likewise.
19029 (inline_small_functions): Likewise.
19030 (inline_to_all_callers_1): Likewise.
19031 (dump_overall_stats): Likewise.
19032 (early_inline_small_functions): Likewise.
19033 (early_inliner): Likewise.
19034 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
19035 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
19036 * ipa-pure-const.c (malloc_candidate_p): Likewise.
19037 * ipa-split.c (execute_split_functions): Likewise.
19038 * symbol-summary.h: Likewise.
19039 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
19041 2018-06-19 Richard Biener <rguenther@suse.de>
19043 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
19044 (vectorize_loops): ... here. Fix dbgcnt handling.
19045 (try_vectorize_loop): Wrap try_vectorize_loop_1.
19047 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
19050 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
19051 ieee128 argument takes up only one (vector) register, not two (floating
19054 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
19056 * gimplify.c (gimplify_init_constructor): Really never clear for an
19057 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
19059 2018-06-19 Richard Biener <rguenther@suse.de>
19061 PR tree-optimization/86179
19062 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
19063 after failed recognition.
19065 2018-06-18 Martin Sebor <msebor@redhat.com>
19067 PR middle-end/85602
19068 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
19069 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
19070 Handle integer subtraction.
19071 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
19072 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
19074 2018-06-18 David Malcolm <dmalcolm@redhat.com>
19076 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
19077 param from rtx to rtx_insn *.
19078 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
19080 (frv_ifcvt_modify_insn): Likwise.
19081 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
19082 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
19083 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
19084 as_a <rtx_insn *> cast to local "unprotected_region" once
19085 it's been established that it's not NULL or pc_rtx.
19086 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
19087 param "sethi" from rtx to rtx_insn *.
19088 (nds32_group_float_insns): Likewise for param "insn".
19089 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
19091 (vax_output_int_subtract): Likewise.
19092 * config/vax/vax.c (vax_output_int_add): Likewise for param
19094 (vax_output_int_subtract): Likewise.
19095 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
19096 (emit_pattern_after): Likewise for param "after".
19097 (emit_insn_after): Likewise.
19098 (emit_jump_insn_after): Likewise.
19099 (emit_call_insn_after): Likewise.
19100 (emit_debug_insn_after): Likewise.
19101 (emit_pattern_before): Likewise for param "before".
19102 (emit_insn_before): Likewise.
19103 (emit_jump_insn_before): Likewise.
19104 * final.c (get_insn_template): Likewise for param "insn", removing
19106 * output.h (get_insn_template): Likewise for 2nd param.
19107 * rtl.h (emit_insn_before): Likewise.
19108 (emit_jump_insn_before): Likewise.
19109 (emit_debug_insn_before_noloc): Likewise.
19110 (emit_insn_after): Likewise.
19111 (emit_jump_insn_after): Likewise.
19112 (emit_call_insn_after): Likewise.
19113 (emit_debug_insn_after): Likewise.
19114 (set_insn_deleted): Likewise for param.
19116 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
19119 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
19120 floating point modes, so that IFmode is numerically greater than
19121 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
19122 to declare the ordering. This prevents IFmode from being
19123 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
19124 machine. Include rs6000-modes.h to share the fractional values
19125 between genmodes* and the rest of the compiler.
19126 (IFmode): Likewise.
19127 (KFmode): Likewise.
19128 (TFmode): Likewise.
19129 * config/rs6000/rs6000-modes.h: New file.
19130 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
19131 meaning of rs6000_long_double_size so that 126..128 selects an
19132 appropriate 128-bit floating point type.
19133 (rs6000_option_override_internal): Likewise.
19134 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
19135 (TARGET_LONG_DOUBLE_128): Change the meaning of
19136 rs6000_long_double_size so that 126..128 selects an appropriate
19137 128-bit floating point type.
19138 (LONG_DOUBLE_TYPE_SIZE): Update comment.
19139 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
19140 source and destination to match the standard usage.
19141 (truncifkf2): Likewise.
19142 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
19143 ISA 2.07 to use an explicit clobber, instead of passing in a
19145 (copysign<mode>3_soft): Likewise.
19147 2018-06-18 David Malcolm <dmalcolm@redhat.com>
19149 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
19150 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
19151 (vect_slp_analyze_instance_dependence): Likewise.
19152 (vect_enhance_data_refs_alignment): Likewise.
19153 (vect_analyze_data_refs_alignment): Likewise.
19154 (vect_slp_analyze_and_verify_instance_alignment
19155 (vect_analyze_data_ref_accesses): Likewise.
19156 (vect_prune_runtime_alias_test_list): Likewise.
19157 (vect_analyze_data_refs): Likewise.
19158 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
19159 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19160 (vect_analyze_scalar_cycles_1): Likewise.
19161 (vect_get_loop_niters): Likewise.
19162 (vect_analyze_loop_form_1): Likewise.
19163 (vect_update_vf_for_slp): Likewise.
19164 (vect_analyze_loop_operations): Likewise.
19165 (vect_analyze_loop): Likewise.
19166 (vectorizable_induction): Likewise.
19167 (vect_transform_loop): Likewise.
19168 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
19169 * tree-vect-slp.c (vect_analyze_slp): Likewise.
19170 (vect_make_slp_decision): Likewise.
19171 (vect_detect_hybrid_slp): Likewise.
19172 (vect_slp_analyze_operations): Likewise.
19173 (vect_slp_bb): Likewise.
19174 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
19175 (vectorizable_bswap): Likewise.
19176 (vectorizable_call): Likewise.
19177 (vectorizable_simd_clone_call): Likewise.
19178 (vectorizable_conversion): Likewise.
19179 (vectorizable_assignment): Likewise.
19180 (vectorizable_shift): Likewise.
19181 (vectorizable_operation): Likewise.
19182 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
19184 2018-06-18 Martin Sebor <msebor@redhat.com>
19186 PR tree-optimization/81384
19187 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
19188 * builtins.c (expand_builtin_strnlen): New function.
19189 (expand_builtin): Call it.
19190 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
19191 * builtins.def (BUILT_IN_STRNLEN): New.
19192 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
19193 Warn for bounds in excess of maximum object size.
19194 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
19195 single-value ranges. Handle strnlen.
19196 (handle_builtin_strlen): Handle strnlen.
19197 (strlen_check_and_optimize_stmt): Same.
19198 * doc/extend.texi (Other Builtins): Document strnlen.
19200 2018-06-18 Maya Rashish <coypu@sdf.org>
19202 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
19203 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
19204 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
19206 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
19208 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
19210 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19212 * tree.c (escaped_string::escape): Replace cast to char * by
19213 const_cast<char *> (unescaped).
19215 2018-06-18 Nick Clifton <nickc@redhat.com>
19218 * tree.c (escaped_string): New class. Converts an unescaped
19219 string into its escaped equivalent.
19220 (warn_deprecated_use): Use the new class to convert the
19221 deprecation message, if present.
19222 (test_escaped_strings): New self test.
19223 (test_c_tests): Add test_escaped_strings.
19224 * doc/extend.texi (deprecated): Add a note that the
19225 deprecation message is affected by the -fmessage-length
19226 option, and that control characters will be escaped.
19227 (#pragma GCC error): Document this pragma.
19228 (#pragma GCC warning): Likewise.
19229 * doc/invoke.texi (-fmessage-length): Document this option's
19230 effect on the #warning and #error preprocessor directives and
19231 the deprecated attribute.
19233 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
19235 * tree.c (decl_value_expr_lookup): Revert latest change.
19236 (decl_value_expr_insert): Likewise.
19238 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
19240 * gimplify.c (nonlocal_vlas): Delete.
19241 (nonlocal_vla_vars): Likewise.
19242 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
19244 (gimplify_body): Do not create and destroy nonlocal_vlas.
19245 * tree-nested.c: Include diagnostic.h.
19246 (use_pointer_in_frame): Tweak.
19247 (lookup_field_for_decl): Add assertion and declare the transformation.
19248 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
19249 internal error when the reference is in a wrong context. Do not
19250 create a debug decl by default.
19251 (note_nonlocal_block_vlas): Delete.
19252 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
19253 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
19254 create a debug decl by default.
19255 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
19256 call is in a wrong context.
19257 (fixup_vla_decls): New function.
19258 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
19259 debug variables were created.
19260 * tree.c (decl_value_expr_lookup): Add checking assertion.
19261 (decl_value_expr_insert): Likewise.
19263 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
19265 PR middle-end/82479
19266 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
19267 * tree-scalar-evolution.c (interpret_expr): Likewise.
19268 (expression_expensive_p): Likewise.
19269 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
19270 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
19271 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
19272 (ssa_defined_by_minus_one_stmt_p): New.
19274 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
19276 PR middle-end/64946
19277 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
19278 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
19279 * dojump.c (do_jump): Likewise.
19280 * expr.c (expand_expr_real_2): Check operand type's sign.
19281 * fold-const.c (const_unop): Handle ABSU_EXPR.
19282 (fold_abs_const): Likewise.
19283 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
19284 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
19285 (strip_sign_op_1): Likesise.
19286 * match.pd: Add new pattern to generate ABSU_EXPR.
19287 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
19288 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
19289 * tree-eh.c (operation_could_trap_helper_p): Likewise.
19290 * tree-inline.c (estimate_operator_cost): Likewise.
19291 * tree-pretty-print.c (dump_generic_node): Likewise.
19292 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
19293 * tree.def (ABSU_EXPR): New.
19295 2018-06-16 Jakub Jelinek <jakub@redhat.com>
19297 PR middle-end/86095
19298 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
19299 documented as preserved for backward compatibility only.
19300 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
19302 PR rtl-optimization/86108
19303 * bb-reorder.c (create_forwarder_block): Renamed to ...
19304 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
19305 jump from new landing pad to the second part.
19306 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
19309 2018-06-15 Jakub Jelinek <jakub@redhat.com>
19311 PR middle-end/85878
19312 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
19313 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
19314 Only call store_expr for halves if the mode is the same.
19316 PR middle-end/86123
19317 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
19318 Fix up comment formatting.
19320 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
19322 * typed-splay-tree.h (typed_splay_tree::remove): New function.
19323 (typed_splay_tree::closure,
19324 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
19325 (typed_splay_tree::typed_splay_tree,
19326 typed_splay_tree::operator =): Declared private.
19327 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
19328 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
19329 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
19330 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
19331 typed_splay_tree::splay_tree_splay,
19332 typed_splay_tree::splay_tree_foreach_helper,
19333 typed_splay_tree::splay_tree_insert,
19334 typed_splay_tree::splay_tree_remove,
19335 typed_splay_tree::splay_tree_lookup,
19336 typed_splay_tree::splay_tree_predecessor,
19337 typed_splay_tree::splay_tree_successor,
19338 typed_splay_tree::splay_tree_min,
19339 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
19340 (typed_splay_tree::root, typed_splay_tree::comp,
19341 typed_splay_tree::delete_key,
19342 typed_splay_tree::delete_value): New data members.
19343 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
19344 typed_splay_tree::remove.
19346 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
19348 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
19349 -mginv and -mno-ginv to the assembler.
19350 * config/mips/mips.opt (-mcrc): New option.
19351 (-mginv): Likewise.
19352 * doc/invoke.text (-mcrc): Document.
19353 (-mginv): Likewise.
19355 2018-06-15 Nick Clifton <nickc@redhat.com>
19358 * tree.c (escaped_string): New class. Converts an unescaped
19359 string into its escaped equivalent.
19360 (warn_deprecated_use): Use the new class to convert the
19361 deprecation message, if present.
19362 (test_escaped_strings): New self test.
19363 (test_c_tests): Add test_escaped_strings.
19364 * doc/extend.texi (deprecated): Add a note that the
19365 deprecation message is affected by the -fmessage-length
19366 option, and that control characters will be escaped.
19367 (#pragma GCC error): Document this pragma.
19368 (#pragma GCC warning): Likewise.
19369 * doc/invoke.texi (-fmessage-length): Document this option's
19370 effect on the #warning and #error preprocessor directives and
19371 the deprecated attribute.
19373 2018-06-15 Richard Biener <rguenther@suse.de>
19375 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
19376 here, also noting vector size used.
19377 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
19378 size used in MSG_OPTIMIZED_LOCATIONS dump.
19379 (pass_slp_vectorize::execute): Adjust.
19381 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
19384 * config/arc/arc.c (arc_return_address_register): Fix
19387 2018-06-15 Richard Biener <rguenther@suse.de>
19389 PR middle-end/86159
19390 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
19391 leave useless conversion stripping to force_gimple_operand_gsi.
19392 (gimplify_build2): Likewise.
19393 (gimplify_build1): Likewise.
19395 2018-06-15 Richard Biener <rguenther@suse.de>
19397 PR middle-end/86076
19398 * tree-cfg.c (move_stmt_op): unshare invariant addresses
19399 before adjusting their block.
19401 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
19403 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
19404 multilibs for *-*-rtems*.
19405 * config/riscv/t-rtems: New file.
19407 2018-06-14 Jakub Jelinek <jakub@redhat.com>
19409 PR middle-end/86122
19410 * match.pd ((A +- CST1) +- CST2): Punt if last resort
19411 unsigned_type_for returns NULL.
19414 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
19415 subregs of multi-word pseudos unless the float mode has word size.
19417 2018-06-14 Richard Biener <rguenther@suse.de>
19419 PR middle-end/86139
19420 * tree-vect-generic.c (build_word_mode_vector_type): Remove
19421 duplicate and harmful type_hash_canon.
19422 * tree.c (type_hash_canon): Assert we didn't find ourselves.
19424 2018-06-14 Richard Biener <rguenther@suse.de>
19427 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
19430 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
19432 * config/rtems.h (STDINT_LONG32): Define.
19434 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
19435 Prachi Godbole <prachi.godbole@imgtec.com>
19437 * config/mips/mips-cpus.def: Define P6600.
19438 * config/mips/mips-tables.opt: Regenerate.
19439 * config/mips/mips.c (mips_ucbranch_type): New enum.
19440 (mips_rtx_cost_data): Add support for P6600.
19441 (mips_issue_rate): Likewise.
19442 (mips_multipass_dfa_lookahead): Likewise.
19443 (mips_avoid_hazard): Likewise.
19444 (mips_reorg_process_insns): Likewise.
19445 (mips_classify_branch_p6600): New function.
19446 * config/mips/mips.h (TUNE_P6600): New define.
19447 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
19448 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
19449 * config/mips/mips.md: Include p6600.md.
19450 (processor): Add p6600.
19451 * config/mips/p6600.md: New file.
19452 * doc/invoke.texi: Add p6600 to supported architectures.
19454 2018-06-13 Martin Sebor <msebor@redhat.com>
19456 PR tree-optimization/86114
19457 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
19459 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
19461 2018-06-13 Richard Biener <rguenther@suse.de>
19463 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
19464 Properly set vector type of the intermediate stmt.
19465 * tree-vect-stmts.c (vectorizable_operation): The destination
19466 var always has vectype_out type.
19468 2018-06-13 Jeff Law <law@redhat.com>
19470 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
19471 integer 0 for argument to print_rtl_with_bb.
19472 (rl78_reorg): Likewise.
19474 2018-06-13 David Malcolm <dmalcolm@redhat.com>
19476 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
19477 from rtx to rtx_insn *.
19478 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
19480 (add_sched_insns_for_speculation): Likewise for local "target",
19481 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
19482 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
19483 from rtx_insn ** to rtx_code_label **.
19484 (reorg_emit_nops): Likewise.
19485 (c6x_reorg): Likewise for local "call_labels".
19486 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
19488 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
19489 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
19490 the loops over LABEL_REFS.
19491 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
19493 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
19494 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
19495 (split_branches): Strengthen local "olabel" from rtx to
19496 rtx_insn *, adding a safe_as_a cast.
19497 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
19499 (add_insn_after): Likewise for first two params.
19500 (add_insn_before): Likewise.
19501 (remove_insn): Likewise for param.
19502 (emit_pattern_before_noloc): Likewise for second and third params.
19503 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
19504 (emit_call_insn_before_noloc): Likewise.
19505 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
19507 (emit_barrier_before): Likewise.
19508 (emit_label_before): Strengthen "label" param from "rtx" to
19509 "rtx_code_label *". Strengthen "before" param from "rtx" to
19511 (emit_insn_after_1): Strengthen "after" param from "rtx" to
19513 (emit_pattern_after_noloc): Likewise.
19514 (emit_insn_after_noloc): Likewise.
19515 (emit_jump_insn_after_noloc): Likewise.
19516 (emit_call_insn_after_noloc): Likewise.
19517 (emit_debug_insn_after_noloc): Likewise.
19518 (emit_barrier_after): Likewise.
19519 (emit_label_after): Likewise for both params.
19520 (emit_pattern_after_setloc): Likewise for "after" param. Convert
19521 "loc" param from "int" to "location_t".
19522 (emit_insn_after_setloc): Likewise.
19523 (emit_jump_insn_after_setloc): Likewise.
19524 (emit_call_insn_after_setloc): Likewise.
19525 (emit_debug_insn_after_setloc): Likewise.
19526 (emit_pattern_before_setloc): Likewise for "before" param. Convert
19527 "loc" param from "int" to "location_t".
19528 (emit_pattern_before): Convert NULL_RTX to NULL.
19529 (emit_insn_before_setloc): Convert "loc" param from "int" to
19531 (emit_jump_insn_before_setloc): Likewise.
19532 (emit_call_insn_before_setloc): Likewise.
19533 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
19534 rtx_insn *. Convert "loc" param from "int" to "location_t".
19535 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
19536 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
19537 Convert 3rd param from "int" to "location_t".
19538 (emit_barrier_before, emit_barrier_after, next_real_insn):
19539 Strengthen param from rtx to rtx_insn *.
19540 (emit_label_before): Strengthen 1st param from "rtx" to
19541 "rtx_code_label *". Strengthen 2nd param from "rtx" to
19543 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
19544 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
19545 Strengthen 2nd param from "rtx" to "rtx_insn *".
19546 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
19547 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
19548 Likewise. Convert 3rd param from "int" to "location_t".
19549 (emit_label_after): Strengthen 1st param from "rtx" to
19550 "rtx_code_label *".
19551 (next_real_insn, remove_insn): Strengthen param from "rtx" to
19553 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
19554 from "rtx" to "rtx_insn *".
19556 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
19558 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
19559 bodies streamed in with -Q.
19560 * dumpfile.c (dump_files): Add lto-stream-out dump file.
19561 * dumpfile.h (tree_dump_index): Add lto_stream_out.
19562 * gimple-streamer-out.c: Include gimple-pretty-print.h
19563 (output_bb): Dump stmts streamed.
19564 * lto-section-out.c: Include print-tree.h
19565 (lto_begin_section): Dump sections created.
19566 (lto_output_decl_index): Dump decl encoded.
19567 * lto-streamer-out.c: Include print-tree.h
19568 (create_output_block): Dump output block created.
19569 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
19570 (output_function): Dump function output.
19571 (output_constructor): Dump constructor streamed.
19572 (write_global_stream): Output indexes encoded.
19573 (produce_asm_for_decls): Dump streams encoded.
19574 * lto-streamer.c (streamer_dump_file): New global var.
19575 * lto-streamer.h (streamer_dump_file): Declare.
19576 * passes.c (ipa_write_summaries): Initialize streamer dump.
19577 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
19580 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
19583 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
19584 offsets for register save directives. Emit a second batch of save
19585 directives, if need be, when the function accesses prior frames.
19587 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19589 * config/arc/fpu.md (fmasf4): Force operand to register.
19590 (fnmasf4): Likewise.
19592 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19594 * config/arc/arc-protos.h (arc_pad_return): Remove.
19595 * config/arc/arc.c (machine_function): Remove force_short_suffix
19597 (arc_print_operand): Adjust printing of '&'.
19598 (arc_verify_short): Remove conditional printing of short suffix.
19599 (arc_final_prescan_insn): Remove reference to size_reason.
19600 (pad_return): New function.
19601 (arc_reorg): Call pad_return.
19602 (arc_pad_return): Remove.
19603 (arc_init_machine_status): Remove reference to force_short_suffix.
19604 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
19605 (attr length): When attribute iscompact is true force to 2
19606 regardless; in the case of maybe check if we want to force the
19607 instruction to have 4 bytes length.
19608 (nopv): Change it to generate 4 byte long nop as well.
19609 (blockage): New pattern.
19610 (simple_return): Remove call to arc_pad_return.
19611 (p_return_i): Likewise.
19613 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19615 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
19617 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19619 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
19622 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19624 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
19625 for ARC700 and ARCv2.
19627 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
19630 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
19631 operands[2] instead of operands[1].
19634 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
19636 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
19637 case, check whether the outer register overlaps an unallocatable
19638 register, not just whether it fits the required class.
19640 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
19642 * poly-int.h (can_div_trunc_p): Add new overload in which all values
19644 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
19645 (memrefs_conflict_p): Likewise.
19646 (init_alias_analysis): Likewise.
19647 * cfgexpand.c (expand_debug_expr): Likewise.
19648 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
19649 * cse.c (fold_rtx): Likewise.
19650 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
19651 * expr.c (emit_block_move_hints): Likewise.
19652 (clear_storage_hints, push_block, emit_push_insn): Likewise.
19653 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
19654 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
19655 (emit_group_store): Likewise.
19656 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
19657 to read the PRE/POST_MODIFY increment.
19658 * calls.c (store_one_arg): Use strip_offset.
19659 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
19661 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
19663 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
19664 (simplify_binary_operation_1): Extend CONST_INT handling to
19666 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
19667 than a HOST_WIDE_INT.
19668 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
19670 (adjust_mems, add_stores): Update accodingly.
19671 (vt_canonicalize_addr): Track polynomial offsets.
19672 (emit_note_insn_var_location): Likewise.
19673 (vt_add_function_parameter): Likewise.
19674 (vt_initialize): Likewise.
19676 2018-06-12 Jeff Law <law@redhat.com>
19678 * config.gcc (alpha*-*-freebsd*): Remove.
19679 * config/alpha/freebsd.h: Remove.
19681 2018-06-12 David Malcolm <dmalcolm@redhat.com>
19684 * spellcheck-tree.c (levenshtein_distance): Rename to...
19685 (get_edit_distance): ...this, and update for underlying renaming.
19686 * spellcheck-tree.h (levenshtein_distance): Rename to...
19687 (get_edit_distance): ...this.
19688 * spellcheck.c (levenshtein_distance): Rename to...
19689 (get_edit_distance): ...this. Convert from Levenshtein distance
19690 to Damerau-Levenshtein distance by supporting transpositions of
19691 adjacent characters. Rename "v1" to "v_next" and "v0" to
19693 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
19694 (selftest::test_edit_distance_unit_test_oneway): ...this, and
19695 update for underlying renaming.
19696 (selftest::levenshtein_distance_unit_test): Rename to...
19697 (selftest::test_get_edit_distance_unit): ...this, and update for
19698 underlying renaming.
19699 (selftest::test_find_closest_string): Add example from PR 69968
19700 where transposition helps
19701 (selftest::test_metric_conditions): Update for renaming.
19702 (selftest::test_metric_conditions): Likewise.
19703 (selftest::spellcheck_c_tests): Likewise.
19704 * spellcheck.h (levenshtein_distance): Rename both overloads to...
19705 (get_edit_distance): ...this.
19706 (best_match::consider): Update for renaming.
19708 2018-06-12 Martin Sebor <msebor@redhat.com>
19710 PR tree-optimization/85259
19711 * builtins.c (compute_objsize): Handle constant offsets.
19712 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
19713 true iff a warning has been issued.
19714 * gimple.h (gimple_nonartificial_location): New function.
19715 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
19716 gimple_nonartificial_location and handle -Wno-system-headers.
19717 (handle_builtin_stxncpy): Same.
19719 2018-06-12 Martin Sebor <msebor@redhat.com>
19722 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
19724 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
19726 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
19727 BUILTIN_VEC_XST entries for pointer to double and long long.
19729 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
19732 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19734 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19737 2018-06-12 Martin Liska <mliska@suse.cz>
19739 * doc/options.texi: Document IntegerRange.
19741 2018-06-12 Martin Liska <mliska@suse.cz>
19743 * config/i386/i386.opt: Make MPX-related options as Deprecated.
19744 * opt-functions.awk: Handle Deprecated flag.
19745 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
19747 (read_cmdline_option): Report warning for a deprecated option.
19748 * opts.h (struct cl_option): Add new field cl_deprecated.
19749 (CL_ERR_DEPRECATED): New.
19751 2018-06-12 Martin Liska <mliska@suse.cz>
19753 * doc/options.texi: Document Deprecated option flag.
19755 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19757 * config/arc/arc-arch.h (arc_extras): New enum.
19758 (arc_cpu_t):Add field extra.
19759 (arc_cpu_types): Consider the extras.
19760 * config/arc/arc-cpus.def: Add extras info.
19761 * config/arc/arc-opts.h (processor_type): Consider extra field.
19762 * config/arc/arc.c (arc_override_options): Handle extra field.
19764 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19766 * config/arc/arc-arch.h: Update ARC_OPTX macro.
19767 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
19769 * config/arc/arc.c (arc_init): Update pic warning.
19770 (irq_range): Update irq range parsing warnings.
19771 (arc_override_options): Update various warning messages.
19772 (arc_handle_aux_attribute): Likewise.
19774 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
19776 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
19778 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
19780 * doc/sourcebuild.texi: Document usage of line number 0 in verify
19781 compiler messages directives.
19783 2018-06-12 Matthew Fortune <mfortune@gmail.com>
19785 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
19786 * config/mips/mips-tables.opt: Regenerate.
19787 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
19789 * doc/invoke.texi: Document -march=i6500.
19791 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
19793 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
19794 (i6400_gpmul): Add cpu_unit.
19795 (i6400_gpdiv): Likewise.
19796 (i6400_msa_add_d): Update reservations.
19797 (i6400_msa_int_add) Likewise.
19798 (i6400_msa_short_logic3) Likewise.
19799 (i6400_msa_short_logic2) Likewise.
19800 (i6400_msa_short_logic) Likewise.
19801 (i6400_msa_move) Likewise.
19802 (i6400_msa_cmp) Likewise.
19803 (i6400_msa_short_float2) Likewise.
19804 (i6400_msa_div_d) Likewise.
19805 (i6400_msa_long_logic1) Likewise.
19806 (i6400_msa_long_logic2) Likewise.
19807 (i6400_msa_mult) Likewise.
19808 (i6400_msa_long_float2) Likewise.
19809 (i6400_msa_long_float4) Likewise.
19810 (i6400_msa_long_float5) Likewise.
19811 (i6400_msa_long_float8) Likewise.
19812 (i6400_fpu_fadd): Include frint type.
19813 (i6400_fpu_store): New define_insn_reservation.
19814 (i6400_fpu_load): Likewise.
19815 (i6400_fpu_move): Likewise.
19816 (i6400_fpu_fcmp): Likewise.
19817 (i6400_fpu_fmadd): Likewise.
19818 (i6400_int_mult): Include imul3nc type and update reservation.
19819 (i6400_int_div): Include idiv3 type and update reservation.
19820 (i6400_int_load): Update to check type not move_type.
19821 (i6400_int_store): Likewise.
19822 (i6400_int_prefetch): Set zero latency.
19824 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
19826 * gcc.c: Document new %@{...} sequence.
19827 (LINK_COMMAND_SPEC): Use it for the -L switches.
19828 (cpp_unique_options): Use it for the -I switches.
19829 (at_file_argbuf): New global variable.
19830 (in_at_file): Likewise.
19831 (alloc_args): Create at_file_argbuf.
19832 (clear_args): Truncate at_file_argbuf.
19833 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
19834 (open_at_file): New function.
19835 (close_at_file): Likewise.
19836 (create_at_file): Delete.
19837 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
19840 (validate_switches_from_spec): Deal with %@{...} sequence.
19841 (validate_switches): Likewise.
19842 (driver::finalize): Call clear_args.
19844 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
19846 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
19848 2018-06-11 Martin Sebor <msebor@redhat.com>
19850 * doc/invoke.texi (-Wall): List -Wc++17-compat.
19851 (Wno-class-memaccess): Add @opindex.
19852 (Wno-templates, Wno-multiple-inheritance): Same.
19853 (Wno-virtual-inheritance, Wno-namespaces): Same.
19854 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
19855 (Wno-format-overflow, Wno-format-truncation): Same.
19856 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
19857 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
19858 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
19859 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
19860 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
19861 (Wno-misspelled-isr): Same.
19863 2018-06-11 Martin Sebor <msebor@redhat.com>
19865 * PR tree-optimization/86083
19866 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
19868 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
19870 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
19872 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
19875 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
19876 on the correct operand.
19877 (*movdi_internal64): Ditto.
19879 2018-06-11 Martin Liska <mliska@suse.cz>
19881 PR tree-optimization/86089
19882 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
19884 2018-06-11 Julia Koval <julia.koval@intel.com>
19886 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
19887 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
19888 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
19890 2018-06-11 Olivier Hainque <hainque@adacore.com>
19892 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
19893 for Ada with strict dwarf2.
19895 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
19898 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
19901 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
19903 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
19905 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
19907 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
19910 2018-06-08 Martin Liska <mliska@suse.cz>
19912 * tree-cfg.h (debug_function): Fix argument type to match
19915 2018-06-08 Martin Liska <mliska@suse.cz>
19917 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
19918 Remove usage of MPX-related (and removed) fields.
19919 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
19921 2018-06-08 David Malcolm <dmalcolm@redhat.com>
19923 * cfg.c (debug): Use TDF_NONE rather than 0.
19924 * cfghooks.c (debug): Likewise.
19925 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
19926 (struct dump_option_value_info): Convert to...
19927 (struct kv_pair): ...this template type.
19928 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
19930 (optinfo_verbosity_options): Likewise.
19931 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
19933 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
19934 than int for "optgroup_flags" param.
19935 (dump_generic_expr_loc): Use dump_flags_t rather than int for
19937 (dump_dec): Likewise.
19938 (dump_finish): Use TDF_NONE rather than 0.
19939 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
19940 rather than int for "optgroup_flags" param. Use TDF_NONE rather
19941 than 0. Update for change to option_ptr.
19942 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
19943 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
19944 0. Update for changes to optinfo_verbosity_options and
19946 (opt_info_switch_p): Convert optgroup_flags from int to
19948 (dump_basic_block): Use dump_flags_t rather than int
19949 for "dump_kind" param.
19950 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
19951 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
19952 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
19953 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
19954 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
19955 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
19956 TDF_NONE): Convert from macros to...
19957 (enum dump_flag): ...this new enum.
19958 (dump_flags_t): Update to use enum.
19959 (operator|, operator&, operator~, operator|=, operator&=):
19960 Implement for dump_flags_t.
19961 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
19962 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
19963 Convert from macros to...
19964 (enum optgroup_flag): ...this new enum.
19965 (optgroup_flags_t): New typedef.
19966 (operator|, operator|=): Implement for optgroup_flags_t.
19967 (struct dump_file_info): Convert field "alt_flags" to
19968 dump_flags_t. Convert field "optgroup_flags" to
19970 (dump_basic_block): Use dump_flags_t rather than int for param.
19971 (dump_generic_expr_loc): Likewise.
19972 (dump_dec): Likewise.
19973 (dump_register): Convert param "optgroup_flags" to
19975 (opt_info_enable_passes): Likewise.
19976 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
19978 * gimple-pretty-print.c (debug): Likewise.
19979 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
19980 (merged_store_group::apply_stores): Likewise.
19981 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
19982 * gimple.c (verify_gimple_pp): Likewise.
19983 * graphite-poly.c (print_pbb_body): Likewise.
19984 * passes.c (pass_manager::register_one_dump_file): Convert
19985 local "optgroup_flags" to optgroup_flags_t.
19986 * print-tree.c (print_node): Use TDF_NONE rather than 0.
19988 (debug_body): Likewise.
19989 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
19990 to optgroup_flags_t.
19991 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
19993 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
19994 (convert_mult_to_fma): Likewise.
19995 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
19996 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
19997 * tree-vect-data-refs.c (dump_lower_bound): Convert param
19998 "dump_kind" to dump_flags_t.
20000 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
20002 * config/rs6000/rs6000.c (min, max): Delete.
20004 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
20006 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
20007 -mabi=spe and -mabi=no-spe.
20009 2018-06-08 Martin Liska <mliska@suse.cz>
20011 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
20012 where we expect an existing summary.
20014 2018-06-08 Martin Liska <mliska@suse.cz>
20016 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
20017 * ipa-inline.h (estimate_edge_growth): Likewise.
20019 2018-06-08 Martin Liska <mliska@suse.cz>
20021 * cgraph.c (function_version_hasher::hash): Use
20022 cgraph_node::get_uid ().
20023 (function_version_hasher::equal):
20024 * cgraph.h (cgraph_node::get_uid): New method.
20025 * ipa-inline.c (update_caller_keys): Use
20026 cgraph_node::get_uid ().
20027 (update_callee_keys): Likewise.
20028 * ipa-utils.c (searchc): Likewise.
20029 (ipa_reduced_postorder): Likewise.
20030 * lto-cgraph.c (input_node): Likewise.
20031 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
20032 * symbol-summary.h (symtab_insertion): Likewise.
20033 (symtab_removal): Likewise.
20034 (symtab_duplication): Likewise.
20035 * tree-pretty-print.c (dump_function_header): Likewise.
20036 * tree-sra.c (convert_callers_for_node): Likewise.
20038 2018-06-08 Martin Liska <mliska@suse.cz>
20040 * cgraph.c (symbol_table::create_edge): Always assign a new
20042 (symbol_table::free_edge): Do not recycle numbers.
20043 * cgraph.h (cgraph_edge::get): New method.
20044 * symbol-summary.h (symtab_removal): Use it.
20045 (symtab_duplication): Likewise.
20046 (call_summary::hashable_uid): Remove.
20048 2018-06-08 Martin Liska <mliska@suse.cz>
20050 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
20051 (initialize_growth_caches): Remove.
20052 (free_growth_caches): Likewise.
20053 (do_estimate_edge_time): Use edge_growth_cache.
20054 (do_estimate_edge_size): Likewise.
20055 (do_estimate_edge_hints): Likewise.
20056 * ipa-inline.c (reset_edge_caches): Likewise.
20057 (recursive_inlining): Likewise.
20058 (inline_small_functions): Likewise.
20059 * ipa-inline.h (initialize_growth_caches): Remove.
20060 (estimate_edge_size): Likewise.
20061 (estimate_edge_time): Likewise.
20062 (estimate_edge_hints): Likewise.
20063 (reset_edge_growth_cache): Likewise.
20064 * symbol-summary.h (call_summary::remove): New method.
20066 2018-06-08 Martin Liska <mliska@suse.cz>
20068 * ipa-cp.c (class edge_clone_summary): New summary.
20069 (grow_edge_clone_vectors): Remove.
20070 (ipcp_edge_duplication_hook): Remove.
20071 (class edge_clone_summary_t): New call_summary class.
20072 (ipcp_edge_removal_hook): Remove.
20073 (edge_clone_summary_t::duplicate): New function.
20074 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
20075 (create_specialized_node): Likewise.
20076 (ipcp_driver): Initialize edge_clone_summaries and do not
20079 2018-06-08 Martin Liska <mliska@suse.cz>
20081 * symbol-summary.h (get): New function.
20082 (call_summary::m_initialize_when_cloning): New class member.
20084 2018-06-08 Martin Liska <mliska@suse.cz>
20086 * cgraph.c (cgraph_node::remove): Do not recycle uid.
20087 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
20088 (symbol_table::allocate_cgraph_symbol): Do not set uid.
20089 * passes.c (uid_hash_t): Record removed_nodes by their uids.
20090 (remove_cgraph_node_from_order): Use the removed_nodes set.
20091 (do_per_function_toporder): Likwise.
20092 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
20093 instead of summary_uid.
20094 (symtab_removal): Likewise.
20095 (symtab_duplication): Likewise.
20097 2018-06-08 Martin Liska <mliska@suse.cz>
20099 * ipa-cp.c (ipcp_store_bits_results): Use
20100 ipcp_transformation_sum.
20101 (ipcp_store_vr_results): Likewise.
20102 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
20104 (ipcp_transformation_initialize): ... this.
20105 (ipa_set_node_agg_value_chain):
20106 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
20107 (write_ipcp_transformation_info): Likewise.
20108 (read_ipcp_transformation_info): Likewise.
20109 (ipcp_update_bits): Likewise.
20110 (ipcp_update_vr): Likewise.
20111 (ipcp_transform_function): Likewise.
20112 * ipa-prop.h: Rename ipcp_transformation_summary to
20113 ipcp_transformation.
20114 (class ipcp_transformation_t): New function summary.
20115 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
20116 (ipa_get_agg_replacements_for_node): Likewise.
20118 2018-06-08 Martin Liska <mliska@suse.cz>
20120 * ipa-pure-const.c (struct funct_state_d): Do it class instead
20122 (class funct_state_summary_t): New function_summary class.
20123 (has_function_state): Remove.
20124 (get_function_state): Likewise.
20125 (set_function_state): Likewise.
20126 (add_new_function): Likewise.
20127 (funct_state_summary_t::insert): New function.
20128 (duplicate_node_data): Remove.
20129 (remove_node_data): Remove.
20130 (funct_state_summary_t::duplicate): New function.
20131 (register_hooks): Create new funct_state_summaries.
20132 (pure_const_generate_summary): Use it.
20133 (pure_const_write_summary): Likewise.
20134 (pure_const_read_summary): Likewise.
20135 (propagate_pure_const): Likewise.
20136 (propagate_nothrow): Likewise.
20137 (dump_malloc_lattice): Likewise.
20138 (propagate_malloc): Likewise.
20139 (execute): Do not register hooks, just remove summary
20141 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
20144 2018-06-08 Martin Liska <mliska@suse.cz>
20146 * ipa-reference.c (remove_node_data): Remove.
20147 (duplicate_node_data): Likewise.
20148 (class ipa_ref_var_info_summary_t): New class.
20149 (class ipa_ref_opt_summary_t): Likewise.
20150 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
20151 (get_reference_optimization_summary): Use
20152 ipa_ref_opt_sum_summaries.
20153 (set_reference_vars_info): Remove.
20154 (set_reference_optimization_summary): Likewise.
20155 (ipa_init): Create summaries.
20156 (init_function_info): Use function summary.
20157 (ipa_ref_opt_summary_t::duplicate): New function.
20158 (ipa_ref_opt_summary_t::remove): New function.
20159 (get_read_write_all_from_node): Fix GNU coding style.
20160 (propagate): Use function summary.
20161 (write_node_summary_p): Fix GNU coding style.
20162 (stream_out_bitmap): Likewise.
20163 (ipa_reference_read_optimization_summary): Use function summary.
20164 (ipa_reference_c_finalize): Do not release hooks.
20166 2018-06-08 Martin Liska <mliska@suse.cz>
20168 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
20169 (analyze_function_body): Extract multiple calls of get_create.
20170 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
20171 * ipa-inline.c (recursive_inlining): Use ::get method.
20172 * ipa-inline.h (estimate_edge_growth): Likewise.
20174 2018-06-08 Martin Liska <mliska@suse.cz>
20176 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
20178 (hsa_function_summary::hsa_function_summary): Use the new enum
20180 (hsa_gpu_implementation_p): Use hsa_summaries::get.
20181 * hsa-gen.c (hsa_get_host_function): Likewise.
20182 (get_brig_function_name): Likewise.
20183 * ipa-hsa.c (process_hsa_functions): Likewise.
20184 (ipa_hsa_write_summary): Likewise.
20185 * symbol-summary.h (symtab_duplication): Use ::get function/
20186 (get): New function.
20188 2018-06-08 Martin Liska <mliska@suse.cz>
20190 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
20192 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
20193 (hsa_register_kernel): Likewise.
20194 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
20195 * hsa-gen.c (hsa_get_host_function): Likewise.
20196 (get_brig_function_name): Likewise.
20197 (generate_hsa): Likewise.
20198 (pass_gen_hsail::execute): Likewise.
20199 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
20200 (devirtualization_time_bonus): Likewise.
20201 (ipcp_propagate_stage): Likewise.
20202 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
20203 (edge_set_predicate): Likewise.
20204 (evaluate_conditions_for_known_args): Likewise.
20205 (evaluate_properties_for_edge): Likewise.
20206 (ipa_fn_summary::reset): Likewise.
20207 (ipa_fn_summary_t::duplicate): Likewise.
20208 (dump_ipa_call_summary): Likewise.
20209 (ipa_dump_fn_summary): Likewise.
20210 (analyze_function_body): Likewise.
20211 (compute_fn_summary): Likewise.
20212 (estimate_edge_devirt_benefit): Likewise.
20213 (estimate_edge_size_and_time): Likewise.
20214 (estimate_calls_size_and_time): Likewise.
20215 (estimate_node_size_and_time): Likewise.
20216 (inline_update_callee_summaries): Likewise.
20217 (remap_edge_change_prob): Likewise.
20218 (remap_edge_summaries): Likewise.
20219 (ipa_merge_fn_summary_after_inlining): Likewise.
20220 (ipa_update_overall_fn_summary): Likewise.
20221 (read_ipa_call_summary): Likewise.
20222 (inline_read_section): Likewise.
20223 (write_ipa_call_summary): Likewise.
20224 (ipa_fn_summary_write): Likewise.
20225 (ipa_free_fn_summary): Likewise.
20226 * ipa-hsa.c (process_hsa_functions): Likewise.
20227 (ipa_hsa_write_summary): Likewise.
20228 (ipa_hsa_read_section): Likewise.
20229 * ipa-icf.c (sem_function::merge): Likewise.
20230 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
20231 (do_estimate_edge_time): Likewise.
20232 (estimate_size_after_inlining): Likewise.
20233 (estimate_growth): Likewise.
20234 (growth_likely_positive): Likewise.
20235 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
20236 (inline_call): Likewise.
20237 * ipa-inline.c (caller_growth_limits): Likewise.
20238 (can_inline_edge_p): Likewise.
20239 (can_inline_edge_by_limits_p): Likewise.
20240 (compute_uninlined_call_time): Likewise.
20241 (compute_inlined_call_time): Likewise.
20242 (want_inline_small_function_p): Likewise.
20243 (edge_badness): Likewise.
20244 (update_caller_keys): Likewise.
20245 (update_callee_keys): Likewise.
20246 (recursive_inlining): Likewise.
20247 (inline_small_functions): Likewise.
20248 (inline_to_all_callers_1): Likewise.
20249 (dump_overall_stats): Likewise.
20250 (early_inline_small_functions): Likewise.
20251 (early_inliner): Likewise.
20252 * ipa-inline.h (estimate_edge_growth): Likewise.
20253 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
20254 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
20255 * ipa-prop.h (IPA_NODE_REF): Likewise.
20256 (IPA_EDGE_REF): Likewise.
20257 * ipa-pure-const.c (malloc_candidate_p): Likewise.
20258 (propagate_malloc): Likewise.
20259 * ipa-split.c (execute_split_functions): Likewise.
20260 * symbol-summary.h: Rename get to get_create.
20262 (get_create): Likewise.
20263 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
20265 2018-06-08 Martin Liska <mliska@suse.cz>
20267 * symbol-summary.h (release): Move definition out of class
20269 (symtab_removal): Likewise.
20270 (symtab_duplication): Likewise.
20272 2018-06-08 Martin Liska <mliska@suse.cz>
20274 * symbol-summary.h (function_summary): Move constructor
20275 implementation out of class declaration.
20276 (release): Likewise.
20277 (symtab_insertion): Likewise.
20278 (symtab_removal): Likewise.
20279 (symtab_duplication): Likewise.
20282 2018-06-08 Martin Liska <mliska@suse.cz>
20284 * Makefile.in: Remove support for MPX (macros, related functions,
20285 fields in cgraph_node, ...).
20286 * builtin-types.def (BT_BND): Likewise.
20287 (BT_FN_BND_CONST_PTR): Likewise.
20288 (BT_FN_CONST_PTR_BND): Likewise.
20289 (BT_FN_VOID_PTR_BND): Likewise.
20290 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
20291 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
20292 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
20293 (expand_builtin_mempcpy_with_bounds): Likewise.
20294 (expand_builtin_memset_with_bounds): Likewise.
20295 (expand_builtin_memset_args): Likewise.
20296 (std_expand_builtin_va_start): Likewise.
20297 (expand_builtin): Likewise.
20298 (expand_builtin_with_bounds): Likewise.
20299 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
20300 (DEF_LIB_BUILTIN_CHKP): Likewise.
20301 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
20302 (DEF_CHKP_BUILTIN): Likewise.
20303 (BUILT_IN_MEMCPY): Likewise.
20304 (BUILT_IN_MEMMOVE): Likewise.
20305 (BUILT_IN_MEMPCPY): Likewise.
20306 (BUILT_IN_MEMSET): Likewise.
20307 (BUILT_IN_STPCPY): Likewise.
20308 (BUILT_IN_STRCAT): Likewise.
20309 (BUILT_IN_STRCHR): Likewise.
20310 (BUILT_IN_STRCPY): Likewise.
20311 (BUILT_IN_STRLEN): Likewise.
20312 (BUILT_IN_MEMCPY_CHK): Likewise.
20313 (BUILT_IN_MEMMOVE_CHK): Likewise.
20314 (BUILT_IN_MEMPCPY_CHK): Likewise.
20315 (BUILT_IN_MEMSET_CHK): Likewise.
20316 (BUILT_IN_STPCPY_CHK): Likewise.
20317 (BUILT_IN_STRCAT_CHK): Likewise.
20318 (BUILT_IN_STRCPY_CHK): Likewise.
20319 * calls.c (store_bounds): Likewise.
20320 (emit_call_1): Likewise.
20321 (special_function_p): Likewise.
20322 (maybe_warn_nonstring_arg): Likewise.
20323 (initialize_argument_information): Likewise.
20324 (finalize_must_preallocate): Likewise.
20325 (compute_argument_addresses): Likewise.
20326 (expand_call): Likewise.
20327 * cfgexpand.c (expand_call_stmt): Likewise.
20328 (expand_return): Likewise.
20329 (expand_gimple_stmt_1): Likewise.
20330 (pass_expand::execute): Likewise.
20331 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
20332 (cgraph_node::remove): Likewise.
20333 (cgraph_node::dump): Likewise.
20334 (cgraph_node::verify_node): Likewise.
20335 * cgraph.h (chkp_function_instrumented_p): Likewise.
20336 (symtab_node::get_alias_target): Likewise.
20337 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
20338 (cgraph_local_p): Likewise.
20339 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
20340 (cgraph_edge::rebuild_references): Likewise.
20341 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
20342 (walk_polymorphic_call_targets): Likewise.
20343 (cgraph_node::expand_thunk): Likewise.
20344 (symbol_table::output_weakrefs): Likewise.
20345 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
20346 (ix86_handle_option): Likewise.
20347 * config/i386/constraints.md: Likewise.
20348 * config/i386/i386-builtin-types.def (BND): Likewise.
20352 * config/i386/i386-builtin.def (BDESC_END): Likewise.
20353 (BDESC_FIRST): Likewise.
20355 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
20356 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
20357 * config/i386/i386.c (enum reg_class): Likewise.
20358 (ix86_target_string): Likewise.
20359 (ix86_option_override_internal): Likewise.
20360 (ix86_conditional_register_usage): Likewise.
20361 (ix86_valid_target_attribute_inner_p): Likewise.
20362 (ix86_set_indirect_branch_type): Likewise.
20363 (ix86_set_current_function): Likewise.
20364 (ix86_function_arg_regno_p): Likewise.
20365 (init_cumulative_args): Likewise.
20366 (ix86_function_arg_advance): Likewise.
20367 (ix86_function_arg): Likewise.
20368 (ix86_pass_by_reference): Likewise.
20369 (ix86_function_value_regno_p): Likewise.
20370 (ix86_function_value_1): Likewise.
20371 (ix86_function_value_bounds): Likewise.
20372 (ix86_return_in_memory): Likewise.
20373 (ix86_setup_incoming_vararg_bounds): Likewise.
20374 (ix86_va_start): Likewise.
20375 (indirect_thunk_need_prefix): Likewise.
20376 (print_reg): Likewise.
20377 (ix86_print_operand): Likewise.
20378 (ix86_expand_call): Likewise.
20379 (ix86_output_function_return): Likewise.
20380 (reg_encoded_number): Likewise.
20381 (BDESC_VERIFYS): Likewise.
20382 (ix86_init_mpx_builtins): Likewise.
20383 (ix86_init_builtins): Likewise.
20384 (ix86_emit_cmove): Likewise.
20385 (ix86_emit_move_max): Likewise.
20386 (ix86_expand_builtin): Likewise.
20387 (ix86_builtin_mpx_function): Likewise.
20388 (ix86_get_arg_address_for_bt): Likewise.
20389 (ix86_load_bounds): Likewise.
20390 (ix86_store_bounds): Likewise.
20391 (ix86_load_returned_bounds): Likewise.
20392 (ix86_store_returned_bounds): Likewise.
20393 (ix86_class_likely_spilled_p): Likewise.
20394 (ix86_hard_regno_mode_ok): Likewise.
20395 (x86_order_regs_for_local_alloc): Likewise.
20396 (ix86_mitigate_rop): Likewise.
20397 (ix86_bnd_prefixed_insn_p): Likewise.
20398 (ix86_mpx_bound_mode): Likewise.
20399 (ix86_make_bounds_constant): Likewise.
20400 (ix86_initialize_bounds): Likewise.
20401 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
20402 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
20403 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
20404 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
20405 (TARGET_CHKP_BOUND_MODE): Likewise.
20406 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
20407 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
20408 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
20409 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
20410 * config/i386/i386.h (TARGET_MPX): Likewise.
20411 (TARGET_MPX_P): Likewise.
20412 (VALID_BND_REG_MODE): Likewise.
20413 (FIRST_BND_REG): Likewise.
20414 (LAST_BND_REG): Likewise.
20415 (enum reg_class): Likewise.
20416 (BND_REG_P): Likewise.
20417 (BND_REGNO_P): Likewise.
20418 (BNDmode): Likewise.
20419 (ADJUST_INSN_LENGTH): Likewise.
20420 * config/i386/i386.md: Likewise.
20421 * config/i386/i386.opt: Likewise.
20422 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
20423 (defined): Likewise.
20424 (LINK_MPX): Likewise.
20425 (MPX_SPEC): Likewise.
20426 (LIBMPX_SPEC): Likewise.
20427 (LIBMPXWRAPPERS_SPEC): Likewise.
20428 (CHKP_SPEC): Likewise.
20429 * config/i386/predicates.md: Likewise.
20430 * dbxout.c (dbxout_type): Likewise.
20431 * doc/extend.texi: Likewise.
20432 * doc/invoke.texi: Likewise.
20433 * doc/md.texi: Likewise.
20434 * doc/tm.texi: Likewise.
20435 * doc/tm.texi.in: Likewise.
20436 * dwarf2out.c (is_base_type): Likewise.
20437 (gen_formal_types_die): Likewise.
20438 (gen_subprogram_die): Likewise.
20439 (gen_type_die_with_usage): Likewise.
20440 (gen_decl_die): Likewise.
20441 (dwarf2out_late_global_decl): Likewise.
20442 * expr.c (expand_assignment): Likewise.
20443 (emit_storent_insn): Likewise.
20444 (store_expr_with_bounds): Likewise.
20445 (store_expr): Likewise.
20446 (expand_expr_real_1): Likewise.
20447 * expr.h (store_expr_with_bounds): Likewise.
20448 * function.c (use_register_for_decl): Likewise.
20449 (struct bounds_parm_data): Likewise.
20450 (assign_parms_augmented_arg_list): Likewise.
20451 (assign_parm_find_entry_rtl): Likewise.
20452 (assign_parm_is_stack_parm): Likewise.
20453 (assign_parm_load_bounds): Likewise.
20454 (assign_bounds): Likewise.
20455 (assign_parms): Likewise.
20456 (expand_function_start): Likewise.
20457 * gcc.c (CHKP_SPEC): Likewise.
20458 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
20459 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
20460 (wrestrict_dom_walker::check_call): Likewise.
20461 * gimple.c (gimple_build_call_from_tree): Likewise.
20462 * gimple.h (enum gf_mask): Likewise.
20463 (gimple_call_with_bounds_p): Likewise.
20464 (gimple_call_set_with_bounds): Likewise.
20465 * gimplify.c (gimplify_init_constructor): Likewise.
20466 * ipa-cp.c (initialize_node_lattices): Likewise.
20467 (propagate_constants_across_call): Likewise.
20468 (find_more_scalar_values_for_callers_subset): Likewise.
20469 * ipa-hsa.c (process_hsa_functions): Likewise.
20470 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
20471 * ipa-icf.c (sem_function::merge): Likewise.
20472 * ipa-inline.c (early_inliner): Likewise.
20473 * ipa-pure-const.c (warn_function_noreturn): Likewise.
20474 (warn_function_cold): Likewise.
20475 (propagate_pure_const): Likewise.
20476 * ipa-ref.h (enum GTY): Likewise.
20477 * ipa-split.c (find_retbnd): Likewise.
20478 (consider_split): Likewise.
20479 (split_function): Likewise.
20480 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
20481 * ipa.c (walk_polymorphic_call_targets): Likewise.
20482 (symbol_table::remove_unreachable_nodes): Likewise.
20483 (process_references): Likewise.
20484 (cgraph_build_static_cdtor_1): Likewise.
20485 * lto-cgraph.c (lto_output_node): Likewise.
20486 (output_refs): Likewise.
20487 (compute_ltrans_boundary): Likewise.
20488 (input_overwrite_node): Likewise.
20489 (input_node): Likewise.
20490 (input_cgraph_1): Likewise.
20491 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
20492 * passes.c (pass_manager::execute_early_local_passes): Likewise.
20493 (class pass_chkp_instrumentation_passes): Likewise.
20494 (make_pass_chkp_instrumentation_passes): Likewise.
20495 * passes.def: Likewise.
20496 * rtl.h (struct GTY): Likewise.
20497 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
20498 * stor-layout.c (layout_type): Likewise.
20499 * symtab.c: Likewise.
20500 * target.def: Likewise.
20501 * targhooks.c (default_chkp_bound_type): Likewise.
20502 (default_chkp_bound_mode): Likewise.
20503 (default_builtin_chkp_function): Likewise.
20504 (default_chkp_function_value_bounds): Likewise.
20505 (default_chkp_make_bounds_constant): Likewise.
20506 (default_chkp_initialize_bounds): Likewise.
20507 * targhooks.h (default_chkp_bound_type): Likewise.
20508 (default_chkp_bound_mode): Likewise.
20509 (default_builtin_chkp_function): Likewise.
20510 (default_chkp_function_value_bounds): Likewise.
20511 (default_chkp_make_bounds_constant): Likewise.
20512 (default_chkp_initialize_bounds): Likewise.
20513 * toplev.c (compile_file): Likewise.
20514 (process_options): Likewise.
20515 * tree-core.h (DEF_BUILTIN): Likewise.
20516 (DEF_BUILTIN_CHKP): Likewise.
20517 * tree-inline.c (declare_return_variable): Likewise.
20518 (remap_gimple_stmt): Likewise.
20519 (copy_bb): Likewise.
20520 (initialize_inlined_parameters): Likewise.
20521 (expand_call_inline): Likewise.
20522 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
20523 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
20524 (make_pass_ipa_chkp_produce_thunks): Likewise.
20525 (make_pass_chkp): Likewise.
20526 (make_pass_chkp_opt): Likewise.
20527 (make_pass_chkp_instrumentation_passes): Likewise.
20528 * tree-pretty-print.c (dump_generic_node): Likewise.
20529 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
20530 * tree-ssa-dce.c (propagate_necessity): Likewise.
20531 (eliminate_unnecessary_stmts): Likewise.
20532 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
20533 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
20534 * tree-ssa-sccvn.h: Likewise.
20535 * tree-ssa-strlen.c (get_string_length): Likewise.
20536 (valid_builtin_call): Likewise.
20537 (adjust_last_stmt): Likewise.
20538 (handle_builtin_strchr): Likewise.
20539 (handle_builtin_strcpy): Likewise.
20540 (handle_builtin_stxncpy): Likewise.
20541 (handle_builtin_memcpy): Likewise.
20542 (handle_builtin_strcat): Likewise.
20543 (strlen_check_and_optimize_stmt): Likewise.
20544 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
20545 * tree-streamer-in.c: Likewise.
20546 * tree-streamer.c (record_common_node): Likewise.
20547 * tree.c (tree_code_size): Likewise.
20548 (wide_int_to_tree_1): Likewise.
20549 (type_contains_placeholder_1): Likewise.
20550 (build_common_tree_nodes): Likewise.
20551 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
20552 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
20553 (POINTER_BOUNDS_P): Likewise.
20554 (BOUNDED_TYPE_P): Likewise.
20555 (BOUNDED_P): Likewise.
20556 (CALL_WITH_BOUNDS_P): Likewise.
20557 (pointer_bounds_type_node): Likewise.
20558 * value-prof.c (gimple_ic): Likewise.
20559 * var-tracking.c (vt_add_function_parameters): Likewise.
20560 * varasm.c (make_decl_rtl): Likewise.
20561 (assemble_start_function): Likewise.
20562 (output_constant): Likewise.
20563 (maybe_assemble_visibility): Likewise.
20564 * varpool.c (ctor_for_folding): Likewise.
20565 * chkp-builtins.def: Remove.
20566 * ipa-chkp.c: Remove.
20567 * ipa-chkp.h: Remove.
20568 * rtl-chkp.c: Remove.
20569 * rtl-chkp.h: Remove.
20570 * tree-chkp-opt.c: Remove.
20571 * tree-chkp.c: Remove.
20572 * tree-chkp.h: Remove.
20574 2018-06-07 Carl Love <cel@us.ibm.com>
20576 * config/rs6000/vsx.md (vextract_fp_from_shorth,
20577 vextract_fp_from_shortl): Add BE support.
20579 2018-06-07 Paul Koning <ni1d@arrl.net>
20581 * compare-elim.c (try_merge_compare): Don't merge compare if
20582 address contains a side effect.
20583 (try_eliminate_compare): Likewise.
20585 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
20587 * config.gcc: Support "tremont".
20588 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
20589 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20591 * config/i386/i386.c (m_TREMONT): Define.
20592 (processor_target_table): Add "tremont".
20593 (PTA_TREMONT): Define.
20594 (ix86_lea_outperforms): Add TARGET_TREMONT.
20595 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
20596 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
20597 and M_INTEL_GOLDMONT_PLUS.
20598 (fold_builtin_cpu): Add "tremont".
20599 (ix86_add_stmt_cost): Add TARGET_TREMONT.
20600 (ix86_option_override_internal): Add "tremont".
20601 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
20602 (processor_type): Add PROCESSOR_TREMONT.
20603 * config/i386/x86-tune.def: Add m_TREMONT.
20604 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
20606 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
20608 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
20609 symbol defined for msp430i* devices to be lower case.
20611 2018-06-07 Richard Biener <rguenther@suse.de>
20613 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
20614 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
20615 Properly wrap signed arithmetic if overflow wraps.
20617 2018-06-07 Jakub Jelinek <jakub@redhat.com>
20619 PR tree-optimization/69615
20620 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
20621 of a cast from a same precision integral SSA_NAME in a bb dominated
20622 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
20623 cast to utype if rhs2 has already a compatible type.
20625 2018-06-07 Richard Biener <rguenther@suse.de>
20627 PR tree-optimization/85935
20628 * graphite-scop-detection.c (find_params_in_bb): Analyze
20629 condition operands with respect to the correct loop. Assert
20630 the analysis doesn't fail.
20632 2018-06-04 Carl Love <cel@us.ibm.com>
20634 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
20635 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
20636 as it is slightly cheaper.
20637 (first_match_or_eos_index_<mode>):
20638 Calculate index using natural element order.
20639 (first_match_index_<mode>):
20640 Calculate index using natural element order.
20641 (first_match_or_eos_index_<mode>):
20642 Calculate index using natural order.
20643 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
20644 for BE and LE modes.
20645 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
20646 P9V_BUILTIN_VCLZLSBB_V16QI.
20647 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
20650 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
20652 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
20653 indentation and line wrap for many prototypes. Add missing
20654 @smallexample directives around block of prototypes for vec_xl and
20657 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
20659 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
20660 track if we pass or return IEEE 128-bit floating point.
20661 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
20662 C++ mangling that is compatible with GCC 8.1.
20663 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
20664 (init_cumulative_args): Note if we pass or return IEEE 128-bit
20665 floating point types.
20666 (rs6000_function_arg_advance_1): Likewise.
20667 (rs6000_mangle_type): Optionally generate mangled names that match
20668 what GCC 8.1 generated for IEEE 128-bit floating point types.
20669 (rs6000_globalize_decl_name): If we have an external function that
20670 passes or returns IEEE 128-bit types, generate a weak reference
20671 from the mangled name used in GCC 8.1 to the current mangled
20673 (rs6000_init_builtins): Make __ibm128 use the long double type if
20674 long double is IBM extended double. Make __float128 use the long
20675 double type if long double is IEEE 128-bit.
20678 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
20679 macro for __ibm128 built-in functions.
20680 (PACK_IF): Add __ibm128 pack/unpack functions.
20681 (UNPACK_IF): Likewise.
20682 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
20683 enable long double built-in functions if long double is IEEE
20684 128-bit floating point.
20685 (rs6000_invalid_builtin): Update long double built-in function
20687 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
20688 functions, adjust the built-in function to use the long double
20689 built-in function if __ibm128 and long double are the same type.
20690 * doc/extend.texi (PowerPC builtins): Update documention for
20691 __builtin_{,un}pack_longdouble. Add documentation for
20692 __builtin_{,un}pack_ibm128.
20694 2018-06-06 Jim Wilson <jimw@sifive.com>
20696 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
20697 (struct machine_function): New field interrupt_mode.
20698 (riscv_handle_type_attribute): New function. Add forward declaration.
20699 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
20700 (riscv_expand_epilogue): Check interrupt_mode field.
20701 (riscv_set_current_function): Check interrupt attribute args and
20702 set interrupt_mode field.
20703 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
20704 (riscv_sret, riscv_uret): New.
20705 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
20706 new arguments to interrupt attribute.
20708 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
20711 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
20712 Don't handle -mcpu=power8 if -mpower9-vector is also used.
20714 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
20716 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
20717 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
20718 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
20719 several redundant entries.
20721 2018-06-06 David Malcolm <dmalcolm@redhat.com>
20723 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
20724 type from "rtx" to "rtx_insn *".
20725 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
20726 for local "call_insn", removing cast.
20727 (ix86_expand_call): Likewise, introducing a "call_insn" local.
20729 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
20731 PR tree-optimization/86066
20732 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
20733 for BIT_INSERT_EXPR stores.
20735 2018-06-06 Richard Biener <rguenther@suse.de>
20737 PR tree-optimization/86062
20738 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
20739 component refs ontop
20740 of to be offsetted base.
20742 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
20744 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
20745 to be static and remove check on interrupt attribute name.
20747 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
20749 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
20750 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
20752 2018-06-05 Steve Ellcey <sellcey@cavium.com>
20755 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
20757 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
20758 Remove second argument, change how error is called.
20759 (aarch64_layout_arg): Remove second argument from
20760 aarch64_err_no_fpadvsimd call.
20761 (aarch64_init_cumulative_args): Ditto.
20762 (aarch64_gimplify_va_arg_expr): Ditto.
20763 * config/aarch64/aarch64.md (mov<mode>): Ditto.
20765 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
20767 * config/i386/i386.md (simple_return_indirect_internal): New expander.
20768 (*simple_return_indirect_internal<mode>): Rename from
20769 simple_return_indirect_internal. Use W mode iterator.
20770 (rstorssp): New expander.
20771 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
20772 (clrssbsy): New expander.
20773 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
20775 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
20777 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
20779 (cmse_check_pointed_object): Likewise.
20781 2018-06-05 Martin Liska <mliska@suse.cz>
20783 PR gcov-profile/47618
20784 * doc/invoke.texi: Document how -fprofile-dir format
20787 2018-06-05 Richard Biener <rguenther@suse.de>
20789 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
20790 removal pretend DOM info isn't available so we do not update
20791 it and only remove edges, not dominated blocks. Actually free
20792 DOM info in case we removed something. Remove unreachable blocks.
20793 (mfb_keep_latches): Work with either DOM info or marked backedges.
20794 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
20795 first. Mark backedges if DOM info isn't available.
20796 (Re-)compute DOM info after cleanup_control_flow_pre.
20798 2018-06-05 Richard Biener <rguenther@suse.de>
20800 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
20801 (locus_discrim_hasher::hash): Adjust.
20802 (locus_discrim_hasher::equal): Likewise.
20803 (next_discriminator_for_locus): Work on line directly.
20804 (same_line_p): Pass in expanded locus1 as well.
20805 (assign_discriminators): Avoid redundant location expansions.
20807 2018-06-05 Richard Biener <rguenther@suse.de>
20809 PR tree-optimization/86046
20810 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
20811 if required after clearing TREE_ADDRESSABLE.
20813 2018-06-05 Richard Biener <rguenther@suse.de>
20815 PR tree-optimization/86047
20816 * tree-ssa-loop.c (for_each_index): Glob handling of all
20817 decls and constants and really handle all of them.
20819 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20822 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
20823 qualifier_void_pointer and qualifier_const_void_pointer.
20824 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
20825 (arm_init_builtins): Handle the above.
20826 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
20827 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
20828 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
20831 2018-06-05 Martin Liska <mliska@suse.cz>
20833 * auto-profile.c (read_autofdo_file): Do not use
20834 gcov_ctr_summary struct.
20835 (afdo_callsite_hot_enough_for_early_inline): Likewise.
20836 * coverage.c (struct counts_entry): Likewise.
20837 (read_counts_file): Read just single summary entry.
20838 (get_coverage_counts): Use gcov_summary struct.
20839 * coverage.h (get_coverage_counts): Likewise.
20840 * gcov-dump.c (dump_working_sets): Likewise.
20841 (tag_summary): Dump just single summary.
20842 * gcov-io.c (gcov_write_summary): Write just histogram
20844 (gcov_read_summary): Read just single summary.
20845 (compute_working_sets): Use gcov_summary struct.
20846 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
20847 of GCOV_COUNTERS_SUMMABLE.
20848 (GCOV_COUNTERS_SUMMABLE): Remove.
20849 (GCOV_FIRST_VALUE_COUNTER): Replace with
20850 GCOV_COUNTER_V_INTERVAL.
20851 (struct gcov_ctr_summary): Remove.
20852 (struct gcov_summary): Directly use fields of former
20854 (compute_working_sets): Use gcov_summary struct.
20855 * gcov.c (read_count_file): Do not use ctrs fields.
20856 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
20858 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
20860 * profile.c: Likewise.
20861 * profile.h: Likewise.
20863 2018-06-05 Martin Liska <mliska@suse.cz>
20865 PR gcov-profile/84846
20866 * gcov.c (output_lines): Print working directory only
20867 in intermediate format.
20869 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
20871 * config/s390/s390-builtin-types.def: Add void function type.
20872 * config/s390/s390-builtins.def: Use the function type for the
20875 2018-06-04 Jim Wilson <jimw@sifive.com>
20877 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
20879 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
20880 and maybe_eh_return. Change regno to unsigned int. Use new args to
20881 handle EH_RETURN_DATA_REGNO registers properly.
20882 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
20883 (riscv_expand_epilogue): Update comment. Change argument name and
20884 type. Update code to use new name and type. Pass new args to
20885 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
20887 * config/riscv/riscv.md (NORMAL_RETURN): New.
20888 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
20889 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
20890 (eh_return): Call gen_eh_return_internal and emit barrier.
20891 (eh_return_internal): Call riscv_expand_epilogue.
20893 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
20895 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
20896 bit_insertion field and declare can_be_merged_into method.
20897 (merged_store_group::can_be_merged_into): New method.
20898 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
20899 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
20900 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
20902 2018-06-04 Richard Biener <rguenther@suse.de>
20904 PR tree-optimization/85955
20905 * builtins.c (fold_builtin_sincos): Convert pointers to
20906 destination to appropriate type before dereferencing.
20908 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
20910 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
20912 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
20914 * expr.c (expand_expr_real_1): Force the operand into memory if
20915 its TYPE_MODE is BLKmode and if there is no integer mode for
20916 the number of bits being extracted.
20918 2018-06-04 Jakub Jelinek <jakub@redhat.com>
20922 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
20923 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
20925 2018-06-04 Richard Biener <rguenther@suse.de>
20927 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
20928 (cleanup_tree_cfg_noloop): ... single caller. Do
20929 start_recording_case_labels later.
20931 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
20933 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
20934 to _IMMINTRIN_H_INCLUDED.
20935 * config/i386/pconfigintrin.h: Ditto.
20936 * config/i386/waitpkgintrin.h: Ditto.
20937 * config/i386/immintrin.h: Add includes for sgxintrin.h,
20938 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
20939 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
20940 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
20941 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
20942 waitpkgintrin.h and cldemoteintrin.h.
20944 2018-06-04 Richard Biener <rguenther@suse.de>
20946 PR tree-optimization/86038
20947 * tracer.c (find_best_successor): Check probability for
20948 being initialized, bail out if not.
20950 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
20953 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
20954 of bits to ignore when comparing architectures.
20956 2018-06-04 Jakub Jelinek <jakub@redhat.com>
20958 PR tree-optimization/69615
20959 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
20960 maximum or minimum of the type, try to merge it also as if
20961 range1 is + [-, x - 1] or + [x + 1, -].
20964 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
20966 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
20968 PR tree-optimization/86034
20969 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
20970 the unsigned bitfield type in a bit insertion sequence if it does not
20971 have a larger precision than the bitfield size.
20972 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
20974 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
20976 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
20978 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
20980 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
20981 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
20982 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
20983 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
20985 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
20987 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
20988 Disable -fdelete-null-pointer-checks for ELF toolchain.
20990 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
20991 Kito Cheng <kito.cheng@gmail.com>
20993 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
20994 (nds32le-*-*, nds32be-*-*): Integrate checking process.
20995 (nds32*-*-*): Add glibc and uclibc conditions.
20996 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
20997 (TARGET_EXCEPT_UNWIND_INFO): Define.
20998 * config/nds32/elf.h: New file.
20999 * config/nds32/linux.h: New file.
21000 * config/nds32/nds32-elf.opt: New file.
21001 * config/nds32/nds32-linux.opt: New file.
21002 * config/nds32/nds32-fp-as-gp.c
21003 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
21004 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
21006 (nds32_asm_file_end): Ditto.
21007 (nds32_print_operand): Ditto.
21008 (nds32_insert_attributes): Ditto.
21009 (nds32_init_libfuncs): New function.
21010 (TARGET_HAVE_TLS): Define.
21011 (TARGET_INIT_LIBFUNCS): Define.
21012 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
21014 (TARGET_ELF): Apply different mcmodel setting.
21015 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
21016 been migrated into elf.h and linux.h files.
21017 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
21018 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
21019 (mcmodel): The content has been migrated into nds32-elf.opt and
21020 nds32-linux.opt files.
21021 * config/nds32/t-elf: New file.
21022 * config/nds32/t-linux: New file.
21024 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
21025 Shiva Chen <shiva0217@gmail.com>
21027 * config/nds32/constants.md (unspec_volatile_element): Add
21028 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
21029 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
21031 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
21032 (make_pass_nds32_fp_as_gp): Declare.
21033 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
21035 (nds32_asm_function_end_prologue): Remove unused asm output.
21036 (nds32_asm_function_begin_epilogue): Remove unused asm output.
21037 (nds32_asm_file_start): Output necessary fp_as_gp information.
21038 (nds32_option_override): Adjust register usage.
21039 (nds32_expand_prologue): Consider fp_as_gp situation.
21040 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
21041 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
21044 (simple_return): Ditto.
21045 (omit_fp_begin): Output special directive for fp_as_gp.
21046 (omit_fp_end): Output special directive for fp_as_gp.
21047 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
21048 mforbid-fp-as-gp): New options.
21050 2018-06-01 Mark Wielaard <mark@klomp.org>
21052 * dwarf2out.c (dwarf2out_finish): Remove generation of
21053 DW_AT_loclists_base.
21055 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
21057 * gimple-ssa-store-merging.c: Include gimple-fold.h.
21058 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
21059 (struct merged_store_group): Add bit_insertion field.
21060 (dump_char_array): Use standard hexadecimal format.
21061 (merged_store_group::merged_store_group): Set bit_insertion to false.
21062 (merged_store_group::apply_stores): Use optimal buffer size. Deal
21063 with BIT_INSERT_EXPR stores. Move up code updating the mask and
21064 also print the mask in the dump file.
21065 (pass_store_merging::gate): Minor tweak.
21066 (imm_store_chain_info::coalesce_immediate): Fix wrong association
21067 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
21068 stores with INTEGER_CST stores.
21069 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
21070 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
21071 and use it throughout. Generate bit insertion sequences if need be.
21072 (pass_store_merging::process_store): Remove redundant condition.
21073 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
21075 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
21077 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
21078 the 128-bit floating point types. Fix function comment.
21080 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21082 * config/aarch64/aarch64-simd.md
21083 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
21085 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
21088 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
21090 PR tree-optimization/85989
21091 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
21093 (backprop::intersect_uses): Check it when deciding whether this
21094 is a backedge reference.
21095 (backprop::process_block): Add each phi to m_visited_phis
21096 after visiting it, then clear it at the end.
21098 2018-06-01 Richard Biener <rguenther@suse.de>
21100 * tree-vectorizer.h (vect_dr_stmt): New function.
21101 (vect_get_load_cost): Adjust.
21102 (vect_get_store_cost): Likewise.
21103 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
21104 Use vect_dr_stmt instead of DR_SMTT.
21105 (vect_record_base_alignments): Likewise.
21106 (vect_calculate_target_alignment): Likewise.
21107 (vect_compute_data_ref_alignment): Likewise and make static.
21108 (vect_update_misalignment_for_peel): Likewise.
21109 (vect_verify_datarefs_alignment): Likewise.
21110 (vector_alignment_reachable_p): Likewise.
21111 (vect_get_data_access_cost): Likewise. Pass down
21112 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
21113 (vect_get_peeling_costs_all_drs): Likewise.
21114 (vect_peeling_hash_get_lowest_cost): Likewise.
21115 (vect_enhance_data_refs_alignment): Likewise.
21116 (vect_find_same_alignment_drs): Likewise.
21117 (vect_analyze_data_refs_alignment): Likewise.
21118 (vect_analyze_group_access_1): Likewise.
21119 (vect_analyze_group_access): Likewise.
21120 (vect_analyze_data_ref_access): Likewise.
21121 (vect_analyze_data_ref_accesses): Likewise.
21122 (vect_vfa_segment_size): Likewise.
21123 (vect_small_gap_p): Likewise.
21124 (vectorizable_with_step_bound_p): Likewise.
21125 (vect_prune_runtime_alias_test_list): Likewise.
21126 (vect_analyze_data_refs): Likewise.
21127 (vect_supportable_dr_alignment): Likewise.
21128 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
21129 (vect_gen_prolog_loop_niters): Likewise.
21130 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
21131 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
21133 (vect_recog_mask_conversion_pattern): Likewise.
21134 (vect_try_gather_scatter_pattern): Likewise.
21135 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
21136 to vect_get_store_cost.
21137 (vect_get_store_cost): Get stmt_info instead of DR.
21138 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
21139 (vect_get_load_cost): Get stmt_info instead of DR.
21141 2018-06-01 Richard Biener <rguenther@suse.de>
21143 PR middle-end/86017
21144 * gimple-fold.c (var_decl_component_p): Also allow offsetted
21145 vars wrapped in MEM_REFs.
21147 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
21149 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
21150 Fix subreg tests so that we only return a choice between
21151 GENERAL_REGS and FP_REGS if the original classes included both.
21153 2018-06-01 Richard Biener <rguenther@suse.de>
21156 * tree-ssa-structalias.c (get_function_part_constraint):
21157 Handle NULL fi->decl.
21158 (find_func_aliases_for_call): Properly handle indirect
21159 fi from direct call.
21160 (find_func_clobbers): Likewise.
21161 (ipa_pta_execute): Likewise.
21162 (create_variable_info_for): For functions that are ifunc_resolver
21163 resolve to a varinfo that contains the result of the resolver call.
21164 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
21167 2018-05-31 Michael Collison <michael.collison@arm.com>
21169 * config/aarch64/aarch64.md:
21170 (*fix_to_zero_extenddfdi2): New pattern.
21171 * gcc.target/aarch64/fix_extend1.c: New testcase.
21173 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
21175 PR middle-end/78809
21176 PR middle-end/83026
21177 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
21178 and BUILT_IN_STRNCMP_EQ.
21179 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
21180 BUILT_IN_STRNCMP_EQ.
21181 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
21182 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
21183 (gimple_fold_builtin): Likewise.
21184 * tree-ssa-strlen.c (compute_string_length): New function.
21185 (determine_min_obsize): New function.
21186 (handle_builtin_string_cmp): New function to handle calls to
21187 string compare functions.
21188 (strlen_optimize_stmt): Add handling to builtin string compare
21190 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
21191 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
21192 * tree.c (build_common_builtin_nodes): Add new defines of
21193 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
21195 2018-05-31 Jakub Jelinek <jakub@redhat.com>
21198 * bb-reorder.c (pass_partition_blocks::gate): Return false for
21199 functions with naked attribute.
21201 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
21203 * config/i386/sse.md (avx_vec_concat<mode>):
21204 Substitute concat_tg_mode mode attribute with xtg_mode.
21205 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
21206 (concat_tg_mode): Remove mode attribute.
21208 2018-05-31 Martin Sebor <msebor@redhat.com>
21211 * calls.c (alloc_max_size): Correct a logic error/typo.
21212 Treat excessive arguments as infinite. Warn for invalid arguments.
21213 * doc/invoke.texi (-Walloc-size-larger-than): Update.
21215 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
21218 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
21219 and movx for Haswell.
21221 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
21222 Cesar Philippidis <cesar@codesourcery.com>
21224 PR middle-end/85879
21225 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
21226 when emitting error on private/firstprivate reductions.
21227 * omp-low.c (lower_omp_target): Avoid reference-type processing
21228 on pointers for firstprivate clause.
21230 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
21232 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
21235 * config/aarch64/aarch64-simd.md
21236 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
21237 (aarch64_ld1_x3_<mode>): Likewise
21238 (aarch64_st1x2<VALLDIF:mode>): Likewise
21239 (aarch64_st1_x2_<mode>): Likewise
21240 (aarch64_st1x3<VALLDIF:mode>): Likewise
21241 (aarch64_st1_x3_<mode>): Likewise
21242 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
21243 (vld1_s8_x3): Likewise.
21244 (vld1_u16_x3): Likewise.
21245 (vld1_s16_x3): Likewise.
21246 (vld1_u32_x3): Likewise.
21247 (vld1_s32_x3): Likewise.
21248 (vld1_u64_x3): Likewise.
21249 (vld1_s64_x3): Likewise.
21250 (vld1_f16_x3): Likewise.
21251 (vld1_f32_x3): Likewise.
21252 (vld1_f64_x3): Likewise.
21253 (vld1_p8_x3): Likewise.
21254 (vld1_p16_x3): Likewise.
21255 (vld1_p64_x3): Likewise.
21256 (vld1q_u8_x3): Likewise.
21257 (vld1q_s8_x3): Likewise.
21258 (vld1q_u16_x3): Likewise.
21259 (vld1q_s16_x3): Likewise.
21260 (vld1q_u32_x3): Likewise.
21261 (vld1q_s32_x3): Likewise.
21262 (vld1q_u64_x3): Likewise.
21263 (vld1q_s64_x3): Likewise.
21264 (vld1q_f16_x3): Likewise.
21265 (vld1q_f32_x3): Likewise.
21266 (vld1q_f64_x3): Likewise.
21267 (vld1q_p8_x3): Likewise.
21268 (vld1q_p16_x3): Likewise.
21269 (vld1q_p64_x3): Likewise.
21270 (vst1_s64_x2): Likewise.
21271 (vst1_u64_x2): Likewise.
21272 (vst1_f64_x2): Likewise.
21273 (vst1_s8_x2): Likewise.
21274 (vst1_p8_x2): Likewise.
21275 (vst1_s16_x2): Likewise.
21276 (vst1_p16_x2): Likewise.
21277 (vst1_s32_x2): Likewise.
21278 (vst1_u8_x2): Likewise.
21279 (vst1_u16_x2): Likewise.
21280 (vst1_u32_x2): Likewise.
21281 (vst1_f16_x2): Likewise.
21282 (vst1_f32_x2): Likewise.
21283 (vst1_p64_x2): Likewise.
21284 (vst1q_s8_x2): Likewise.
21285 (vst1q_p8_x2): Likewise.
21286 (vst1q_s16_x2): Likewise.
21287 (vst1q_p16_x2): Likewise.
21288 (vst1q_s32_x2): Likewise.
21289 (vst1q_s64_x2): Likewise.
21290 (vst1q_u8_x2): Likewise.
21291 (vst1q_u16_x2): Likewise.
21292 (vst1q_u32_x2): Likewise.
21293 (vst1q_u64_x2): Likewise.
21294 (vst1q_f16_x2): Likewise.
21295 (vst1q_f32_x2): Likewise.
21296 (vst1q_f64_x2): Likewise.
21297 (vst1q_p64_x2): Likewise.
21298 (vst1_s64_x3): Likewise.
21299 (vst1_u64_x3): Likewise.
21300 (vst1_f64_x3): Likewise.
21301 (vst1_s8_x3): Likewise.
21302 (vst1_p8_x3): Likewise.
21303 (vst1_s16_x3): Likewise.
21304 (vst1_p16_x3): Likewise.
21305 (vst1_s32_x3): Likewise.
21306 (vst1_u8_x3): Likewise.
21307 (vst1_u16_x3): Likewise.
21308 (vst1_u32_x3): Likewise.
21309 (vst1_f16_x3): Likewise.
21310 (vst1_f32_x3): Likewise.
21311 (vst1_p64_x3): Likewise.
21312 (vst1q_s8_x3): Likewise.
21313 (vst1q_p8_x3): Likewise.
21314 (vst1q_s16_x3): Likewise.
21315 (vst1q_p16_x3): Likewise.
21316 (vst1q_s32_x3): Likewise.
21317 (vst1q_s64_x3): Likewise.
21318 (vst1q_u8_x3): Likewise.
21319 (vst1q_u16_x3): Likewise.
21320 (vst1q_u32_x3): Likewise.
21321 (vst1q_u64_x3): Likewise.
21322 (vst1q_f16_x3): Likewise.
21323 (vst1q_f32_x3): Likewise.
21324 (vst1q_f64_x3): Likewise.
21325 (vst1q_p64_x3): Likewise.
21327 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
21329 * config/msp430/msp430.c (msp430_output_labelref): Prepend
21330 user_label_prefix to name.
21332 * tree-core.h: Update comment about the format of NAME string
21333 passed to handler in attribute_spec.
21335 * config/msp430/msp430.md: Remove erroneous subreg expression from
21336 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
21337 zero_extend{q,h}isi2.
21339 2018-05-30 Borislav Petkov <bp@suse.de>
21341 * doc/extend.texi: Document some architecture specific
21342 constraints and sort entries.
21344 2018-05-30 Martin Sebor <msebor@redhat.com>
21346 PR middle-end/85369
21347 * builtins.c (expand_builtin_stpcpy_1): New function.
21348 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
21349 only if the former succeeds.
21351 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
21353 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
21356 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21358 * doc/invoke.texi (-flinker-output): Document
21360 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21362 * passes.c (ipa_write_summaries): Only modify statements if body
21364 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
21365 incrementally linking.
21366 (ipa_passes): Likewise.
21367 * lto-cgraph.c (lto_output_node): When incrementally linking do not
21368 pass down resolution info.
21369 * common.opt (flag_incremental_link): Update info.
21370 * gcc.c (plugin specs): Turn flinker-output=* to
21371 -plugin-opt=-linker-output-known
21372 * toplev.c (compile_file): Also cut compilation when doing incremental
21374 * flag-types. (enum lto_partition_model): Add
21375 LTO_LINKER_OUTPUT_NOLTOREL.
21376 (invoke.texi): Add -flinker-output docs.
21377 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
21378 link same way as WPA; do not stream in dead initializers.
21380 * dwarf2out.c (dwarf2out_die_ref_for_decl,
21381 darf2out_register_external_decl): Support incremental link.
21383 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21385 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
21387 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21389 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
21390 it down to simple_object_copy_lto_debug_sections.
21391 (run_gcc): Determine incremental LTO link time and configure
21392 lto1 into non-wpa mode, disable renaming of debug sections.
21394 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
21396 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
21397 descriptions of various incorrectly documented functions.
21399 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
21402 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
21403 address check not strict.
21405 2018-05-30 Richard Biener <rguenther@suse.de>
21407 PR tree-optimization/85964
21408 * tracer.c (better_p): Drop initialized count check, we only
21409 call the function with initialized counts now.
21410 (find_best_successor): Do find a best edge if one
21411 has uninitialized count.
21412 (find_best_predecessor): Likewise. Do BB frequency check only
21413 if count is initialized.
21415 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
21417 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
21418 (aarch64_ldrstr_offset_compare): New.
21419 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
21420 load/store orderings.
21421 (aarch64_gen_adjusted_ldpstp): Likewise.
21423 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
21425 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
21426 Check for subset of GENERAL_REGS and FP_REGS.
21427 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
21430 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
21432 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
21433 and wi::to_poly_offset. Add the current offset and then check
21434 whether the sum fits, rather than using an unchecked addition of
21435 a checked term. Check for a shwi rather than a uhwi.
21436 * expr.c (get_bit_range): Use tree_to_poly_uint64.
21437 (store_constructor): Use poly_int_tree_p.
21438 (expand_expr_real_1): Likewise.
21439 * function.c (assign_temp): Likewise.
21440 * fold-const.c (const_binop): Use poly_int_tree_p and
21441 wi::to_poly_offset.
21442 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
21444 * ipa-icf-gimple.c (func_checker::compare_operand): Use
21445 to_poly_offset for MEM offsets.
21446 * ipa-icf.c (sem_variable::equals): Likewise.
21447 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
21448 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
21449 wi::to_poly_offset for BIT_FIELD_REF offsets.
21450 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
21451 wi::to_poly_offset.
21452 * var-tracking.c (emit_note_insn_var_location): Use
21453 tree_to_poly_uint64.
21455 2018-05-29 Jim Wilson <jimw@sifive.com>
21457 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
21459 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
21462 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
21463 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
21465 (sse4_1_round<mode>2): Use nonimmediate_operand
21466 for operand 1 predicate.
21468 2018-05-29 Martin Sebor <msebor@redhat.com>
21469 Richard Biener <rguenther@suse.de>
21472 * calls.c (get_size_range): Call determine_value_range instead
21473 of get_value_range..
21474 * tree-vrp.h (determine_value_range): Declared new function.
21475 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
21477 2018-05-29 Richard Biener <rguenther@suse.de>
21479 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
21480 sure to use non-pattern stmts for get_earlier_stmt arguments.
21481 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
21482 called on pattern stmts.
21483 (get_later_stmt): Likewise.
21485 2018-05-29 Martin Liska <mliska@suse.cz>
21487 PR gcov-profile/85759
21488 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
21491 2018-05-29 Jakub Jelinek <jakub@redhat.com>
21493 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
21495 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
21499 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
21500 VEC_PACK_FLOAT_EXPR): New tree codes.
21501 * tree-pretty-print.c (op_code_prio): Handle
21502 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
21503 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
21504 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
21505 * tree-inline.c (estimate_operator_cost): Likewise.
21506 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
21507 * fold-const.c (const_binop): Likewise.
21508 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
21509 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
21510 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
21511 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
21512 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
21513 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
21514 * expr.c (expand_expr_real_2): Likewise.
21515 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
21516 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
21517 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
21519 * optabs.c (expand_widen_pattern_expr): For
21520 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
21521 sign from result type rather than operand's type.
21522 (expand_binop_directly): For vec_packu_float_optab and
21523 vec_packs_float_optab allow result type to be different from operand's
21525 * optabs-tree.c (optab_for_tree_code): Handle
21526 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
21527 VEC_PACK_FLOAT_EXPR. Formatting fixes.
21528 * tree-vect-generic.c (expand_vector_operations_1): Handle
21529 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
21530 VEC_PACK_FLOAT_EXPR.
21531 * tree-vect-stmts.c (supportable_widening_operation): Handle
21533 (supportable_narrowing_operation): Handle FLOAT_EXPR.
21534 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
21535 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
21536 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
21537 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
21539 (vec_pack<floatprefix>_float_<mode>): New expander.
21540 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
21542 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
21543 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
21544 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
21545 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
21546 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
21548 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
21549 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
21550 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
21551 VEC_PACK_FLOAT_EXPR): Document.
21553 2018-05-29 Richard Biener <rguenther@suse.de>
21555 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
21557 (stmt_vec_info_vec): Make pointer.
21558 (init_stmt_vec_info_vec): Remove.
21559 (free_stmt_vec_info_vec): Likewise.
21560 (set_stmt_vec_info_vec): New function.
21561 (free_stmt_vec_infos): Likewise.
21562 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
21563 (set_vinfo_for_stmt): Likewise.
21564 (get_earlier_stmt): Likewise.
21565 (get_later_stmt): Likewise.
21566 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
21567 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
21568 (vec_info::~vec_info): Free stmt_vec_infos.
21569 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
21570 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
21571 (pass_slp_vectorize::execute): Likewise.
21572 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
21573 (free_stmt_vec_info_vec): Likewise.
21574 (set_stmt_vec_info_vec): New function.
21575 (free_stmt_vec_infos): Likewise.
21576 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
21577 the global stmt_vec_info_vec.
21578 * tree-parloops.c (gather_scalar_reductions): Use
21579 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
21582 2018-05-29 Richard Biener <rguenther@suse.de>
21584 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
21586 2018-05-29 Martin Liska <mliska@suse.cz>
21587 David Malcolm <dmalcolm@redhat.com>
21589 * vec.c (test_reverse): New.
21590 (vec_c_tests): Add new test.
21591 * vec.h (vl_ptr>::reverse): New function.
21593 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
21595 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
21597 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
21600 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
21602 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
21604 2018-05-28 Richard Biener <rguenther@suse.de>
21606 PR tree-optimization/85933
21607 * tree-vect-data-refs.c (vect_record_base_alignments): Only
21608 look at stmts marked as vectorizable.
21610 2018-05-28 Richard Biener <rguenther@suse.de>
21612 PR tree-optimization/85934
21613 * tree-vect-generic.c (expand_vector_operations_1): Hoist
21614 vector boolean check before scalar optimization.
21616 2018-05-28 Jakub Jelinek <jakub@redhat.com>
21618 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
21621 2018-05-28 Mark Wielaard <mark@klomp.org>
21623 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
21624 if it is an expression containing a minus sign.
21626 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
21628 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
21630 2018-05-27 Paul Koning <ni1d@arrl.net>
21632 * config/pdp11/pdp11.md (truncsihi2): Remove.
21634 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
21635 Chung-Ju Wu <jasonwucj@gmail.com>
21637 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
21639 (unaligned_store_dw): Ditto.
21640 * config/nds32/nds32-memory-manipulation.c
21641 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
21642 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
21643 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
21644 (emit_setmem_word_loop): Rename to ...
21645 (emit_setmem_doubleword_loop): ... this.
21646 (nds32_gen_dup_4_byte_to_word_value): New function.
21647 (nds32_gen_dup_8_byte_to_double_word_value): New function.
21648 (nds32_expand_setmem_loop): Refine implementation.
21649 (nds32_expand_setmem_loop_v3m): Ditto.
21650 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
21653 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
21655 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
21657 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
21659 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
21660 (nds32_init_machine_status): Initialize machine->attr_naked_p and
21661 machine->attr_no_prologue_p.
21662 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
21663 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
21664 (nds32_expand_epilogue): Consider attr_naked_p.
21665 (nds32_expand_epilogue_v3pop): Likewise.
21666 (nds32_can_use_return_insn): Likewise.
21667 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
21668 attr_no_prologue_p fields.
21669 * config/nds32/nds32.opt (mret-in-naked-func): New option.
21671 2018-05-27 Jakub Jelinek <jakub@redhat.com>
21674 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
21676 * config/i386/sse.md
21677 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
21679 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
21681 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
21683 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
21685 (*<floatsuffix>floatv2div2sf2): Rename to ...
21686 (*float<floatunssuffix>v2div2sf2): ... this.
21687 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
21688 (float<floatunssuffix>v2div2sf2_mask): ... this.
21689 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
21690 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
21691 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
21693 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
21695 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21697 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21699 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21701 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21703 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
21704 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
21705 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
21706 gen_ufix_truncv8dfv8si2.
21707 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
21708 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
21709 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
21710 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
21711 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
21712 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
21713 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
21714 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
21716 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
21720 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
21722 2018-05-25 Jim Wilson <jimw@sifive.com>
21724 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
21725 * config/riscv/riscv.c (struct machine_function): Add
21726 interrupt_handler_p and attribute_checked_p fields.
21727 (riscv_attribute_table): Add interrupt.
21728 (riscv_interrupt_type_p): New.
21729 (riscv_save_reg_p): Save extra regs for interrupt handler.
21730 (riscv_use_save_libcall): Return false for interrupt handler.
21731 (riscv_first_stack_step): Add forward declaration.
21732 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
21733 for interrupt handler with large frame. Use it for saved reg list.
21734 (riscv_expand_prologue): Move flag_stack_usage_info support to
21735 eliminate duplication.
21736 (riscv_expand_epilogue): Generate mret for interrupt handler.
21737 (riscv_epilogue_uses): New.
21738 (riscv_can_use_return_insn): Return false for interrupt handler.
21739 (riscv_function_ok_for_sibcall): Likewise.
21740 (riscv_set_current_function): Add interrupt handler support.
21741 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
21742 * config/riscv/riscv.md (UNSPECV_MRET): New.
21744 (riscv_frflags, riscv_fsflags): Use tab after opcode.
21746 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
21748 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
21750 PR tree-optimization/85712
21751 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
21752 this candidate has already been replaced in-situ by a copy.
21754 2018-05-25 Jason Merrill <jason@redhat.com>
21756 PR c++/80485 - inline function non-zero address.
21757 * symtab.c (nonzero_address): Check DECL_COMDAT.
21759 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
21762 * config/alpha/alpha.md (ashlsi3): New insn pattern.
21763 (*ashlsi_se): Rename from *ashldi_se. Define as sign
21764 extension of SImode operation. Use const123_operand predicate.
21765 (*saddsi_1): Remove.
21766 (*saddl_se_1): Ditto.
21767 (*ssubsi_1): Ditto.
21768 (*ssubl_se_1): Ditto.
21769 * config/alpha/predicates.md (const123_operand): New predicate.
21770 * config/alpha/constraints.md (P): Use IN_RANGE.
21772 2018-05-25 Richard Biener <rguenther@suse.de>
21774 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
21776 (ref_maybe_used_by_stmt_p): Likewise.
21777 (stmt_may_clobber_ref_p): Likewise.
21778 (stmt_may_clobber_ref_p_1): Likewise.
21779 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
21781 (ref_maybe_used_by_stmt_p): Likewise.
21782 (stmt_may_clobber_ref_p): Likewise.
21783 (stmt_may_clobber_ref_p_1): Likewise.
21784 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
21785 the alias oracle to disambiguate DRs with stmts DR analysis
21787 (vect_analyze_data_refs): Do not give up on not analyzable
21788 DRs for BB vectorization. Remove code truncating the dataref
21791 2018-05-25 Jakub Jelinek <jakub@redhat.com>
21794 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
21795 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
21796 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
21798 2018-05-25 Richard Biener <rguenther@suse.de>
21800 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
21801 function, combining stmt data ref gathering and fatal analysis
21803 (vect_analyze_data_refs): Remove now redudnant code and simplify.
21804 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
21805 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
21806 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
21807 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
21809 2018-05-25 Bin Cheng <bin.cheng@arm.com>
21811 PR tree-optimization/85720
21812 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
21813 SCC if all partitions are builtins.
21814 (version_loop_by_alias_check): New parameter. Generate cancelable
21815 runtime alias check if all partitions are builtins.
21816 (distribute_loop): Update call to above function.
21818 2018-05-25 Bin Cheng <bin.cheng@arm.com>
21820 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
21821 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
21822 (parm_default_def_partition_arg): Ditto.
21823 (set_parm_default_def_partition): Ditto.
21824 (get_parm_default_def_partitions): Ditto and make it static.
21825 (get_undefined_value_partitions): Ditto and make it static.
21826 (remove_ssa_form): Refactor call to init_var_map here.
21827 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
21828 computation for loop region.
21829 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
21830 (register_default_def): Delete.
21831 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
21832 (parm_default_def_partition_arg): Ditto.
21833 (set_parm_default_def_partition): Ditto.
21834 (get_parm_default_def_partitions): Ditto and make it static.
21835 (get_undefined_value_partitions): Ditto and make it static.
21836 (coalesce_with_default, coalesce_with_default): Update comment.
21837 (create_coalesce_list_for_region): New func factored out from
21838 create_outofssa_var_map.
21839 (populate_coalesce_list_for_outofssa): New func factored out from
21840 create_outofssa_var_map and coalesce_ssa_name.
21841 (create_outofssa_var_map): Delete.
21842 (coalesce_ssa_name): Refactor to support live range computation.
21843 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
21844 (get_parm_default_def_partitions): Delete.
21845 (get_undefined_value_partitions): Ditto.
21846 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
21847 computation for loop region.
21848 (new_tree_live_info, loe_visit_block): Ditto.
21849 (live_worklist, set_var_live_on_entry): Ditto.
21850 (calculate_live_on_exit, verify_live_on_entry): Ditto.
21851 * tree-ssa-live.h (struct _var_map): New fields.
21852 (init_var_map): Change decl.
21853 (region_contains_p): New.
21855 2018-05-25 Bin Cheng <bin.cheng@arm.com>
21857 * tree-ssa-live.h (live_merge_and_clear): Delete.
21859 2018-05-25 Richard Biener <rguenther@suse.de>
21862 * tree-dump.c (dequeue_and_dump): Remove access to removed
21863 operand 2 of a SWITCH_EXPR.
21865 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21867 * doc/sourcebuild.texi (vect_double_cond_arith): Include
21868 multiplication and division.
21869 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
21870 (cond_udiv@var{m}, cond_umod@var{m}): Document.
21871 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
21872 (cond_udiv_optab, cond_umod_optab): New optabs.
21873 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
21874 (IFN_COND_RDIV): New internal functions.
21875 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
21876 TRUNC_MOD_EXPR and RDIV_EXPR.
21877 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
21878 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
21880 (SVE_INT_BINARY): Include mult.
21881 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
21882 (optab, sve_int_op): Handle mult.
21883 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
21885 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
21886 for SVE_INT_BINARY_SD.
21888 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21890 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
21891 (optab, sve_int_op): Handle div and udiv.
21892 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
21893 for SVE_INT_BINARY_SD.
21894 (*<optab><mode>3): New insn for the same.
21896 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21898 * tree-vect-patterns.c: Include predict.h.
21899 (vect_recog_divmod_pattern): Restrict check for division support
21900 to when optimizing for size.
21902 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21904 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
21905 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
21906 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
21907 (gimple_match_op::set_op): Likewise.
21908 (gimple_resimplify4): Declare.
21909 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
21910 (expr::gen_transform): Likewise.
21911 (decision_tree::gen): Generate a simplification routine for 4 operands.
21912 * gimple-match-head.c (gimple_simplify): Add an overload for
21913 4 operands. In the top-level function, handle up to 4 call
21914 arguments and call gimple_resimplify4.
21915 (gimple_resimplify4): New function.
21916 (build_call_internal): Pass a fourth operand.
21917 (maybe_push_to_seq): Likewise.
21918 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
21919 Fold VEC_COND_EXPRs of an operation and a default value into
21920 an IFN_COND_* function if possible.
21921 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
21923 (SVE_COND_FP_BINARY): Include them.
21924 (optab, sve_fp_op): Handle them.
21925 (SVE_INT_BINARY_REV): New code iterator.
21926 (SVE_COND_FP_BINARY_REV): New int iterator.
21927 (commutative): New int attribute.
21928 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
21930 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
21932 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
21933 (*cond_<optab><mode>): New patterns for reversed operands.
21935 2018-05-25 Richard Biener <rguenther@suse.de>
21937 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
21938 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
21939 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
21940 (STMT_VINFO_GROUPED_ACCESS): Adjust.
21941 * tree-vect-data-refs.c (everywhere): Adjust users.
21942 * tree-vect-loop.c (everywhere): Likewise.
21943 * tree-vect-slp.c (everywhere): Likewise.
21944 * tree-vect-stmts.c (everywhere): Likewise.
21945 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
21947 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21949 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
21951 (gcc_cv_as_section_exclude): ... this.
21952 Try Solaris as #exclude syntax.
21953 * configure: Regenerate.
21954 * config.in: Regenerate.
21955 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
21957 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
21958 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
21960 * varasm.c (default_elf_asm_named_section): Don't check if
21961 HAVE_GAS_SECTION_EXCLUDE is defined.
21963 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21965 * doc/md.texi: Update the documentation of the cond_* optabs
21966 to mention the new final operand. Fix GET_MODE_NUNITS call.
21967 Describe the scalar case too.
21968 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
21969 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
21971 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
21972 (get_conditional_internal_fn): Update comment.
21973 * tree-vect-loop.c (vectorizable_reduction): Pass the original
21974 accumulator value as a final argument to conditional functions.
21975 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
21976 a define_expand and add an "else" operand. Assert for now that
21977 the else operand is equal to operand 2. Use SVE_INT_BINARY and
21978 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
21979 (*cond_<optab><mode>): New patterns.
21980 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
21981 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21982 (UNSPEC_COND_EOR): Delete.
21983 (optab): Remove associated mappings.
21984 (SVE_INT_BINARY): New code iterator.
21985 (sve_int_op): Remove int attribute and add "minus" to the code
21987 (SVE_COND_INT_OP): Delete.
21988 (SVE_COND_FP_OP): Rename to...
21989 (SVE_COND_FP_BINARY): ...this.
21991 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21993 * optabs.c (can_reuse_operands_p): New function.
21994 (maybe_legitimize_operands): Try to reuse the results for
21997 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
21999 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
22000 Add {q} suffix to insn mnemonic.
22002 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
22004 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
22005 (msp430_warn_func_return): New.
22007 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
22009 * fold-const.c (tree_nonzero_bits): New function.
22010 * fold-const.h (tree_nonzero_bits): Likewise.
22011 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
22012 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
22014 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
22018 * varasm.c (assemble_alias): Check ifunc_resolver only on
22021 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
22024 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
22025 when memory input operand is handled.
22027 2018-05-24 Luis Machado <luis.machado@linaro.org>
22029 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
22031 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
22033 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
22035 * match.pd: Delay FMA folds until after vectorization.
22037 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
22040 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
22041 address check not strict.
22043 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
22045 * gimple-match.h (gimple_match_op): New class.
22046 (mprts_hook): Replace parameters with a gimple_match_op *.
22047 (maybe_build_generic_op): Likewise.
22048 (gimple_simplified_result_is_gimple_val): Replace parameters with
22049 a const gimple_match_op *.
22050 (gimple_simplify): Replace code_helper * and tree * parameters with
22051 a gimple_match_op * parameter.
22052 (gimple_resimplify1): Replace code_helper *, tree and tree *
22053 parameters with a gimple_match_op * parameter.
22054 (gimple_resimplify2): Likewise.
22055 (gimple_resimplify3): Likewise.
22056 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
22057 parameters with a gimple_match_op * parameter.
22058 * gimple-match-head.c (gimple_simplify): Change prototypes of
22059 auto-generated functions to take a gimple_match_op * instead of
22060 separate code_helper * and tree * parameters. Make the same
22061 change in the top-level overload and update calls to the
22062 gimple_resimplify routines. Update calls to the auto-generated
22063 functions and to maybe_push_res_to_seq in the publicly-facing
22064 operation-specific gimple_simplify overloads.
22065 (gimple_match_op::MAX_NUM_OPS): Define.
22066 (gimple_resimplify1): Replace rcode and ops with a single res_op
22067 parameter. Update call to gimple_simplify.
22068 (gimple_resimplify2): Likewise.
22069 (gimple_resimplify3): Likewise.
22070 (mprts_hook): Replace parameters with a gimple_match_op *.
22071 (maybe_build_generic_op): Likewise.
22072 (build_call_internal): Replace type, nargs and ops with
22073 a gimple_match_op *.
22074 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
22075 with a single gimple_match_op *. Update calls to mprts_hook,
22076 build_call_internal and gimple_simplified_result_is_gimple_val.
22077 Factor out code that is common to the tree_code and combined_fn cases.
22078 * genmatch.c (expr::gen_transform): Replace tem_code and
22079 tem_ops with a gimple_match_op called tem_op. Update calls
22080 to the gimple_resimplify functions and maybe_push_res_to_seq.
22081 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
22082 res_ops. Update call to the gimple_resimplify functions.
22083 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
22084 (decision_tree::gen): Make the functions take a gimple_match_op *
22085 called res_op instead of separate res_code and res_ops parameters.
22086 Update call accordingly.
22087 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
22088 and ops with a single res_op parameter. Update calls to
22089 maybe_build_generic_op and maybe_push_res_to_seq.
22090 (fold_stmt_1): Update calls to gimple_simplify and
22091 replace_stmt_with_simplification.
22092 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
22093 and gimple_simplified_result_is_gimple_val.
22094 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
22096 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
22097 with a gimple_match_op *.
22098 (vn_nary_build_or_lookup): Likewise. Update call to
22099 vn_nary_build_or_lookup_1.
22100 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
22101 gimple_match_op *. Update calls to the gimple_resimplify routines
22102 and to gimple_simplified_result_is_gimple_val.
22103 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
22104 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
22105 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
22106 (visit_nary_op): Likewise.
22107 (visit_reference_op_load): Likewise.
22109 2018-05-23 Luis Machado <luis.machado@linaro.org>
22111 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
22112 modifier for printing the step amount.
22114 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
22117 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
22120 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
22122 * doc/sourcebuild.texi (Endianness): New subsubsection.
22124 2018-05-23 Luis Machado <luis.machado@linaro.org>
22126 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22127 <prefetch_dynamic_strides>: New const bool field.
22128 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22129 prefetch_dynamic_strides.
22130 (exynosm1_prefetch_tune): Likewise.
22131 (thunderxt88_prefetch_tune): Likewise.
22132 (thunderx_prefetch_tune): Likewise.
22133 (thunderx2t99_prefetch_tune): Likewise.
22134 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
22136 (aarch64_override_options_internal): Update to set
22137 PARAM_PREFETCH_DYNAMIC_STRIDES.
22138 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
22139 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
22140 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
22141 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22142 prefetch-dynamic-strides setting.
22144 2018-05-23 Luis Machado <luis.machado@linaro.org>
22146 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22147 <minimum_stride>: New const int field.
22148 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22149 minimum_stride field defaulting to -1.
22150 (exynosm1_prefetch_tune): Likewise.
22151 (thunderxt88_prefetch_tune): Likewise.
22152 (thunderx_prefetch_tune): Likewise.
22153 (thunderx2t99_prefetch_tune): Likewise.
22154 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
22155 <default_opt_level>: Set to 3.
22156 (aarch64_override_options_internal): Update to set
22157 PARAM_PREFETCH_MINIMUM_STRIDE.
22158 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
22159 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
22160 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
22161 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
22162 stride is constant and is below the minimum stride threshold.
22164 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22166 * config/arm/arm-cpus.in (mode26): Delete.
22167 (armv4): Delete mode26 reference.
22168 * config/arm/arm.c (arm_configure_build_target): Delete use of
22171 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
22173 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
22175 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
22176 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
22178 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
22179 (floatunsdidf2): Ditto.
22181 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
22183 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
22184 (fixuns_trunc<mode>si2_avx512f): Ditto.
22185 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
22186 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
22187 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
22189 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
22191 PR rtl-optimization/79985
22192 * df-scan.c (df_insn_refs_collect): Remove special case for
22193 global registers and asm statements.
22195 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
22197 * extend.texi (Global Register Variables): Rewrite the bullet list.
22198 Note that the register is available for allocation. Note that access
22199 via inline asm must use constraints. Add note about async-signal
22200 handlers. Remove paragraph about automagic register selection.
22202 2018-05-23 Richard Biener <rguenther@suse.de>
22204 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
22205 of fixed offset from memset VN.
22207 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
22209 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
22210 first_interp field.
22211 (alloc_cand_and_find_basis): Initialize first_interp field.
22212 (slsr_process_mul): Modify first_interp field.
22213 (slsr_process_add): Likewise.
22214 (slsr_process_cast): Modify first_interp field for each new
22216 (slsr_process_copy): Likewise.
22217 (dump_candidate): Dump first_interp field.
22218 (replace_mult_candidate): Process all interpretations, not just
22220 (replace_rhs_if_not_dup): Likewise.
22221 (replace_one_candidate): Likewise.
22223 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
22225 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
22227 (aarch64_needs_frame_chain): New function.
22228 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
22230 2018-05-23 Sudakshina Das <sudi.das@arm.com>
22233 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
22234 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
22235 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
22236 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
22237 as true for strict-align.
22238 (aarch64_can_inline_p): Perform checks even when callee has no
22239 attributes to check for strict alignment.
22240 * doc/extend.texi (AArch64 Function Attributes): Document
22242 * doc/invoke.texi: (AArch64 Options): Likewise.
22244 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
22246 PR tree-optimization/85853
22247 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
22248 the handling of the root of the node to...
22249 (vect_slp_analyze_node_operations_1): ...this new function,
22250 and run the whole thing with the child nodes' def types
22251 set according to their SLP node's def type.
22253 2018-05-23 Richard Biener <rguenther@suse.de>
22255 PR middle-end/85874
22256 * tree-data-ref.c (create_runtime_alias_checks): Defer
22257 and ignore overflow warnings.
22259 2018-05-23 Yury Gribov <tetra2005@gmail.com>
22261 PR tree-optimization/85822
22262 * tree-vrp.c (is_masked_range_test): Fix handling of negative
22265 2018-05-23 Richard Biener <rguenther@suse.de>
22267 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
22268 memset constants via native_interpret_expr.
22270 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
22273 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
22275 (cgraph_node::create_alias): Likewise.
22276 (cgraph_node::get_availability): Check ifunc_resolver instead
22277 of looking up ifunc attribute.
22278 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
22279 * varasm.c (do_assemble_alias): Likewise.
22280 (assemble_alias): Likewise.
22281 (default_binds_local_p_3): Likewise.
22282 * cgraph.h (cgraph_node): Add ifunc_resolver.
22283 (cgraph_node::only_called_directly_or_aliased_p): Return false
22284 for IFUNC resolver.
22285 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
22287 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
22288 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
22289 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
22290 instead of looking up ifunc attribute.
22292 2018-05-22 Luis Machado <luis.machado@linaro.org>
22294 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
22296 2018-05-22 Martin Sebor <msebor@redhat.com>
22298 PR middle-end/85359
22299 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
22300 only when expasion succeeds.
22301 (expand_builtin_strcmp): Same.
22302 (expand_builtin_strncmp): Same.
22304 2018-05-22 Martin Sebor <msebor@redhat.com>
22306 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
22308 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
22309 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22311 * config/aarch64/aarch64-ldpstp.md: Replace uses of
22312 aarch64_mem_pair_operand with memory_operand and delete operand swapping
22314 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
22315 Add check for legitimate_address.
22316 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
22317 (aarch64_swap_ldrstr_operands): New.
22318 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
22321 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
22322 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22324 * config/aarch64/aarch64.md: New patterns to generate stp
22326 (store_pair_sw, store_pair_dw): New patterns to generate stp for
22327 single words and double words.
22328 (load_pair_sw, load_pair_dw): Likewise.
22329 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
22331 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
22333 * config/aarch64/aarch64-ldpstp.md: Modify peephole
22334 for different mode ldpstp and add peephole for merged zero stores.
22335 Likewise for loads.
22336 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
22338 (aarch64_gen_store_pair): Rename calls to match new patterns.
22339 (aarch64_gen_load_pair): Rename calls to match new patterns.
22340 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
22341 (load_pair<DREG:mode><DREG2:mode>): ... This.
22342 (store_pair<mode>): Rename to...
22343 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
22344 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
22345 New mode iterators.
22346 (V_INT_EQUIV): Handle SImode.
22347 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
22350 2018-05-22 Martin Sebor <msebor@redhat.com>
22353 * calls.c (maybe_warn_nonstring_arg): Use string length to set
22354 or ajust the presumed bound on an operation to avoid unnecessary
22357 2018-05-22 Martin Sebor <msebor@redhat.com>
22359 PR tree-optimization/85826
22360 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
22361 assuming that a DECL necesarily has a constant size.
22363 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
22365 PR middle-end/85862
22366 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
22368 2018-05-22 Richard Biener <rguenther@suse.de>
22370 PR tree-optimization/85834
22371 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
22372 non-constant and non-zero memset arguments.
22374 2018-05-22 Martin Liska <mliska@suse.cz>
22377 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
22379 2018-05-22 Richard Biener <rguenther@suse.de>
22381 PR tree-optimization/85863
22382 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
22383 comparisons when vectype is specified.
22384 (vectorizable_condition): Do not specify vectype for
22385 vect_is_simple_cond when SLP vectorizing.
22387 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
22390 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
22391 define __ibm128 as long double.
22392 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
22393 as a distinct type when IEEE 128-bit support is enabled.
22394 (init_float128_ieee): Fix up conversions between IFmode and IEEE
22395 128-bit types to use the correct functions.
22396 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
22397 convert between 128-bit floating point types that have different
22398 modes but the same representation, instead of using gen_lowpart to
22400 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
22402 (IFKF_reg): New attributes to give the register constraints for
22404 (extend<mode>tf2_internal): New insns to mark an explicit
22405 conversion between 128-bit floating point types that have a
22406 different mode but share the same representation.
22408 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
22410 PR tree-optimization/85814
22411 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
22412 a null return from get_strinfo when unsharing the next
22413 strinfo in the chain.
22415 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
22418 * varasm.c (weak_finish): Clean up weak_decls.
22420 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22422 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
22423 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
22424 UNSPEC_UADALP values.
22425 * config/aarch64/iterators.md (ABAL): New int iterator.
22428 (sur): Add mappings for the above.
22429 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22431 (aarch64_<sur>abal<mode>_4): Likewise.
22432 (aarch64_<sur>adalp<mode>_3): Likewise.
22433 (<sur>sadv16qi): New define_expand.
22435 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
22437 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
22438 (*movdf_internal): Ditto.
22439 (*rcpsf2_sse): Ditto.
22440 (*rsqrtsf2_sse): Ditto.
22441 (*sqrt<mode>2_sse): Ditto.
22443 2018-05-21 Tamar Christina <tamar.christina@arm.com>
22445 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
22447 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
22448 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
22449 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
22450 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
22452 * config/aarch64/arm_neon.h: Likewise.
22453 * config/aarch64/iterators.md (VQ_I): New.
22455 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
22457 * config.gcc: Add arc/t-multilib-linux to tmake_file for
22459 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
22462 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
22464 * config/nds32/constraints.md (S): New constraint.
22465 * config/nds32/nds32.md (call_internal): Use constraint S.
22466 (call_value_internal): Likewise.
22467 (sibcall_internal): Likewise.
22468 (sibcall_value_internal): Likewise.
22470 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
22471 Chung-Ju Wu <jasonwucj@gmail.com>
22473 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
22474 into consideration.
22476 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
22477 Chung-Ju Wu <jasonwucj@gmail.com>
22479 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
22480 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
22481 (nds32_rtx_costs_impl): Simplify.
22482 (nds32_address_cost_impl): Simplify.
22483 (nds32_init_rtx_costs): New function.
22484 (nds32_rtx_costs_speed_prefer): Likewise.
22485 (nds32_rtx_costs_size_prefer): Likewise.
22486 (nds32_address_cost_speed_prefer): Likewise.
22487 (nds32_address_cost_speed_fwprop): Likewise.
22488 (nds32_address_cost_size_prefer): Likewise.
22489 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
22490 * config/nds32/nds32.c (nds32_option_override): Use
22491 nds32_init_rtx_costs function.
22493 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
22495 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
22496 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
22497 (TARGET_PIPELINE_N8): Likewise.
22498 (TARGET_PIPELINE_N10): Likewise.
22499 (TARGET_PIPELINE_N13): Likewise.
22500 (TARGET_PIPELINE_GRAYWOLF): Likewise.
22502 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
22504 * config/nds32/nds32-fpu.md: Update copyright year.
22506 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22508 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
22510 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22512 * config/nds32/nds32.c
22513 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
22514 * config/nds32/nds32.opt (minline-asm-r15): New option.
22516 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22518 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
22520 * config/nds32/nds32.md (abssi2): New pattern.
22522 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
22524 * config/i386/i386.md (rex64namesuffix): New mode attribute.
22525 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
22526 Merge insn pattern from sse_cvtsi2ss<round_name> and
22527 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
22528 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
22529 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
22530 using SWI48 mode iterator.
22531 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
22532 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
22533 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
22534 pattern from sse_cvttss2si<round_saeonly_name>
22535 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
22536 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
22537 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
22538 using SWI48 mode iterator.
22539 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
22540 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
22541 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
22542 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
22543 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
22544 using SWI48 mode iterator.
22545 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
22546 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
22547 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
22548 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
22549 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
22550 SWI48 mode iterator.
22551 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
22552 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
22553 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
22554 pattern from sse_cvttsd2si<round_saeonly_name>
22555 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
22557 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22559 * config/nds32/nds32-md-auxiliary.c
22560 (nds32_valid_smw_lwm_base_p): Refine.
22561 (nds32_output_smw_single_word): Refine.
22562 (nds32_output_smw_double_word): New.
22563 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
22565 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22567 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
22568 (nds32_output_stack_pop): Refine.
22569 (nds32_expand_unaligned_load): Refine.
22570 (nds32_expand_unaligned_store): Refine.
22572 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22573 Chung-Ju Wu <jasonwucj@gmail.com>
22575 * config/nds32/constants.md: Add TP_REGNUM constant.
22576 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
22577 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
22579 * config/nds32/nds32-doubleword.md: Consider flag_pic.
22580 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
22581 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
22582 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
22583 and PIC code generation.
22584 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
22586 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
22588 * config/nds32/nds32.md: Support TLS and PIC.
22589 * config/nds32/nds32.c: Support TLS and PIC.
22590 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
22591 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
22594 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22596 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
22597 mode with E_ prefix.
22599 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22600 Chung-Ju Wu <jasonwucj@gmail.com>
22602 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
22603 * config/nds32/nds32-md-auxiliary.c
22604 (symbolic_reference_mentioned_p): New.
22605 (nds32_legitimize_ict_address): New.
22606 (nds32_expand_ict_move): New.
22607 (nds32_indirect_call_referenced_p): New.
22608 (nds32_symbol_binds_local_p): Delete.
22609 (nds32_long_call_p): Modify.
22610 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
22611 * config/nds32/nds32-protos.h
22612 (symbolic_reference_mentioned_p): Declare.
22613 (nds32_legitimize_ict_address): Declare.
22614 (nds32_expand_ict_move): Declare.
22615 (nds32_indirect_call_referenced_p): Declare.
22616 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
22617 (nds32_relax_group): Use nds32_ict_const_p as condition.
22618 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
22619 (nds32_asm_file_start): Output ict_model directive in asm code.
22620 (nds32_legitimate_address_p): Consider indirect call.
22621 (nds32_print_operand): Consider indirect call.
22622 (nds32_print_operand_address): Consider indirect call.
22623 (nds32_insert_attributes): Handle "indirect_call" attribute.
22624 (TARGET_LEGITIMATE_ADDRESS_P): Define.
22625 (TARGET_LEGITIMATE_CONSTANT_P): Define.
22626 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
22627 (TARGET_DELEGITIMIZE_ADDRESS): Define.
22628 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
22629 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
22630 (TARGET_ICT_MODEL_SMALL): Define.
22631 (TARGET_ICT_MODEL_LARGE): Define.
22632 * config/nds32/nds32.md (movsi): Consider ict model.
22633 (call, call_value): Consider ict model.
22634 (sibcall, sibcall_value): Consider ict model.
22635 * config/nds32/nds32.opt (mict-model): New option.
22636 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
22639 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
22640 Monk Chiang <sh.chiang04@gmail.com>
22641 Jim Wilson <jimw@sifive.com>
22643 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
22644 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
22645 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
22646 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
22647 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
22648 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
22649 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
22650 compute save_libcall_adjustment properly.
22651 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
22652 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
22653 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
22654 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
22655 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
22656 (ABI_SPEC): Handle mabi=ilp32e.
22657 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
22658 (RVE): Add RVE mask.
22659 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
22660 <-march>: Add rv32e as an example.
22662 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
22665 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
22666 (intra_create_variable_infos): Handle C++ constructors.
22668 2018-05-18 Martin Liska <mliska@suse.cz>
22670 * passes.def: Remove a redundant pass.
22672 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
22675 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
22677 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22679 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
22681 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
22683 (armv2, armv2a, armv3, armv3m): Delete architectures.
22684 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
22685 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
22686 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
22688 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
22689 (*mulsidi3adddi): Likewise.
22690 (mulsidi3): Likewise.
22691 (*mulsidi3_nov6): Likewise.
22692 (umulsidi3): Likewise.
22693 (umulsidi3_nov6): Likewise.
22694 (umaddsidi4): Likewise.
22695 (*umulsidi3adddi): Likewise.
22696 (smulsi3_highpart): Likewise.
22697 (*smulsi3_highpart_nov6): Likewise.
22698 (umulsi3_highpart): Likewise.
22699 (*umulsi3_highpart_nov6): Likewise.
22700 * config/arm/arm.h (arm_arch3m): Delete.
22701 * config/arm/arm.c (arm_arch3m): Delete.
22702 (arm_option_override_internal): Update armv3-related comment.
22703 (arm_configure_build_target): Delete use of isa_bit_mode32.
22704 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
22705 (arm_rtx_costs_internal): Delete check of arm_arch3m.
22706 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
22707 (mulsa3): Likewise.
22708 (mulusa3): Likewise.
22709 * config/arm/arm-protos.h (arm_arch3m): Delete.
22710 * config/arm/arm-tables.opt: Regenerate.
22711 * config/arm/arm-tune.md: Likewise.
22712 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
22713 deleted architectures.
22715 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22717 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
22718 (armv5t, armv5te): New features.
22719 (ARMv5, ARMv5e): Delete fgroups.
22720 (ARMv5t, ARMv5te): Adjust for above changes.
22721 (ARMv6m): Likewise.
22722 (armv5, armv5e): Delete arches.
22723 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
22725 (*call_reg_arm): Likewise.
22726 (*call_value_reg_armv5): Likewise.
22727 (*call_value_reg_arm): Likewise.
22728 (*call_symbol): Likewise.
22729 (*call_value_symbol): Likewise.
22730 (*sibcall_insn): Likewise.
22731 (*sibcall_value_insn): Likewise.
22732 (clzsi2): Likewise.
22733 (prefetch): Likewise.
22734 (define_split and define_peephole2 dependent on arm_arch5):
22736 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
22738 (TARGET_ARM_QBIT): Likewise.
22739 (TARGET_DSP_MULTIPLY): Likewise.
22740 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
22741 (arm_arch5, arm_arch5e): Delete.
22742 (arm_arch5t, arm_arch5te): Declare.
22743 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
22744 (arm_arch5t): Declare.
22745 (arm_option_reconfigure_globals): Update for the above.
22746 (arm_options_perform_arch_sanity_checks): Update comment, replace
22747 use of arm_arch5 with arm_arch5t.
22748 (use_return_insn): Likewise.
22749 (arm_emit_call_insn): Likewise.
22750 (output_return_instruction): Likewise.
22751 (arm_final_prescan_insn): Likewise.
22752 (arm_coproc_builtin_available): Likewise.
22753 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
22754 arm_arch5e with arm_arch5t and arm_arch5te.
22755 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
22756 (arm_arch5t, arm_arch5te): Declare.
22757 * config/arm/arm-tables.opt: Regenerate.
22758 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
22759 * config/arm/t-multilib: Likewise.
22760 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
22761 instead of arm_arch5.
22762 (*call_reg_thumb1): Likewise.
22763 (*call_value_reg_thumb1_v5): Likewise.
22764 (*call_value_reg_thumb1): Likewise.
22765 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
22767 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
22769 2018-05-18 Martin Liska <mliska@suse.cz>
22771 PR gcov-profile/84846
22772 * doc/gcov.texi: Document -t option of gcov tool.
22774 2018-05-18 Martin Liska <mliska@suse.cz>
22776 PR gcov-profile/84846
22777 * gcov.c (print_usage): Add new -t option.
22778 (process_args): Handle the option.
22779 (generate_results): Use stdout as output when requested by
22782 2018-05-18 Martin Liska <mliska@suse.cz>
22784 PR gcov-profile/84846
22785 * coverage.c (coverage_init): Write PWD to .gcno file.
22786 * doc/gcov.texi: Document how working directory is printed.
22787 * gcov-dump.c (dump_gcov_file): Print PWD.
22788 * gcov.c (output_intermediate_file): Likewise.
22789 (read_graph_file): Read PWD string.
22790 (output_lines): Print PWD.
22792 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22794 PR middle-end/85817
22795 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
22796 for retval and return false if all args to phi are zero.
22798 2018-05-18 Richard Biener <rguenther@suse.de>
22800 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
22802 (evrp_dom_walker::before_dom_children): Call it.
22804 2018-05-18 Richard Biener <rguenther@suse.de>
22806 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
22807 results when processing array refs with variable index.
22809 2018-05-18 Toon Moene <toon@moene.org>
22811 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
22812 directly after that of -floop-interchange. Indicate that both
22813 options are enabled by default when specifying -O3.
22815 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22817 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
22818 iterator. Delete separate integer-mode vec_set<mode> expander.
22819 (aarch64_simd_vec_setv2di): Delete.
22820 (vec_setv2di): Delete.
22821 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
22822 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
22823 the "w, r" alternative.
22825 2018-05-18 Martin Liska <mliska@suse.cz>
22827 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
22828 * tree-pass.h (make_pass_lower_switch_O0): New function.
22829 * tree-switch-conversion.c (node_has_low_bound): Remove.
22830 (node_has_high_bound): Likewise.
22831 (node_is_bounded): Likewise.
22832 (class pass_lower_switch): Make it a template type and create
22834 (pass_lower_switch::execute): Add template argument.
22835 (make_pass_lower_switch): New function.
22836 (make_pass_lower_switch_O0): New function.
22837 (do_jump_if_equal): Remove.
22838 (emit_case_nodes): Simplify to just handle all 3 cases and leave
22839 all the hard work to tree optimization passes.
22841 2018-05-18 Martin Liska <mliska@suse.cz>
22843 * dbgcnt.c (limit_low): Renamed from limit.
22844 (limit_high): New variable.
22845 (dbg_cnt_is_enabled): Check for upper limit.
22846 (dbg_cnt): Adjust dumping.
22847 (dbg_cnt_set_limit_by_index): Add new argument for high
22849 (dbg_cnt_set_limit_by_name): Likewise.
22850 (dbg_cnt_process_single_pair): Parse new format.
22851 (dbg_cnt_process_opt): Use strtok.
22852 (dbg_cnt_list_all_counters): Remove 'value' and add
22854 * doc/invoke.texi: Document changes.
22856 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
22858 * doc/sourcebuild.texi (scalar_all_fma): Document.
22859 * tree.def (FMA_EXPR): Delete.
22860 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
22861 * internal-fn.c (ternary_direct): New macro.
22862 (expand_ternary_optab_fn): Likewise.
22863 (direct_ternary_optab_supported_p): Likewise.
22864 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
22865 * builtins.c (fold_builtin_fma): Delete.
22866 (fold_builtin_3): Don't call it.
22867 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
22868 * expr.c (expand_expr_real_2): Likewise.
22869 * fold-const.c (operand_equal_p): Likewise.
22870 (fold_ternary_loc): Likewise.
22871 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
22872 * gimple.c (DEFTREECODE): Likewise.
22873 * gimplify.c (gimplify_expr): Likewise.
22874 * optabs-tree.c (optab_for_tree_code): Likewise.
22875 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
22876 * tree-eh.c (operation_could_trap_p): Likewise.
22877 (stmt_could_throw_1_p): Likewise.
22878 * tree-inline.c (estimate_operator_cost): Likewise.
22879 * tree-pretty-print.c (dump_generic_node): Likewise.
22880 (op_code_prio): Likewise.
22881 * tree-ssa-loop-im.c (stmt_cost): Likewise.
22882 * tree-ssa-operands.c (get_expr_operands): Likewise.
22883 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
22884 * fold-const-call.h (fold_fma): Delete.
22885 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
22886 CFN_FNMA and CFN_FNMS.
22887 (fold_fma): Delete.
22888 * genmatch.c (combined_fn): New enum.
22889 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
22890 (commutative_op): New function.
22891 (commutate): Use it. Handle more than 2 operands.
22892 (dt_operand::gen_gimple_expr): Use commutative_op.
22893 (parser::parse_expr): Allow :c to be used with non-binary
22894 operators if the commutative operand is known.
22895 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
22896 CFN_FMS, CFN_FNMA and CFN_FNMS.
22897 (backprop::process_assign_use): Remove FMA_EXPR handling.
22898 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
22899 (gen_hsa_fma): New function.
22900 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
22901 IFN_FNMA and IFN_FNMS.
22902 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
22903 * gimple-fold.h (follow_all_ssa_edges): Declare.
22904 * gimple-fold.c (follow_all_ssa_edges): New function.
22905 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
22906 gimple_build interface and use follow_all_ssa_edges to fold the result.
22907 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
22908 instead of checking for optabs directly.
22909 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
22910 rather than FMA_EXPRs.
22911 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
22912 call to IFN_FMA instead of an FMA_EXPR.
22914 2018-05-17 Jim Wilson <jimw@sifive.com>
22916 * expr.c (do_tablejump): When converting index to Pmode, if we have a
22917 sign extended promoted subreg, and the range does not have the sign bit
22918 set, then do a sign extend.
22920 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
22921 test, check for sign extended subreg and/or constant operands, and
22922 do a sign extend in that case.
22924 2018-05-17 Steve Ellcey <sellcey@cavium.com>
22926 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
22927 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
22929 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
22930 Change logics_shift_reg to logics_shift_imm.
22931 (thunderx2t99_fp_loadpair_basic): Delete.
22932 (thunderx2t99_fp_storepair_basic): Delete.
22933 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
22934 (thunderx2t99_asimd_polynomial): Delete.
22935 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
22936 and neon_fp_mul_d_scalar_q.
22937 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
22938 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
22939 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
22940 (thunderx2t99_asimd_lut): Add missing tbl types.
22941 (thunderx2t99_asimd_ext): Delete.
22942 (thunderx2t99_asimd_load1_1_mult): Delete.
22943 (thunderx2t99_asimd_load1_2_mult): Delete.
22944 (thunderx2t99_asimd_load1_ldp): New.
22945 (thunderx2t99_asimd_load1): New.
22946 (thunderx2t99_asimd_load2): Add missing *load2* types.
22947 (thunderx2t99_asimd_load3): New.
22948 (thunderx2t99_asimd_load4): New.
22949 (thunderx2t99_asimd_store1_1_mult): Delete.
22950 (thunderx2t99_asimd_store1_2_mult): Delete.
22951 (thunderx2t99_asimd_store2_mult): Delete.
22952 (thunderx2t99_asimd_store2_onelane): Delete.
22953 (thunderx2t99_asimd_store_stp): New.
22954 (thunderx2t99_asimd_store1): New.
22955 (thunderx2t99_asimd_store2): New.
22956 (thunderx2t99_asimd_store3): New.
22957 (thunderx2t99_asimd_store4): New.
22959 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
22961 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
22962 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
22964 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
22965 Segher Boessenkool <segher@kernel.crashing.org>
22968 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
22971 2018-05-17 Richard Biener <rguenther@suse.de>
22973 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
22974 for pruning loop and prune defs feeding only already visited PHIs.
22976 2018-05-17 Richard Biener <rguenther@suse.de>
22978 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
22980 2018-05-17 Bin Cheng <bin.cheng@arm.com>
22981 Richard Biener <rguenther@suse.de>
22983 PR tree-optimization/85793
22984 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
22985 for VMAT_ELEMENTWISE.
22987 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
22989 * internal-fn.h (lookup_internal_fn): Declare
22990 * internal-fn.c (lookup_internal_fn): New function.
22991 * gimple.c (gimple_build_call_from_tree): Handle calls to
22992 internal functions.
22993 * gimple-pretty-print.c (dump_gimple_call): Print "." before
22994 internal function names.
22995 * tree-pretty-print.c (dump_generic_node): Likewise.
22996 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
22998 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
23000 * gimple-fold.h (gimple_build): Make the function forms take
23001 combined_fn rather than built_in_function.
23002 (gimple_simplify): Likewise.
23003 * gimple-match-head.c (gimple_simplify): Likewise.
23004 * gimple-fold.c (gimple_build): Likewise.
23005 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
23006 rather than gimple_build_call_internal.
23007 (get_initial_defs_for_reduction): Likewise.
23008 (vect_create_epilog_for_reduction): Likewise.
23009 (vectorizable_live_operation): Likewise.
23011 2018-05-17 Martin Liska <mliska@suse.cz>
23013 * gimple-ssa-sprintf.c (format_directive): Do not use
23014 space in between 'G_' and '('.
23016 2018-05-17 Jakub Jelinek <jakub@redhat.com>
23019 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
23020 even if the mask is not all ones.
23023 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
23025 (ix86_gimple_fold_builtin): Likewise.
23028 * config/i386/i386.c: Include tree-vector-builder.h.
23029 (ix86_vector_shift_count): New function.
23030 (ix86_fold_builtin): Fold shift builtins by scalar count.
23031 (ix86_gimple_fold_builtin): Likewise.
23033 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
23034 _mm512_setzero): New intrinsics.
23036 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
23037 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23039 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
23040 code generation for cases where splatting a value is not useful.
23041 * simplify-rtx.c (simplify_ternary_operation): Simplify
23042 vec_merge across a vec_duplicate and a paradoxical subreg forming
23043 a vector mode to a vec_concat.
23045 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
23047 * config.gcc: Support "goldmont-plus".
23048 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
23050 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23051 PROCESSOR_GOLDMONT_PLUS.
23052 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
23053 (processor_target_table): Add "goldmont-plus".
23054 (PTA_GOLDMONT_PLUS): Define.
23055 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
23056 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
23057 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
23058 (fold_builtin_cpu): Add "goldmont-plus".
23059 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
23060 (ix86_option_override_internal): Add "goldmont-plus".
23061 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
23062 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
23063 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
23064 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
23066 2018-05-17 Richard Biener <rguenther@suse.de>
23068 PR tree-optimization/85757
23069 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
23070 remove defs that only feed that PHI from further processing.
23072 2018-05-16 Jim Wilson <jimw@sifive.com>
23074 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
23076 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
23078 2018-05-16 Mark Wielaard <mark@klomp.org>
23080 * dwarf2out.c (count_index_strings): New function.
23081 (output_indirect_strings): Call count_index_strings and generate
23082 header for dwarf_version >= 5.
23084 2018-05-16 Mark Wielaard <mark@klomp.org>
23086 * dwarf2out.c (dwarf_FORM): New function.
23087 (set_indirect_string): Use dwarf_FORM.
23088 (reset_indirect_string): Likewise.
23089 (size_of_die): Likewise.
23090 (value_format): Likewise.
23091 (output_die): Likewise.
23092 (add_skeleton_AT_string): Likewise.
23093 (output_macinfo_op): Likewise.
23094 (index_string): Likewise.
23095 (output_index_string_offset): Likewise.
23096 (output_index_string): Likewise.
23097 (count_index_strings): Likewise.
23099 2018-05-16 Carl Love <cel@us.ibm.com>
23101 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
23102 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
23104 2018-05-16 Martin Jambor <mjambor@suse.cz>
23106 * ipa-prop.c (ipa_free_all_edge_args): Remove.
23107 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
23109 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
23111 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
23112 (fnma<mode>4): Likewise.
23113 (fms<mode>4): Likewise.
23114 (fnms<mode>4): Likewise.
23115 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
23116 (aarch64_fnma<mode>4): Likewise.
23117 (aarch64_fms<mode>4): Likewise.
23118 (aarch64_fnms<mode>4): Likewise.
23119 (aarch64_fnmadd<mode>4): Likewise.
23121 2018-05-16 Jason Merrill <jason@redhat.com>
23123 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
23125 2018-05-16 Richard Biener <rguenther@suse.de>
23127 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
23128 (dump_stmt_cost): Declare.
23129 (add_stmt_cost): Dump cost we add.
23130 (add_stmt_costs): New function.
23131 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
23132 No longer exported.
23133 (vect_analyze_stmt): Adjust prototype.
23134 (vectorizable_condition): Likewise.
23135 (vectorizable_live_operation): Likewise.
23136 (vectorizable_reduction): Likewise.
23137 (vectorizable_induction): Likewise.
23138 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
23139 cost vector to pass to vectorizable_ and record afterwards.
23140 (vect_model_reduction_cost): Take cost vector argument and adjust.
23141 (vect_model_induction_cost): Likewise.
23142 (vectorizable_reduction): Likewise.
23143 (vectorizable_induction): Likewise.
23144 (vectorizable_live_operation): Likewise.
23145 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
23146 SLP_TREE_NUMBER_OF_VEC_STMTS.
23147 (vect_analyze_slp_cost_1): Remove.
23148 (vect_analyze_slp_cost): Likewise.
23149 (vect_slp_analyze_node_operations): Take visited args and
23150 a target cost vector. Avoid processing already visited stmt sets.
23151 (vect_slp_analyze_operations): Use a local cost vector to gather
23152 costs and register those of non-discarded instances.
23153 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
23154 (vect_schedule_slp_instance): Remove copying of
23155 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
23157 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
23158 adding cost. Record cost entry location.
23159 (vect_prologue_cost_for_slp_op): Function to compute cost of
23160 a constant or invariant generated for SLP vect in the prologue,
23161 split out from vect_analyze_slp_cost_1.
23162 (vect_model_simple_cost): Make static. Adjust for SLP costing.
23163 (vect_model_promotion_demotion_cost): Likewise.
23164 (vect_model_store_cost): Likewise, make static.
23165 (vect_model_load_cost): Likewise.
23166 (vectorizable_bswap): Add cost vector arg and adjust.
23167 (vectorizable_call): Likewise.
23168 (vectorizable_simd_clone_call): Likewise.
23169 (vectorizable_conversion): Likewise.
23170 (vectorizable_assignment): Likewise.
23171 (vectorizable_shift): Likewise.
23172 (vectorizable_operation): Likewise.
23173 (vectorizable_store): Likewise.
23174 (vectorizable_load): Likewise.
23175 (vectorizable_condition): Likewise.
23176 (vectorizable_comparison): Likewise.
23177 (can_vectorize_live_stmts): Likewise.
23178 (vect_analyze_stmt): Likewise.
23179 (vect_transform_stmt): Adjust calls to vectorizable_*.
23180 * tree-vectorizer.c: Include gimple-pretty-print.h.
23181 (dump_stmt_cost): New function.
23183 2018-05-16 Richard Biener <rguenther@suse.de>
23185 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
23186 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
23187 * tree-ssa-dse.c: Include tree-ssa-loop.h.
23188 (check_name): New callback.
23189 (dse_classify_store): Track cycles via a visited bitmap of PHI
23190 defs and simplify handling of in-loop and across loop dead stores
23191 and properly fail for loop-variant refs. Handle byte-tracking with
23192 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
23195 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
23197 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
23198 (vect_get_mask_type_for_stmt): Likewise.
23199 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
23201 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
23202 to determine the statement's vector type and the vector type that
23203 should be used for calculating nunits. Deal with cases in which
23204 the type has to be deferred.
23205 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
23206 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
23207 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
23208 (vect_determine_vf_for_stmt): New functions, split out from...
23209 (vect_determine_vectorization_factor): ...here.
23210 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
23211 (vect_get_mask_type_for_stmt): New functions, split out from
23212 vect_determine_vectorization_factor.
23214 2018-05-16 Richard Biener <rguenther@suse.de>
23216 * tree-cfg.c (verify_gimple_assign_ternary): Properly
23217 verify the [VEC_]COND_EXPR embedded comparison.
23219 2018-05-15 Martin Sebor <msebor@redhat.com>
23221 PR tree-optimization/85753
23222 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
23223 RECORD_TYPE in addition to ARRAY_TYPE.
23225 2018-05-15 Martin Sebor <msebor@redhat.com>
23227 PR middle-end/85643
23228 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
23230 2018-05-15 Richard Biener <rguenther@suse.de>
23232 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
23233 add by_clobber_p one. Change algorithm to collect all defs
23234 representing uses we need to walk and try reducing them to
23235 a single one before failing.
23236 (dse_dom_walker::dse_optimize_stmt): Adjust.
23238 2018-05-13 Mark Wielaard <mark@klomp.org>
23240 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
23241 (size_of_loc_descr): Likewise.
23242 (output_loc_operands): Likewise.
23243 (output_loc_operands_raw): Likewise.
23244 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
23245 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
23246 (hash_loc_operands): Likewise.
23247 (compare_loc_operands): Likewise.
23249 2018-05-14 Mark Wielaard <mark@klomp.org>
23251 * dwarf2out.c (count_index_addrs): New function.
23252 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
23254 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23256 PR tree-optimization/83648
23257 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
23258 return value as malloc candidate.
23260 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23263 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
23264 param as true in call to suggest_attribute.
23266 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
23268 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
23269 -mreadonly-in-sdata.
23271 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23273 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
23275 (aarch64_crypto_aesd_fused): Likewise.
23277 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
23279 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
23280 (movsi_aarch64): Likewise.
23281 (load_pairsi): Likewise.
23282 (load_pairdi): Likewise.
23283 (store_pairsi): Likewise.
23284 (store_pairdi): Likewise.
23285 (load_pairsf): Likewise.
23286 (load_pairdf): Likewise.
23287 (store_pairsf): Likewise.
23288 (store_pairdf): Likewise.
23289 (zero_extend): Likewise.
23290 (trunc): Swap alternatives.
23291 (fcvt_target): Add '?' to prefer w over r.
23293 2018-05-14 Jakub Jelinek <jakub@redhat.com>
23296 * config/i386/i386.md: Disallow non-commutative arithmetics in
23297 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
23298 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
23299 in the peephole2 before it.
23301 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
23303 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
23304 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
23305 (ix86_handle_option): Handle -mcldemote.
23306 * config.gcc: New header.
23307 * config/i386/cldemoteintrin.h: New file.
23308 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
23309 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
23311 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23312 OPTION_MASK_ISA_CLDEMOTE.
23313 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
23314 (ix86_valid_target_attribute_inner_p): Ditto.
23315 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
23316 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
23317 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
23318 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
23319 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
23321 * config/i386/i386.opt: Add -mcldemote.
23322 * config/i386/x86intrin.h: New header.
23323 * doc/invoke.texi: Add -mcldemote.
23325 2018-05-14 Richard Biener <rguenther@suse.de>
23327 * doc/match-and-simplify.texi: Adjust :s documentation.
23329 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
23331 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
23332 intended memcpy size.
23333 (REORDER_45): Likewise.
23335 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
23337 * sort.cc: New file.
23338 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
23339 * vec.c (qsort_chk): Use gcc_qsort.
23340 * Makefile.in (OBJS-libcommon): Add sort.o.
23341 (build/sort.o): New target. Use it...
23342 (BUILD_RTL): ... here, and...
23343 (build/gencfn-macros): ... here, and...
23344 (build/genmatch): ... here.
23346 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
23347 Chung-Ju Wu <jasonwucj@gmail.com>
23349 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
23350 * config/nds32/nds32-graywolf.md: New file.
23351 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
23352 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
23354 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
23355 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
23356 * config/nds32/nds32.md (pipeline_model): Add graywolf.
23357 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
23358 * config/nds32/pipelines.md: Include n15 settings.
23360 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
23361 Chung-Ju Wu <jasonwucj@gmail.com>
23363 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
23364 * config/nds32/nds32-n13.md: New file.
23365 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
23366 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
23368 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
23369 * config/nds32/nds32.md (pipeline_model): Add n13.
23370 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
23371 * config/nds32/pipelines.md: Include n13 settings.
23373 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
23374 Chung-Ju Wu <jasonwucj@gmail.com>
23376 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
23377 * config/nds32/nds32-n10.md: New file.
23378 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
23379 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
23381 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
23382 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
23383 * config/nds32/nds32.md (pipeline_model): Add n10.
23384 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
23385 * config/nds32/pipelines.md: Include n10 settings.
23387 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
23388 Kito Cheng <kito.cheng@gmail.com>
23389 Chung-Ju Wu <jasonwucj@gmail.com>
23391 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
23392 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23393 Add enum values for DSP extension instructions.
23394 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
23396 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
23397 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
23398 New code iterators.
23399 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
23400 * config/nds32/nds32-dspext.md: New file for DSP implementation.
23401 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
23402 * config/nds32/nds32-intrinsic.md: Likewise.
23403 * config/nds32/nds32_intrinsic.h: Likewise.
23404 * config/nds32/nds32-md-auxiliary.c: Likewise.
23405 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
23406 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
23407 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
23408 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
23409 * config/nds32/nds32-protos.h: New declarations for DSP extension.
23410 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
23411 TYPE_DMAC in switch statement.
23412 * config/nds32/nds32.c: New checking and implementation for DSP
23413 extension instructions.
23414 * config/nds32/nds32.h: Likewise.
23415 * config/nds32/nds32.md: Likewise.
23416 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
23417 * config/nds32/predicates.md: Implement new predicates for DSP
23420 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
23422 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
23423 Reformat alternatives and attributes so it is easier to identify
23424 which constraints/attributes go with which instruction.
23425 (mov<mode>_hardfloat32, FMOVE64): Likewise.
23426 (mov<mode>_softfloat32, FMOVE64): Likewise.
23427 (mov<mode>_hardfloat64, FMOVE64): Likewise.
23428 (mov<mode>_softfloat64, FMOVE64): Likewise.
23430 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
23432 * doc/extend.texi (PowerPC Built-in Functions): Rename this
23434 (Basic PowerPC Built-in Functions): The new name of the
23435 subsection previously known as "PowerPC Built-in Functions".
23436 (Basic PowerPC Built-in Functions Available on all Configurations):
23438 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
23439 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
23440 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
23441 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
23443 2018-05-11 Martin Jambor <mjambor@suse.cz>
23446 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
23449 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
23452 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
23454 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
23456 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
23457 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
23458 (ix86_handle_option): Handle -mwaitpkg.
23459 * config.gcc: New header.
23460 * config/i386/cpuid.h (bit_WAITPKG): New bit.
23461 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
23462 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
23464 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23465 OPTION_MASK_ISA_WAITPKG.
23466 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
23467 (ix86_option_override_internal): Add PTA_WAITPKG.
23468 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
23469 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
23470 IX86_BUILTIN_TPAUSE.
23471 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
23472 __builtin_ia32_umwait and __builtin_ia32_tpause.
23473 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
23474 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
23475 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
23476 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
23477 UNSPECV_TPAUSE): New.
23478 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
23479 * config/i386/i386.opt: Add -mwaitpkg.
23480 * config/i386/waitpkgintrin.h: New file.
23481 * config/i386/x86intrin.h: New header.
23482 * doc/invoke.texi: Add -mwaitpkg.
23484 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
23487 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
23489 (cortex-m0): Use armv6s-m isa.
23490 (cortex-m0plus): Likewise.
23491 (cortex-m1): Likewise.
23492 (cortex-m0.small-multiply): Likewise.
23493 (cortex-m0plus.small-multiply): Likewise.
23494 (cortex-m1.small-multiply): Likewise.
23496 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
23497 Jakub Jelinek <jakub@redhat.com>
23499 PR tree-optimization/85692
23500 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
23501 source permute as well.
23503 2018-05-11 Martin Liska <mliska@suse.cz>
23506 * doc/extend.texi: Document LLVM style format for no_sanitize
23509 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23511 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
23512 mode_supports_vsx_dform_quad to mode_supports_dq_form.
23513 (mode_supports_vsx_dform_quad): Likewise.
23514 (mode_supports_vmx_dform): Move these functions to be next to the
23515 other mode_supports functions.
23516 (mode_supports_dq_form): Likewise.
23517 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
23518 mode_supports_dq_form.
23519 (reg_offset_addressing_ok_p): Likewise.
23520 (offsettable_ok_by_alignment): Likewise.
23521 (rs6000_legitimate_offset_address_p): Likewise.
23522 (legitimate_lo_sum_address_p): Likewise.
23523 (rs6000_legitimize_address): Likewise.
23524 (rs6000_legitimize_reload_address): Likewise.
23525 (rs6000_secondary_reload_inner): Likewise.
23526 (rs6000_preferred_reload_class): Likewise.
23527 (rs6000_output_move_128bit): Likewise.
23529 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
23531 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
23532 Generate SImode target register for null target.
23533 <case IX86_BUILTIN_XGETBV>: Ditto.
23534 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
23535 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
23537 2018-05-10 Carl Love <cel@us.ibm.com>
23539 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
23540 dcbtt and dcbtstt if operands[2] is 0.
23542 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
23545 * config/i386/sse.md (usadv64qi): New expander.
23547 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
23549 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
23550 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
23551 -maltivec=be support.
23552 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
23553 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
23554 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
23555 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
23556 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
23557 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
23558 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
23559 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
23560 altivec_vsumsws): Adjust.
23561 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
23562 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
23563 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
23565 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
23566 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
23567 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
23568 (altivec_lve<VI_char>x): Delete expand.
23569 (*altivec_lve<VI_char>x_internal): Rename to...
23570 (altivec_lve<VI_char>x): ... this.
23571 (altivec_lvxl_<mode>): Delete expand.
23572 (*altivec_lvxl_<mode>_internal): Rename to ...
23573 (altivec_lvxl_<mode>): ... this.
23574 (altivec_stvxl_<mode>): Delete expand.
23575 (*altivec_stvxl_<mode>_internal): Rename to ...
23576 (altivec_stvxl_<mode>): ... this.
23577 (altivec_stve<VI_char>x): Delete expand.
23578 (*altivec_stve<VI_char>x_internal): Rename to ...
23579 (altivec_stve<VI_char>x): ... this.
23580 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
23581 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
23582 reduc_plus_scal_<mode>): Adjust.
23583 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
23585 (rs6000_cpu_cpp_builtins): Adjust.
23586 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
23587 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
23588 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
23589 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
23590 -maltivec=be support.
23591 (rs6000_split_vec_extract_var): Adjust.
23592 (rs6000_split_v4si_init): Adjust.
23593 (swap_selector_for_mode): Delete.
23594 (altivec_expand_lvx_be, altivec_expand_stvx_be,
23595 altivec_expand_stvex_be): Delete.
23596 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
23597 -maltivec=be support.
23598 (rs6000_gimple_fold_builtin): Ditto.
23599 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
23601 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
23602 (TARGET_DIRECT_MOVE_64BIT): Adjust.
23603 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
23604 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
23605 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
23606 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
23607 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
23608 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
23609 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
23610 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
23611 anonymous split): Adjust.
23612 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
23613 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
23615 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
23617 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
23618 when --with-gxx-include-dir is also specified.
23619 * configure: Regenerate.
23621 2018-05-09 Jim Wilson <jimw@sifive.com>
23624 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
23625 * config/riscv/t-withmultilib: New.
23626 * config/riscv/withmultilib.h: New.
23627 * doc/install.texi: Document RISC-V --with-multilib-list support.
23629 2018-05-09 Richard Biener <rguenther@suse.de>
23631 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
23633 (vect_bb_vectorization_profitable_p): Adjust. Compute
23634 actual scalar cost using the cost vector and the add_stmt_cost
23637 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23639 PR rtl-optimization/85645
23640 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
23641 in the REG_CFA_REGISTER note for LR, don't leave it empty.
23643 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23645 PR rtl-optimization/85645
23646 * shrink-wrap.c (spread_components): Return a boolean saying if
23647 anything was changed.
23648 (try_shrink_wrapping_separate): Iterate spread_components until
23649 nothing changes anymore.
23651 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23653 PR rtl-optimization/85645
23654 * regrename.c (build_def_use): Also kill the chains that include the
23655 destination of a REG_CFA_REGISTER note.
23657 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23659 PR rtl-optimization/85645
23660 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
23661 insn that has a REG_CFA_REGISTER note.
23663 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
23665 * cfgexpand.c (expand_clobber): New function.
23666 (expand_gimple_stmt_1): Use it.
23667 * tree-vect-stmts.c (vect_clobber_variable): New function,
23669 (vectorizable_simd_clone_call): ...here.
23670 (vectorizable_store): Emit a clobber either side of an
23671 IFN_STORE_LANES sequence.
23672 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
23674 2018-05-09 Tom de Vries <tom@codesourcery.com>
23677 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
23678 (define_insn "trap_if_false"): Add exit after trap.
23680 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
23682 PR rtl-optimization/85638
23683 * bb-reorder.c: Include common/common-target.h.
23684 (create_forwarder_block): New function extracted from...
23685 (fix_up_crossing_landing_pad): ...here. Rename into...
23686 (dw2_fix_up_crossing_landing_pad): ...this.
23687 (sjlj_fix_up_crossing_landing_pad): New function.
23688 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
23689 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
23690 from both partitions and exit the loop after one iteration.
23692 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
23695 * doc/extend.texi (PowerPC Built-in Functions): Rename this
23697 (Basic PowerPC Built-in Functions): The new name of the
23698 subsection previously known as "PowerPC Built-in Functions".
23699 (Basic PowerPC Built-in Functions Available on all Configurations):
23701 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23703 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23705 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23707 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23710 2018-05-08 Jim Wilson <jimw@sifive.com>
23712 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
23713 (LD_EMUL_SUFFIX): New.
23714 (LINK_SPEC): Use it.
23716 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
23718 * doc/extend.texi (PowerPC Built-in Functions): Rename this
23720 (Basic PowerPC Built-in Functions): The new name of the
23721 subsection previously known as "PowerPC Built-in Functions".
23722 (Basic PowerPC Built-in Functions Available on all Configurations):
23724 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23726 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23728 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23730 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23733 2018-05-08 Jakub Jelinek <jakub@redhat.com>
23736 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
23737 after cmpelim optimization.
23739 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
23741 * config.gcc: Support "goldmont".
23742 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
23743 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23744 PROCESSOR_GOLDMONT.
23745 * config/i386/i386.c (m_GOLDMONT): Define.
23746 (processor_target_table): Add "goldmont".
23747 (PTA_GOLDMONT): Define.
23748 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
23749 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
23750 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
23751 (fold_builtin_cpu): Add "goldmont".
23752 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
23753 (ix86_option_override_internal): Add "goldmont".
23754 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
23755 (processor_type): Add PROCESSOR_GOLDMONT.
23756 * config/i386/i386.md: Add CPU "glm".
23757 * config/i386/glm.md: New file.
23758 * config/i386/x86-tune.def: Add m_GOLDMONT.
23759 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
23761 2018-05-08 Jakub Jelinek <jakub@redhat.com>
23764 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
23766 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
23767 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
23768 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
23771 * config/i386/i386.c (ix86_fold_builtin): Handle
23772 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
23775 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
23776 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
23778 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
23781 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
23782 (check_arch): Likewise.
23783 (check_fpu): Return the result rather than printing it.
23784 (end arch): Fix operator precedence.
23785 (end cpu): Likewise.
23786 (END): Print the result from check_fpu.
23788 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
23789 Alan Hayward <alan.hayward@arm.com>
23790 David Sherwood <david.sherwood@arm.com>
23792 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
23793 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
23794 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
23795 (*fcmuo<mode>_and): New patterns.
23797 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
23799 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
23800 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
23801 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
23802 (cmp_op, sve_imm_con): New code attributes.
23803 (SVE_COND_INT_CMP, imm_con): Delete.
23804 (cmp_op): Remove above unspecs from int attribute.
23805 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
23807 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
23808 comparison-specific unspecs.
23809 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
23810 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
23811 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
23812 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
23813 (*vec_fcm<cmp_op><mode>): Rename to...
23814 (*fcm<cmp_op><mode>): ...this and adjust likewise.
23815 (*vec_fcmuo<mode>): Rename to...
23816 (*fcmuo<mode>): ...this and adjust likewise.
23817 (*pred_fcm<cmp_op><mode>): New pattern.
23818 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
23819 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
23821 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
23823 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
23824 (aarch64_emit_sve_predicated_cond): New function.
23825 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
23826 (aarch64_emit_unspec_cond_or): Replace with...
23827 (aarch64_emit_sve_or_conds): ...this new function. Use
23828 aarch64_emit_sve_ptrue_op for the individual comparisons and
23829 aarch64_emit_binop to OR them together.
23830 (aarch64_emit_inverted_unspec_cond): Replace with...
23831 (aarch64_emit_sve_inverted_cond): ...this new function. Use
23832 aarch64_emit_sve_ptrue_op for the comparison and
23833 aarch64_emit_unop to invert the result.
23834 (aarch64_expand_sve_vec_cmp_float): Update after the above
23835 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
23837 2018-05-07 Nathan Sidwell <nathan@acm.org>
23839 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
23840 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
23841 (Backwards Compatibility): Likewise.
23843 2018-05-07 Luis Machado <luis.machado@linaro.org>
23847 2018-05-07 Luis Machado <luis.machado@linaro.org>
23849 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23850 <prefetch_dynamic_strides>: New const bool field.
23851 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23852 prefetch_dynamic_strides.
23853 (exynosm1_prefetch_tune): Likewise.
23854 (thunderxt88_prefetch_tune): Likewise.
23855 (thunderx_prefetch_tune): Likewise.
23856 (thunderx2t99_prefetch_tune): Likewise.
23857 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
23859 (aarch64_override_options_internal): Update to set
23860 PARAM_PREFETCH_DYNAMIC_STRIDES.
23861 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
23862 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
23863 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
23864 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
23865 prefetch-dynamic-strides setting.
23867 2018-05-07 Luis Machado <luis.machado@linaro.org>
23869 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23870 <minimum_stride>: New const int field.
23871 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23872 minimum_stride field.
23873 (exynosm1_prefetch_tune): Likewise.
23874 (thunderxt88_prefetch_tune): Likewise.
23875 (thunderx_prefetch_tune): Likewise.
23876 (thunderx2t99_prefetch_tune): Likewise.
23877 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
23878 (aarch64_override_options_internal): Update to set
23879 PARAM_PREFETCH_MINIMUM_STRIDE.
23880 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
23881 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
23882 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
23883 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
23884 stride is constant and is below the minimum stride threshold.
23886 2018-05-07 Luis Machado <luis.machado@linaro.org>
23888 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
23891 2018-05-07 Luis Machado <luis.machado@linaro.org>
23893 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23894 <prefetch_dynamic_strides>: New const bool field.
23895 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23896 prefetch_dynamic_strides.
23897 (exynosm1_prefetch_tune): Likewise.
23898 (thunderxt88_prefetch_tune): Likewise.
23899 (thunderx_prefetch_tune): Likewise.
23900 (thunderx2t99_prefetch_tune): Likewise.
23901 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
23903 (aarch64_override_options_internal): Update to set
23904 PARAM_PREFETCH_DYNAMIC_STRIDES.
23905 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
23906 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
23907 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
23908 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
23909 prefetch-dynamic-strides setting.
23911 2018-05-07 Luis Machado <luis.machado@linaro.org>
23913 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23914 <minimum_stride>: New const int field.
23915 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23916 minimum_stride field.
23917 (exynosm1_prefetch_tune): Likewise.
23918 (thunderxt88_prefetch_tune): Likewise.
23919 (thunderx_prefetch_tune): Likewise.
23920 (thunderx2t99_prefetch_tune): Likewise.
23921 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
23922 (aarch64_override_options_internal): Update to set
23923 PARAM_PREFETCH_MINIMUM_STRIDE.
23924 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
23925 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
23926 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
23927 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
23928 stride is constant and is below the minimum stride threshold.
23930 2018-05-06 Jakub Jelinek <jakub@redhat.com>
23933 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
23934 the type is addressable. Don't force op into register if it has
23937 2018-05-05 Roland McGrath <mcgrathr@google.com>
23940 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
23941 any section for which we don't know a specific type it should have,
23942 regardless of name. Previously this was done only for the exact
23943 names ".init_array", ".fini_array", and ".preinit_array".
23944 (default_elf_asm_named_section): Add comment about
23945 relationship with default_section_type_flags and SECTION_NOTYPE.
23946 (get_section): Don't consider it a type conflict if one side has
23947 SECTION_NOTYPE and the other doesn't, as long as neither has the
23948 SECTION_BSS et al used in the default_section_type_flags logic.
23950 2018-05-05 Tom de Vries <tom@codesourcery.com>
23953 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
23954 (workaround_barsyncs): New function.
23955 (nvptx_reorg): Use workaround_barsyncs.
23956 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
23957 (define_expand "nvptx_membar_cta"): New define_expand.
23958 (define_insn "*nvptx_membar_cta"): New insn.
23960 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
23962 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
23963 To improve optimization opportunities.
23964 * builtin-types.def: The new needed builtin types for the above.
23966 2018-05-04 Richard Biener <rguenther@suse.de>
23968 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
23969 * gimple-ssa-store-merging.c
23970 (imm_store_chain_info::output_merged_store): Remove redundant create,
23971 release split_store vector contents on failure.
23972 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
23973 scalar stmt vector on cache hit.
23975 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
23977 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
23979 * config.gcc (powerpc-xilinx-eabi*): Remove.
23980 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
23982 (fusion_addis_mem_combo_load): Ditto.
23983 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
23985 (rs6000_cpu_cpp_builtins): Ditto.
23986 * config/rs6000/rs6000-linux.c
23987 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
23988 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
23989 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
23991 (rs6000_setup_reg_addr_masks): Ditto.
23992 (rs6000_init_hard_regno_mode_ok): Ditto.
23993 (rs6000_option_override_internal): Ditto.
23994 (legitimate_lo_sum_address_p): Ditto.
23995 (rs6000_legitimize_address): Ditto.
23996 (rs6000_legitimize_reload_address): Ditto.
23997 (rs6000_legitimate_address_p): Ditto.
23998 (abi_v4_pass_in_fpr): Ditto.
23999 (setup_incoming_varargs): Ditto.
24000 (rs6000_gimplify_va_arg): Ditto.
24001 (rs6000_split_multireg_move): Ditto.
24002 (rs6000_savres_strategy): Ditto.
24003 (rs6000_emit_prologue_components): Ditto.
24004 (rs6000_emit_epilogue_components): Ditto.
24005 (rs6000_emit_prologue): Ditto.
24006 (rs6000_emit_epilogue): Ditto.
24007 (rs6000_elf_file_end): Ditto.
24008 (rs6000_function_value): Ditto.
24009 (rs6000_libcall_value): Ditto.
24010 * config/rs6000/rs6000.h: Ditto.
24011 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
24012 (TARGET_MINMAX): ... this. New.
24013 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
24014 * config/rs6000/rs6000.md: Remove Xilinx FP support.
24015 (*movsi_internal1_single): Delete.
24016 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
24017 mfpu=, mxilinx-fpu): Delete.
24018 * config/rs6000/singlefp.h: Delete.
24019 * config/rs6000/sysv4.h: Remove Xilinx FP support.
24020 * config/rs6000/t-rs6000: Ditto.
24021 * config/rs6000/t-xilinx: Delete.
24022 * config/rs6000/titan.md: Adjust for fp_type removal.
24023 * config/rs6000/vsx.md: Remove Xilinx FP support.
24024 (VStype_simple): Delete.
24025 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
24026 * config/rs6000/xfpu.h: Delete.
24027 * config/rs6000/xfpu.md: Delete.
24028 * config/rs6000/xilinx.h: Delete.
24029 * config/rs6000/xilinx.opt: Delete.
24030 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
24031 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
24033 2018-05-04 Tom de Vries <tom@codesourcery.com>
24036 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
24039 2018-05-04 Richard Biener <rguenther@suse.de>
24041 PR middle-end/85627
24042 * tree-complex.c (update_complex_assignment): We are always in SSA form.
24043 (expand_complex_div_wide): Likewise.
24044 (expand_complex_operations_1): Likewise.
24045 (expand_complex_libcall): Preserve EH info of the original stmt.
24046 (tree_lower_complex): Handle removed blocks.
24047 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
24048 on complex multiplication and division libcall builtins.
24050 2018-05-04 Richard Biener <rguenther@suse.de>
24052 PR middle-end/85574
24053 * fold-const.c (negate_expr_p): Restrict negation of operand
24054 zero of a division to when we know that can happen without
24056 (fold_negate_expr_1): Likewise.
24058 2018-05-04 Jakub Jelinek <jakub@redhat.com>
24061 * real.h (real_nextafter): Declare.
24062 * real.c (real_nextafter): New function.
24063 * fold-const-call.c (fold_const_nextafter): New function.
24064 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
24065 CASE_CFN_NEXTTOWARD.
24066 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
24067 even when arg1_mode is different from arg0_mode.
24069 2018-05-03 Nathan Sidwell <nathan@acm.org>
24071 * doc/extend.texi (Deprecated Features): Remove
24072 -ffriend-injection.
24073 (Backwards Compatibility): Likewise.
24074 * doc/invoke.texi (C++ Language Options): Likewise.
24075 (C++ Dialect Options): Likewise.
24077 2018-05-03 Jakub Jelinek <jakub@redhat.com>
24080 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
24081 _mm512_mask_mullox_epi64): New intrinsics.
24083 2018-05-03 Tom de Vries <tom@codesourcery.com>
24086 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24087 dump files): Add offload-tree.
24089 2018-05-03 Richard Biener <rguenther@suse.de>
24091 PR tree-optimization/85615
24092 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
24093 to loops not nested in BBs loop father to avoid creating multi-entry
24096 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24098 PR tree-optimization/70291
24099 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
24100 arguments. Change return type to tree. Emit libcall as a new
24101 statement rather than replacing existing one when inplace_p is true.
24102 (expand_complex_multiplication_components): New function.
24103 (expand_complex_multiplication): Expand floating-point complex
24104 multiplication using the above.
24105 (expand_complex_division): Rename inner_type parameter to type.
24106 Update expand_complex_libcall call-site.
24107 (expand_complex_operations_1): Update expand_complex_multiplication
24108 and expand_complex_division call-sites.
24110 2018-05-02 Jakub Jelinek <jakub@redhat.com>
24113 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
24114 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
24115 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
24116 the highest significant bit of the shift count mask is clear. In
24117 check whether and[sq]i3 is needed verify that all significant bits
24118 of the shift count other than the highest are set.
24120 2018-05-02 Tom de Vries <tom@codesourcery.com>
24123 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
24124 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
24125 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
24126 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
24127 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
24128 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
24129 __builtin_goacc_parlevel_size.
24131 2018-05-02 Richard Biener <rguenther@suse.de>
24133 PR tree-optimization/85597
24134 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
24135 do not use split vect_get_vec_defs call but call vect_get_slp_defs
24138 2018-05-02 Tom de Vries <tom@codesourcery.com>
24141 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24142 dump files): Add ltrans-tree.
24144 2018-05-02 Tom de Vries <tom@codesourcery.com>
24147 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24148 dump files): Add wpa-ipa.
24150 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
24152 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
24153 powerpc*-*-linux*paired* target.
24154 * config/rs6000/750cl.h: Delete.
24155 * config/rs6000/paired.h: Delete.
24156 * config/rs6000/paired.md: Delete.
24157 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
24159 * config/rs6000/rs6000-builtin.def: Remove paired float support.
24160 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
24161 comment. Remove paired float support.
24162 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
24163 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
24165 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
24166 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
24168 * config/rs6000/rs6000.c: Remove paired float support.
24169 (paired_expand_vector_init, paired_expand_vector_move,
24170 paired_emit_vector_compare, paired_emit_vector_cond_expr,
24171 (paired_expand_lv_builtin, paired_expand_stv_builtin,
24172 paired_expand_builtin, paired_expand_predicate_builtin,
24173 paired_init_builtins): Delete.
24174 * config/rs6000/rs6000.h: Remove paired float support.
24175 * config/rs6000/rs6000.md: Remove paired float support.
24176 (move_from_CR_ov_bit): Delete.
24177 * config/rs6000/rs6000.opt (mpaired): Delete.
24178 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
24179 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
24181 2018-05-02 Richard Biener <rguenther@suse.de>
24183 PR middle-end/85567
24184 * gimplify.c (gimplify_save_expr): When in SSA form allow
24185 SAVE_EXPRs to compute to SSA vars.
24187 2018-05-02 Jakub Jelinek <jakub@redhat.com>
24190 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
24191 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
24192 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
24193 clobber operands[2], instead use a new pseudo. Formatting fixes.
24195 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
24197 PR tree-optimization/85586
24198 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
24199 exit early for statements in the same group if the accesses are
24202 2018-05-02 Tom de Vries <tom@codesourcery.com>
24205 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
24208 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
24210 PR tree-optimization/85143
24211 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
24213 2018-05-01 Tom de Vries <tom@codesourcery.com>
24216 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
24217 not found" error message.
24219 2018-05-01 Tom de Vries <tom@codesourcery.com>
24222 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
24223 * vec.c (test_ordered_remove_if): New function.
24224 (vec_c_tests): Call test_ordered_remove_if.
24225 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
24226 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
24227 * tree-vect-patterns.c (vect_pattern_recog_1): Use
24228 VEC_ORDERED_REMOVE_IF.
24230 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24232 PR tree-optimization/82665
24233 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
24234 pointer subtraction where arguments come from a memchr call.
24236 2018-05-01 Jakub Jelinek <jakub@redhat.com>
24238 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
24239 --push-state --as-needed and --pop-state instead of --as-needed and
24240 --no-as-needed if ld supports it.
24241 * configure: Regenerated.
24244 * doc/install.texi2html: Replace _002d with - and _002a with * in
24245 generated html files using sed.
24247 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24250 * gcc-rich-location.c (blank_line_before_p): New function.
24251 (use_new_line): New function.
24252 (gcc_rich_location::add_fixit_insert_formatted): New function.
24253 * gcc-rich-location.h
24254 (gcc_rich_location::add_fixit_insert_formatted): New function.
24256 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24258 * selftest.c (assert_streq): Rename "expected" and "actual" to
24259 "val1" and "val2". Extend NULL-handling to cover both inputs
24260 symmetrically, while still requiring both to be non-NULL for a pass.
24261 * selftest.h (assert_streq): Rename "expected" and "actual" to
24263 (ASSERT_EQ): Likewise.
24264 (ASSERT_EQ_AT): Likewise.
24265 (ASSERT_KNOWN_EQ): Likewise.
24266 (ASSERT_KNOWN_EQ_AT): Likewise.
24267 (ASSERT_NE): Likewise.
24268 (ASSERT_MAYBE_NE): Likewise.
24269 (ASSERT_MAYBE_NE_AT): Likewise.
24270 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
24271 the assertion to pass.
24272 (ASSERT_STREQ_AT): Likewise.
24274 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
24276 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
24277 interaction with -pie.
24279 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24281 * selftest.h: Fix alphabetization of per-source-file selftest
24284 2018-04-30 Jason Merrill <jason@redhat.com>
24286 PR c++/61982 - dead stores to destroyed objects.
24287 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
24290 2018-04-30 Jason Merrill <jason@redhat.com>
24292 * tree.c (build_clobber): New.
24293 * tree.h: Declare it.
24294 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
24296 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24298 * diagnostic-show-locus.c (layout::layout): Update for
24299 location_get_source_line returning a char_span.
24300 (struct char_span): Move to input.h.
24301 (struct correction): Update for fields in char_span becoming
24303 (struct source_line): Update for location_get_source_line
24304 returning a char_span.
24305 (layout::print_line): Likewise.
24306 * edit-context.c (edited_file::print_content): Likewise.
24307 (edited_file::print_diff_hunk): Likewise.
24308 (edited_file::print_run_of_changed_lines): Likewise.
24309 (edited_file::get_num_lines): Likewise.
24310 (edited_line::edited_line): Likewise.
24311 * final.c (asm_show_source): Likewise.
24312 * input.c (location_get_source_line): Convert return type
24313 from const char * to char_span, losing the final "line_len"
24315 (dump_location_info): Update for the above.
24316 (get_substring_ranges_for_loc): Likewise. Use a char_span
24317 when handling the literal within the line.
24318 (test_reading_source_line): Update for location_get_source_line
24319 returning a char_span.
24320 * input.h (class char_span): Move here from
24321 diagnostic-show-locus.c, converting from a struct to a class.
24322 Make data members private.
24323 (char_span::operator bool): New.
24324 (char_span::length): New.
24325 (char_span::get_buffer): New.
24326 (char_span::operator[]): New.
24327 (char_span::subspan): Make const.
24328 (char_span::xstrdup): New.
24329 (location_get_source_line): Convert return type from const char *
24330 to char_span, losing the final "line_size" param.
24332 2018-04-30 Jan Hubicka <jh@suse.cz>
24334 * lto-wrapper.c (ltrans_priorities): New static var.
24335 (cmp_priority): New.
24336 (run_gcc): Read priorities and if doing parallel build order
24337 the Makefile by them.
24339 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24341 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
24343 2018-04-30 Richard Biener <rguenther@suse.de>
24345 * tree-cfg.c (verify_address): Remove base argument, add
24346 flag whether to check TREE_ADDRESSABLE and do that.
24347 (verify_expr): Remove.
24348 (verify_types_in_gimple_reference): Add pieces from verify_expr.
24349 (verify_gimple_assign_single): Likewise.
24350 (verify_gimple_switch): Likewise.
24351 (verify_expr_location_1): Dereference tp once. Add (disabled)
24352 piece from verify_expr.
24353 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
24355 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
24357 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
24359 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
24361 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
24362 (small_data_pattern): Likewise.
24363 (arc_rewrite_small_data): Likewise.
24364 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
24365 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
24366 (get_symbol_alignment): New function.
24367 (legitimate_small_data_address_p): Likewise.
24368 (legitimate_scaled_address): Update, call
24369 legitimate_small_data_address_p.
24370 (output_sdata): New static variable.
24371 (arc_print_operand): Update how we handle small data operands.
24372 (arc_print_operand_address): Likewise.
24373 (arc_legitimate_address_p): Update, use
24374 legitimate_small_data_address_p.
24375 (arc_rewrite_small_data_p): Remove.
24376 (arc_rewrite_small_data_1): Likewise.
24377 (arc_rewrite_small_data): Likewise.
24378 (small_data_pattern): Likewise.
24379 (compact_sda_memory_operand): Update to use
24380 legitimate_small_data_address_p and get_symbol_alignment.
24381 (prepare_move_operands): Don't rewite sdata pattern.
24382 (prepare_extend_operands): Remove.
24383 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
24385 (zero_extendqisi2): Likewise.
24386 (zero_extendhisi2): Likewise.
24387 (extendqihi2): Likewise.
24388 (extendqisi2): Likewise.
24389 (extendhisi2): Likewise.
24390 (addsi3): Likewise.
24391 (subsi3): Likewise.
24392 (andsi3): Likewise.
24393 * config/arc/constraints.md (Usd): Change it to memory constraint.
24395 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
24397 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
24398 as source of std instructions.
24399 * config/arc/arc.md (movsi_insn): Update pattern predicate to
24400 allow 6-bit constants as source for store instructions.
24401 (movdi_insn): Update instruction pattern to allow 6-bit constants
24402 as source for store instructions.
24404 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
24406 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
24408 2018-04-30 Nathan Sidwell <nathan@acm.org>
24409 Sandra Loosemore <sandra@codesourcery.com>
24411 * dumpfile.c (dump_open): Allow '-' for stdout.
24412 * doc/invoke.texi (Developer Options): Document dump filename
24413 determination early. Document stdin/stdout selection.
24415 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
24417 Microblaze Target: PIC data text relative
24419 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
24420 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
24422 * config/microblaze/microblaze.h (microblaze_constant_address_p):
24423 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
24424 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
24425 New addressing mode for data-text relative position indepenedent code.
24426 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
24427 'ADDRESS_SYMBOLIC_TXT_REL'.
24428 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
24429 (microblaze_legitimate_pic_operand): Exclude function calls from
24430 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
24431 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
24433 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
24434 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
24435 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
24436 for 'address + offset'.
24437 (microblaze_expand_prologue): Add new function prologue call for
24439 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
24440 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
24441 table in case of TARGET_PIC_DATA_TEXT_REL.
24442 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
24443 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
24444 Add new macros 'UNSPEC_TEXT',
24445 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
24446 + exclude function calls from 'UNSPEC_PLT' in case of data text
24448 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
24449 new target hook for generating address diff vector tables in case of
24451 * doc/tm.texi : Regenerate.
24452 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
24453 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
24454 of addr diff vector generation.
24455 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
24456 target hook definition.
24457 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
24458 Add default function for generate_pic_addr_diff_vec -> flag_pic.
24459 * doc/invoke.texi (Add new pic option): Add new microblaze pic
24460 option for data text relative.
24462 2018-04-30 Richard Biener <rguenther@suse.de>
24464 * tree-chrec.h (evolution_function_is_constant_p): Remove
24466 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
24468 2018-04-30 Richard Biener <rguenther@suse.de>
24471 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
24473 2018-04-30 Richard Biener <rguenther@suse.de>
24475 PR tree-optimization/28364
24476 PR tree-optimization/85275
24477 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
24478 copying first exit test.
24480 2018-04-28 Mark Wielaard <mark@klomp.org>
24482 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
24483 dwarf_version >= 5.
24484 (dwarf_AT): Handle DW_AT_addr_base.
24485 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
24487 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
24490 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
24491 (*ashl<dwi>3_doubleword_mask_1): Ditto.
24492 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
24493 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
24495 2018-04-28 Richard Biener <rguenther@suse.de>
24497 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
24498 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
24499 to reflect use. Only add interesting stmts.
24501 2018-04-27 Martin Jambor <mjambor@suse.cz>
24504 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
24505 the jump function allows for passing through aggregate values.
24507 2018-04-27 David Malcolm <dmalcolm@redhat.com>
24509 * input.h (in_system_header_at): Convert from macro to inline
24511 (from_macro_expansion_at): Likewise.
24512 (from_macro_definition_at): Likewise.
24514 2018-04-27 Jeff Law <law@redhat.com>
24516 * config.gcc: Mark tile* targets as deprecated/obsolete.
24518 2018-04-27 Richard Biener <rguenther@suse.de>
24520 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
24523 2018-04-27 Richard Biener <rguenther@suse.de>
24525 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
24527 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
24529 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
24530 with Yd constraint. Set "preferred_for_speed" attribute from
24531 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
24532 with Yd constraint.
24533 (*movdi_internal): Ditto.
24534 (movti_interunit splitters): Remove
24535 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
24536 (movdi_interunit splitters): Ditto.
24537 * config/i386/constraints.md (Ye): Remove.
24538 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
24540 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24543 * config/aarch64/constraints.md (Usg): Limit to 31.
24544 (Usj): Limit to 63.
24546 2018-04-27 Jakub Jelinek <jakub@redhat.com>
24548 PR tree-optimization/85529
24549 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
24550 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
24551 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
24552 zero extension or masking of the MSB bit.
24553 (optimize_range_tests): Add FIRST_BB argument, pass it through
24554 to optimize_range_tests_var_bound.
24555 (maybe_optimize_range_tests, reassociate_bb): Adjust
24556 optimize_range_tests callers.
24558 2018-04-26 Richard Biener <rguenther@suse.de>
24559 Jakub Jelinek <jakub@redhat.com>
24561 * cgraph.h (symbol_table): Just declare debug method here.
24562 * symtab.c (symbol_table::debug): Define.
24564 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
24566 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
24568 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
24570 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
24571 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
24572 (*movdi_internal): Substitute Yi and Yj constraint with x
24573 and Ym and Yn constraint with y constraint. Update "isa"
24574 attribute and set "preferred_for_speed" attribute from
24575 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
24576 (*movsi_internal): Ditto.
24577 (*movdf_internal): Ditto.
24578 (*movsf_internal): Ditto.
24579 (*zero_extendsidi2): Ditto.
24580 * config/i386/sse.md (vec_set<mode>_0): Ditto.
24581 (sse2_loadld): Ditto.
24582 (*vec_extract<ssevecmodelower>_0): Ditto.
24583 (*vec_extractv4si_0_zext_sse4): Ditto.
24584 (vec_concatv2di): Ditto.
24585 (*vec_dup<mode>): Ditto.
24586 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
24587 * config/i386/constraints.md (Yi): Remove.
24592 2018-04-26 Nathan Sidwell <nathan@acm.org>
24594 * dumpfile.c (dump_open): New.
24595 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
24596 (dump_finish): Detect stdio/stderr by value not name.
24598 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
24600 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
24602 2018-04-26 Tom de Vries <tom@codesourcery.com>
24605 * config/nvptx/nvptx.c (verify_neutering_jumps)
24606 (verify_neutering_labels): New function
24607 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
24609 2018-04-26 Tom de Vries <tom@codesourcery.com>
24612 * config/nvptx/nvptx.c (needs_neutering_p): New function.
24613 (nvptx_single): Use needs_neutering_p to skip over insns that do not
24616 2018-04-26 Richard Biener <rguenther@suse.de>
24617 Tom de Vries <tom@codesourcery.com>
24620 * lto-streamer-out.c (output_function): Fixup loops if required to match
24621 discovery done in the reader.
24623 2018-04-26 Richard Biener <rguenther@suse.de>
24625 PR tree-optimization/85116
24626 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
24627 have a loop exit from the single latch predecessor. Remove
24628 case of header with just condition.
24629 (ch_base::copy_headers): Exclude infinite loops from any
24631 (pass_ch::execute): Record exits.
24633 2018-04-26 Richard Biener <rguenther@suse.de>
24635 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
24636 prologue cost vector and pass it to vect_get_load_cost.
24637 (vect_get_peeling_costs_all_drs): Likewise.
24638 (vect_peeling_hash_get_lowest_cost): Likewise.
24639 (vect_enhance_data_refs_alignment): Likewise.
24641 2018-04-26 Richard Biener <rguenther@suse.de>
24643 PR middle-end/85450
24644 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
24645 checking of integer<->pointer conversions.
24646 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
24647 sign-/zero-extending pointer types.
24648 (expand_omp_for_static_chunk): Likewise.
24650 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
24651 Jean Lee <xiaoyur347@gmail.com>
24653 * config/mips/mips.c (mips_asan_shadow_offset): New function.
24654 (TARGET_ASAN_SHADOW_OFFSET): Define.
24655 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
24656 true for -fsanitize=address.
24658 2018-04-25 Mark Wielaard <mark@klomp.org>
24660 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
24663 2018-04-25 Jakub Jelinek <jakub@redhat.com>
24665 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
24666 than "alu", remove explicit "memory" and "imm_disp" attributes.
24667 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
24669 PR middle-end/85414
24670 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
24671 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
24672 gen_lowpart_no_emit.
24674 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
24677 * config/i386/i386.c (ix86_expand_builtin): Change memory
24678 operand to XI, extend p0 to Pmode.
24679 * config/i386/i386.md: Change unspec volatile and operand
24680 1 mode to XI, change operand 0 mode to P.
24682 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24684 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
24685 GET_MODE_MASK before any checking.
24686 (nds32_can_use_bset_p): Likewise.
24687 (nds32_can_use_btgl_p): Likewise.
24689 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24691 * config/nds32/nds32-doubleword.md: New define_split pattern for
24692 illegal register number.
24694 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24696 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
24698 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24700 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
24702 2018-04-25 Richard Biener <rguenther@suse.de>
24704 * lto-streamer.h (LTO_major_version): Bump to 8.
24706 2018-04-25 Jakub Jelinek <jakub@redhat.com>
24708 * BASE-VER: Set to 9.0.0.
24710 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
24712 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
24713 in __abskf2 and __powikf2.
24715 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24718 * config/aarch64/constraints.md (Usg, Usj): New constraints.
24719 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
24720 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
24721 Use the above on operand 2. Reindent.
24722 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
24724 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
24727 * common/config/i386/i386-common.c (ix86_handle_option): Don't
24729 * config/i386/i386.opt (mcet): Removed.
24730 * doc/install.texi: Remove -mcet documentation.
24731 * doc/invoke.texi: Likewise.
24733 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
24736 * doc/install.texi: Remove -mcet from bootstrap-cet.
24738 2018-04-24 Jakub Jelinek <jakub@redhat.com>
24741 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
24742 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
24746 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
24747 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
24748 containing a CONST_VECTOR.
24750 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
24752 * doc/install.texi: Update newlib dependency for nvptx.
24754 2018-04-24 Jakub Jelinek <jakub@redhat.com>
24757 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
24758 instead of INTVAL when shifting x left.
24760 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
24762 PR tree-optimization/85478
24763 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
24764 vect_grouped_store_supported for single element vectors.
24766 2018-04-24 Richard Biener <rguenther@suse.de>
24769 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
24770 load cost increase to the case of non-constant step.
24772 2018-04-24 Jakub Jelinek <jakub@redhat.com>
24775 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
24776 destination if any_malformed_asm.
24778 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
24780 PR middle-end/85496
24781 * expr.c (store_field): In the bitfield case, if the value comes from
24782 a function call and is returned in registers by means of a PARALLEL,
24783 do not change the mode of the temporary unless BLKmode and VOIDmode.
24785 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
24787 PR rtl-optimization/85423
24788 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
24789 dependencies to debug insns when the previous insn is non-debug.
24791 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
24793 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
24794 enums into a single definition.
24795 (fls): Fix predicates and printing.
24798 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
24800 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
24801 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
24802 and short u6 immediate.
24803 (check_if_valid_sleep_operand): Remove.
24804 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
24806 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24808 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
24809 flag_always_save_lp condition.
24810 * config/nds32/nds32.opt (malways-save-lp): New option.
24812 2018-04-22 Shiva Chen <shiva0217@gmail.com>
24814 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
24815 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
24816 * config/nds32/nds32.h
24817 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
24818 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
24820 2018-04-22 Shiva Chen <shiva0217@gmail.com>
24822 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
24823 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
24825 2018-04-22 Shiva Chen <shiva0217@gmail.com>
24826 Chung-Ju Wu <jasonwucj@gmail.com>
24828 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
24830 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
24831 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
24833 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24835 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
24837 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24839 * config/nds32/nds32-protos.h (nds32_data_alignment,
24840 nds32_local_alignment): Declare.
24841 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
24842 nds32_local_alignment): New functions.
24843 (TARGET_CONSTANT_ALIGNMENT): Define.
24844 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
24846 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24848 * config/nds32/nds32.c
24849 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
24850 (TARGET_MODES_TIEABLE_P): Likewise.
24852 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24854 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
24855 level Ofast and Og.
24857 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
24858 Chung-Ju Wu <jasonwucj@gmail.com>
24860 * config/nds32/constants.md (unspec_volatile_element): Add enum values
24861 for unaligned access.
24862 * config/nds32/nds32-intrinsic.c: Implementation of expanding
24864 * config/nds32/nds32-intrinsic.md: Likewise.
24865 * config/nds32/nds32_intrinsic.h: Likewise.
24866 * config/nds32/nds32.h (nds32_builtins): Likewise.
24867 * config/nds32/nds32.opt (munaligned-access): New option.
24868 * config/nds32/nds32.c (nds32_asm_file_start): Display
24869 flag_unaligned_access status.
24871 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
24873 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
24874 -mno-relax is present.
24875 * config/riscv/linux.h (LINK_SPEC): Ditto.
24877 2018-04-20 Martin Sebor <msebor@redhat.com>
24880 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
24882 (gimple_fold_builtin_stxcpy_chk): Same.
24883 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
24885 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
24888 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
24889 __powikf2 when long double is IEEE 128-bit.
24891 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
24893 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
24894 step to make sure stack always aligned.
24896 2018-04-20 Carl Love <cel@us.ibm.com>
24899 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
24900 size check for arg0.
24902 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
24903 Tom de Vries <tom@codesourcery.com>
24906 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
24907 Emit insns for calls too.
24908 (nvptx_find_par): Always look for worker-level predecessor insn.
24909 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
24911 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
24912 (nvptx_process_pars): Propagate frames for calls.
24914 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
24917 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
24919 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
24920 (ix86_handle_option): Don't handle OPT_mibt.
24921 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
24923 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
24925 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
24926 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
24927 (ix86_target_macros): Define __CET__ with flag_cf_protection
24928 for -fcf-protection.
24929 * config/i386/i386.c (isa2_opts): Remove -mibt.
24930 * config/i386/i386.h (TARGET_IBT): Removed.
24931 (TARGET_IBT_P): Likewise.
24932 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
24933 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
24934 * config/i386/i386.opt (mcet): Update help message.
24935 (mshstk): Likewise.
24937 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
24938 -mcet as an alias for -mshstk.
24940 2018-04-20 Richard Biener <rguenther@suse.de>
24942 PR middle-end/85475
24943 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
24944 complexity by forcing a single use of the multiply operand.
24946 2018-04-20 Martin Jambor <mjambor@suse.cz>
24949 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
24950 recursion dependency to only apply to non-clones.
24952 2018-04-20 Martin Jambor <mjambor@suse.cz>
24955 * ipa-cp.c (create_specialized_node): Check that clones of
24956 self-recursive edges exist during IPA-CP.
24958 2018-04-19 Toon Moene <toon@moene.org>
24960 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
24963 2018-04-19 Jakub Jelinek <jakub@redhat.com>
24965 PR tree-optimization/85467
24966 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
24967 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
24968 VECTOR_CST element to type.
24970 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
24973 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
24974 * config/i386/i386.md (builtin_setjmp_setup): Removed.
24975 (builtin_longjmp): Likewise.
24976 (save_stack_nonlocal): New pattern.
24977 (restore_stack_nonlocal): Likewise.
24979 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
24982 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
24983 Replace ASM_OUTPUT_LABEL with fprintf.
24985 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
24988 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
24989 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
24990 * config/i386/i386-c.c (ix86_target_macros_internal): Also
24991 define __IBT__ and __SHSTK__ for -fcf-protection.
24992 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
24994 (ix86_trampoline_init): Likewise.
24995 (x86_output_mi_thunk): Likewise.
24996 (ix86_notrack_prefixed_insn_p): Likewise.
24997 (ix86_option_override_internal): Don't disallow -fcf-protection.
24998 * config/i386/i386.md (rdssp<mode>): Also enable for
25000 (incssp<mode>): Likewise.
25001 (nop_endbr): Likewise.
25002 * config/i386/i386.opt (mcet): Change help message to built-in
25005 (mshstk): Likewise.
25006 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
25007 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
25008 enable CET built-in functions.
25010 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
25012 * common/config/i386/i386-common.c
25013 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
25014 OPTION_MASK_ISA_MOVDIRI_UNSET,
25015 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
25016 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
25017 * config.gcc (movdirintrin.h): New header.
25018 * config/i386/cpuid.h (bit_MOVDIRI,
25019 bit_MOVDIR64B): New bits.
25020 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
25022 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
25023 (VOID, PVOID, PCVOID)): New function types.
25024 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
25025 __builtin_ia32_directstoreu_u64,
25026 __builtin_ia32_movdir64b): New builtins.
25027 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
25028 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
25030 (ix86_valid_target_attribute_inner_p): Ditto.
25031 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
25032 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
25033 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
25034 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
25035 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
25036 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
25037 (movdiri<mode>, movdir64b_<mode>): New.
25038 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
25039 * config/i386/immintrin.h: Include movdirintrin.h.
25040 * config/i386/movdirintrin.h: New file.
25041 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
25043 2018-04-19 Richard Biener <rguenther@suse.de>
25045 PR middle-end/85455
25046 * cfg.c (clear_bb_flags): When loop state says we have
25047 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
25049 2018-04-19 Richard Biener <rguenther@suse.de>
25051 PR tree-optimization/84737
25052 * tree-vect-data-refs.c (vect_copy_ref_info): New function
25053 copying restrict info.
25054 (vect_setup_realignment): Use it.
25055 * tree-vectorizer.h (vect_copy_ref_info): Declare.
25056 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
25057 the first DR to all generated stores.
25058 (vectorizable_load): Likewise for loads.
25060 2018-04-19 Jakub Jelinek <jakub@redhat.com>
25062 PR tree-optimization/85446
25063 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
25064 the integral and pointer types to have the same precision.
25066 * doc/install.texi: Document --disable-cet being the default and
25069 2018-04-18 Martin Liska <mliska@suse.cz>
25071 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
25074 2018-04-18 Martin Liska <mliska@suse.cz>
25077 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
25080 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
25081 arguments if they are comparable.
25083 2018-04-18 Martin Liska <mliska@suse.cz>
25086 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
25089 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
25092 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
25095 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
25096 ENDBR after calling __morestack.
25098 2018-04-18 David Malcolm <dmalcolm@redhat.com>
25101 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
25102 by using gcc_base_ver to generate a gcc_driver_version, and use
25103 it when generating GCC_DRIVER_NAME.
25104 * configure: Regenerate.
25106 2018-04-18 Jakub Jelinek <jakub@redhat.com>
25109 * config.gcc: Obsolete powerpc*-*-*spe*.
25111 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25114 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
25115 (stabstr_D): Change type of unum from unsigned int to
25116 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
25119 2018-04-17 Jim Wilson <jimw@sifive.com>
25122 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
25123 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
25124 Set arg_pointer_offset after using pretend_args_size.
25126 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25128 PR rtl-optimization/85431
25129 * dse.c (record_store): Ignore zero width stores.
25132 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
25133 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
25134 __builtin_stack_restore rather than after it.
25135 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
25136 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
25137 argument instead of virtual_dynamic_stack_rtx.
25139 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
25141 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
25143 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25144 Add note to error message to explain internal mapping of overloaded
25145 built-in function name to non-overloaded built-in function name.
25146 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
25149 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
25152 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
25153 where the inputs overlap with the output.
25155 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25157 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
25158 (=v, v) alternative and explicit "memory" attribute.
25159 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
25160 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25162 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
25163 "sselog1" type instead of "sselog".
25164 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
25165 "sselog". Remove explicit "memory" attribute.
25166 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
25167 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25169 (vec_extract_hi_v32hi): Merge all alternatives into one, use
25170 "sselog1" type instead of "sselog". Remove explicit "memory"
25172 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
25173 use "sselog1" type instead of "sselog". Remove explicit "memory"
25175 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
25176 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25178 (vec_extract_hi_v64qi): Merge all alternatives into one, use
25179 "sselog1" type instead of "sselog". Remove explicit "memory"
25181 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
25182 use "sselog1" type instead of "sselog". Remove explicit "memory"
25186 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
25188 PR middle-end/85414
25189 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
25192 2018-04-17 Martin Jambor <mjambor@suse.cz>
25195 * ipa-cp.c (create_specialized_node): Call
25196 expand_all_artificial_thunks if necessary.
25198 2018-04-17 Martin Liska <mliska@suse.cz>
25201 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
25202 in message, remote space in between '_G' and '('.
25204 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25207 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
25208 avx512f_vmcmp<mode>3<round_saeonly_name>,
25209 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
25210 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
25211 avx512f_rndscale<mode><round_saeonly_name>,
25212 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
25213 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
25214 Use %<iptr>2 instead of %2 for -masm=intel.
25215 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
25216 avx512f_vcvttss2usi<round_saeonly_name>,
25217 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
25219 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
25220 avx512f_vcvttsd2usi<round_saeonly_name>,
25221 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
25222 Use %q1 instead of %1 for -masm=intel.
25223 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
25224 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
25225 of %3 for -masm=intel.
25226 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
25228 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
25230 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
25232 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
25233 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
25235 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
25237 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
25238 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
25239 %g1 and one with %0 and %1.
25240 (avx512er_vmrcp28<mode><round_saeonly_name>,
25241 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
25242 %1 for -masm=intel.
25243 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
25244 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
25245 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
25246 of %0 and %{%4%} for -masm=intel.
25247 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
25248 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
25249 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
25250 order of %0 and %{%5%}%{z%} for -masm=intel.
25252 2018-04-17 Jan Hubicka <jh@suse.cz>
25255 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
25257 2018-04-17 Martin Liska <mliska@suse.cz>
25260 * multiple_target.c (create_dispatcher_calls): Set apostrophes
25261 for target_clone error message. Make default implementation
25262 clone to be a local declaration.
25263 (separate_attrs): Add new argument and check for an empty
25265 (expand_target_clones): Handle it.
25266 (ipa_target_clone): Make redirection just for target_clones
25269 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
25270 Tom de Vries <tom@codesourcery.com>
25272 PR middle-end/84955
25273 * omp-expand.c (expand_oacc_for): Add dummy false branch for
25274 tiled basic blocks without omp continue statements.
25276 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
25279 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
25280 vec_extract expression as having side effects to make sure it gets
25283 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
25286 * config/i386/i386.c (get_builtin_code_for_version): Check
25289 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
25292 * config.gcc: Support "skylake".
25293 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
25295 * config/i386/i386.c (m_SKYLAKE): Define.
25296 (processor_target_table): Add "skylake".
25297 (ix86_option_override_internal): Add "skylake".
25298 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
25299 PROCESSOR_CANNONLAKE.
25300 (get_builtin_code_for_version): Fix priority for
25301 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
25302 PROCESSOR_SKYLAKE-AVX512.
25303 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
25304 (processor_type): Add PROCESSOR_SKYLAKE.
25306 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
25307 Jason Merrill <jason@redhat.com>
25310 * convert.c (convert_to_integer_1): Use direct recursion for
25311 enumeral types and types with a precision less than the number
25312 of bits in their mode.
25314 2018-04-16 Julia Koval <julia.koval@intel.com>
25317 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
25318 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
25320 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
25323 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
25324 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
25325 and -mno-direct-move.
25327 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
25330 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
25331 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
25332 Ensure negative shifts result in {0}.
25334 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
25336 PR rtl-optimization/79916
25337 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
25338 regs (if any) to define how to gnerate SD moves when LRA is in
25341 2018-04-13 Jakub Jelinek <jakub@redhat.com>
25343 PR rtl-optimization/85393
25344 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
25345 * except.c (expand_dw2_landing_pad_for_region): Make static.
25346 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
25347 a label and unconditional jump to old_bb, rather than
25348 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
25351 PR rtl-optimization/85376
25352 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
25353 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
25354 instead of a specific value.
25356 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
25357 Bin Cheng <bin.cheng@arm.com>
25359 PR tree-optimization/82965
25360 PR tree-optimization/83991
25361 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
25362 by_profile_only parameter.
25363 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
25364 information if the loop was predicted to iterate too many times.
25365 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
25367 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
25370 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
25373 2018-04-13 Martin Liska <mliska@suse.cz>
25374 Jakub Jelinek <jakub@redhat.com>
25376 PR middle-end/81657
25377 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
25378 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
25379 * builtins.c (expand_builtin_memory_copy_args): Use
25380 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
25381 handle dest_addr == pc_rtx.
25383 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
25386 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
25387 asked to not generate direct moves.
25388 (fix_trunc<mode>si2_stfiwx): Similar.
25389 (fix_trunc<mode>si2_internal): Similar.
25391 2018-04-12 Jakub Jelinek <jakub@redhat.com>
25394 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
25395 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
25396 lookup if dest in some wider mode is known to be const0_rtx and
25397 if so, record permanent equivalence for it to be ZERO_EXTEND of
25398 the narrower mode destination.
25400 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
25402 * lto-streamer-out.c (output_function): Revert 259346.
25403 * omp-expand.c (expand_oacc_for): Likewise.
25405 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
25407 PR rtl-optimization/85354
25408 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
25409 * sel-sched.c (sel_global_init): ... here.
25411 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
25414 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
25415 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
25416 mode for PE-COFF targets.
25417 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
25418 (i386_pe_asm_lto_end): Likewise.
25419 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
25420 (TARGET_ASM_LTO_END): Likewise.
25421 * config/i386/winnt.c (saved_debug_info_level): New static variable.
25422 (i386_pe_asm_lto_start): New function.
25423 (i386_pe_asm_lto_end): Likewise.
25425 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
25426 Richard Biener <rguenther@suse.de>
25428 PR middle-end/84955
25429 * lto-streamer-out.c (output_function): Fix CFG loop state before
25431 * omp-expand.c (expand_oacc_for): Handle calls to internal
25432 functions like regular functions.
25434 2018-04-12 Richard Biener <rguenther@suse.de>
25437 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
25438 for the early LTO debug to properly generate references to it
25439 during DIE emission. Do not re-use that for the skeleton for
25441 (dwarf2out_early_finish): Likewise.
25443 2018-04-12 Jakub Jelinek <jakub@redhat.com>
25446 * config/i386/sse.md
25447 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
25448 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
25449 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
25450 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
25451 and output is a reg, avoid creating invalid lowpart subreg, but
25452 instead split into a 512-bit move. Don't split if not AVX512VL,
25453 input is xmm16+ reg and output is a mem.
25454 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
25455 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
25456 xmm16+ reg and output is a mem.
25458 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
25460 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
25461 also for flag_dwarf2_cfi_asm.
25463 2018-04-12 Jakub Jelinek <jakub@redhat.com>
25465 PR rtl-optimization/85342
25466 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
25467 a bool scalar var inside of the loop instead. Don't try to update
25468 recog_data.operand after failed apply_change_group.
25470 2018-04-12 Tom de Vries <tom@codesourcery.com>
25473 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
25474 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
25475 array with flexible array member as array without given dimension.
25476 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
25477 argument for undefined param to true.
25479 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
25482 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
25483 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
25484 from PowerPC section.
25485 * config/rs6000/sysv4.opt (mcall-): Improve help text.
25486 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
25487 help text that is too long.
25488 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
25489 help text that is too long.
25490 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
25491 help text that is too long.
25493 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
25495 * config/alpha/alpha.md (stack_probe_internal): Rename
25496 from "probe_stack". Update all callers.
25498 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
25500 PR rtl-optimization/84566
25501 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
25502 sched_macro_fuse_insns.
25504 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
25507 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
25508 (compute_block_dependences): ... from here.
25510 2018-04-11 Jakub Jelinek <jakub@redhat.com>
25512 PR tree-optimization/85331
25513 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
25514 from int to HOST_WIDE_INT.
25516 2018-04-11 Martin Jambor <mjambor@suse.cz>
25519 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
25520 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
25521 not the same as the source val.
25522 (cgraph_edge_brings_value_p): New parameter.
25523 (gather_edges_for_value): Pass destination value to
25524 cgraph_edge_brings_value_p.
25525 (perhaps_add_new_callers): Likewise.
25526 (get_info_about_necessary_edges): Likewise and exclude values brought
25527 only by self-recursive edges.
25528 (create_specialized_node): Redirect only clones of self-calling edges.
25529 (+self_recursive_pass_through_p): New function.
25530 (find_more_scalar_values_for_callers_subset): Use it.
25531 (find_aggregate_values_for_callers_subset): Likewise.
25532 (known_aggs_to_agg_replacement_list): Removed.
25533 (decide_whether_version_node): Re-calculate known constants for all
25534 remaining context clones.
25536 2018-04-11 Richard Biener <rguenther@suse.de>
25539 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
25540 from early DWARF output.
25541 (dwarf2out_early_finish): Output line info unconditionally into
25542 early DWARF and add reference to it.
25544 2018-04-11 Jakub Jelinek <jakub@redhat.com>
25547 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
25548 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
25549 other than V2DFmode using iptr mode attribute.
25550 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
25552 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
25554 PR rtl-optimization/84659
25555 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
25557 2018-04-11 Jakub Jelinek <jakub@redhat.com>
25560 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
25562 (output_loc_list): Pass address of a dummy size variable even in the
25563 locview handling loop.
25564 (index_location_lists): Add comment on why skip_loc_list_entry can't
25567 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
25570 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
25573 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
25576 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
25577 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
25578 and -mstring-compare-inline-limit.
25580 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
25583 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
25584 for stack clash protection in a register whenever we need it to be in
25587 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
25589 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
25590 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
25592 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
25595 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
25597 (mlong-double-): Ditto.
25598 * config/rs6000/sysv4.opt (msdata=): Ditto.
25599 (mtls-size=): Ditto.
25601 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
25603 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
25604 erroneous entries for
25605 "vector int vec_ldl (int, long int *)", and
25606 "vector unsigned int vec_ldl (int, unsigned long int *)".
25607 Add comments and entries for
25608 "vector bool char vec_ldl (int, bool char *)",
25609 "vector bool short vec_ldl (int, bool short *)",
25610 "vector bool int vec_ldl (int, bool int *)",
25611 "vector bool long long vec_ldl (int, bool long long *)",
25612 "vector pixel vec_ldl (int, pixel *)",
25613 "vector long long vec_ldl (int, long long *)",
25614 "vector unsigned long long vec_ldl (int, unsigned long long *)".
25615 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
25616 type tree bool_long_long_type_node and correct definition of
25617 bool_V2DI_type_node to make reference to this new type tree.
25618 (rs6000_mangle_type): Replace erroneous reference to
25619 bool_long_type_node with bool_long_long_type_node.
25620 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
25621 comments to emphasize sign distinctions for char and int types and
25622 replace RS6000_BTI_bool_long constant with
25623 RS6000_BTI_bool_long_long constant. Also add comment to restrict
25624 use of RS6000_BTI_pixel.
25625 (bool_long_type_node): Remove this macro definition.
25626 (bool_long_long_type_node): New macro definition
25628 2018-04-10 Jakub Jelinek <jakub@redhat.com>
25630 PR rtl-optimization/85300
25631 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
25632 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
25633 simplify_unary_operation fails.
25635 2018-04-10 Martin Liska <mliska@suse.cz>
25637 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
25638 cgraph_edge and ipa_ref.
25640 2018-04-10 Jakub Jelinek <jakub@redhat.com>
25644 * config/i386/sse.md
25645 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
25646 computation of the VEC_MERGE selector from mask.
25647 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
25648 Fix decoding of the VEC_MERGE selector into mask.
25650 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
25652 PR tree-optimization/85286
25653 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
25655 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
25657 * final.c (final_1): Set insn_last_address as well as
25658 insn_current_address.
25660 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25663 * explow.c (emit_stack_probe): Call validize_mem on memory location
25664 before passing it to gen_probe_stack. Create address operand and
25665 legitimize it for the probe_stack_address case.
25667 2018-04-09 Jan Hubicka <jh@suse.cz>
25670 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
25671 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
25672 * tree.c (free_lang_data_in_type): Fix handling of binfos;
25674 (free_lang_data): Rebuild type inheritance graph.
25676 2018-04-09 Martin Sebor <msebor@redhat.com>
25678 * invoke.texi (-finline-small-functions): Mention other optimization
25680 (-findirect-inlining, -fpartial-inlining): Same.
25681 (-finline-functions-called-once): Same.
25682 (-freorder-blocks-and-partition): Same.
25684 2018-04-09 Jan Hubicka <jh@suse.cz>
25687 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
25688 jumps; choose last target that matches the criteria (i.e.
25689 no partition changes for non-crossing jumps).
25690 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
25691 support for redirecting crossing jumps to non-crossing.
25693 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
25695 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
25696 also for naked functions.
25698 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
25700 * config/arc/arc.md (add_shift): New pattern.
25701 (add_shift2): Likewise.
25702 (sub_shift): Likewise.
25703 (sub_shift_cmp0_noout): Likewise.
25704 (compare_si_ashiftsi): Likewise.
25705 (xbfu_cmp0_noout): New combine pattern.
25706 (xbfu_cmp0"): Likewise.
25707 (movsi_set_cc_insn): Place the predicable variant first.
25708 (commutative_binary_cmp0_noout): Remove clobber.
25709 (commutative_binary_cmp0): New pattern.
25710 (noncommutative_binary_cmp0): Likewise.
25711 (noncommutative_binary_cmp0_noout): Likewise.
25712 (noncommutative_binary_comparison_result_used): Removed.
25713 (rsub_cmp0): New pattern.
25714 (rsub_cmp0_noout): Likewise.
25715 (extzvsi): Changed, keep only meaningful variants.
25716 (SQH, SEZ): New iterators.
25717 (SQH_postfix): New mode attribute.
25718 (SEZ_prefix): New code attribute.
25719 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
25720 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
25721 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
25722 of numerical value.
25723 (noncommutative_operator): Check the availability of barrel
25726 2018-04-09 Richard Biener <rguenther@suse.de>
25728 PR tree-optimization/85284
25729 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
25730 Only use the niter constraining form of simple_iv when the exit
25731 is always executed.
25733 2018-04-09 Tom de Vries <tom@codesourcery.com>
25736 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
25737 (define_expand "*memory_barrier"): New define_expand.
25738 (define_insn "memory_barrier"): New insn.
25740 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25742 PR rtl-optimization/80463
25743 PR rtl-optimization/83972
25744 PR rtl-optimization/83480
25746 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
25747 correct producer for the insn.
25748 (tidy_control_flow): Fixup seqnos in case of debug insns.
25750 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25752 PR rtl-optimization/83913
25754 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
25755 different sched-times when merging exprs.
25757 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25759 PR rtl-optimization/83962
25761 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
25762 tidy_fallthru_edge and tidy_control_flow.
25764 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25766 PR rtl-optimization/83530
25768 * sel-sched.c (force_next_insn): New global variable.
25769 (remove_insn_for_debug): When force_next_insn is true, also leave only
25770 next insn in the ready list.
25771 (sel_sched_region): When the region wasn't scheduled, make another pass
25772 over it with force_next_insn set to 1.
25774 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
25776 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
25778 * config/nds32/constants.md (unspec_volatile_element): Add enum values
25779 for interrupt control.
25780 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
25781 functions for interrupt control.
25782 * config/nds32/nds32-intrinsic.md: Likewise.
25783 * config/nds32/nds32_intrinsic.h: Likewise.
25784 * config/nds32/nds32.h (nds32_builtins): Likewise.
25786 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
25788 * config/nds32/nds32.c (nds32_init_machine_status,
25789 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
25790 strict_aligned_p field.
25791 (nds32_expand_to_rtl_hook): New function.
25792 (TARGET_EXPAND_TO_RTL_HOOK): Define.
25793 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
25795 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25796 Chung-Ju Wu <jasonwucj@gmail.com>
25798 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
25799 * config/nds32/nds32-n7.md: New file.
25800 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
25801 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
25803 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
25804 * config/nds32/nds32.md (pipeline_model): Add n7.
25805 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
25806 * config/nds32/pipelines.md: Include n7 settings.
25808 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25809 Chung-Ju Wu <jasonwucj@gmail.com>
25811 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
25812 * config/nds32/nds32-e8.md: New file.
25813 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
25814 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
25816 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
25817 * config/nds32/nds32.md (pipeline_model): Add e8.
25818 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
25819 * config/nds32/pipelines.md: Include e8 settings.
25821 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25822 Chung-Ju Wu <jasonwucj@gmail.com>
25824 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
25825 * config/nds32/nds32-n8.md: New file.
25826 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
25827 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
25829 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
25830 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
25831 * config/nds32/nds32.md (pipeline_model): Add n8.
25832 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
25833 * config/nds32/pipelines.md: Include n8 settings.
25835 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25836 Chung-Ju Wu <jasonwucj@gmail.com>
25838 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
25839 * config/nds32/nds32-n9-2r1w.md: New file.
25840 * config/nds32/nds32-n9-3r2w.md: New file.
25841 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
25842 nds32_register_ports): New or modify for cpu n9.
25843 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
25845 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
25846 * config/nds32/nds32-utils.c: New file.
25847 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
25848 TARGET_MUL_SLOW): Define.
25849 * config/nds32/nds32.md (pipeline_model): New attribute.
25850 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
25851 New options that support cpu n9.
25852 * config/nds32/pipelines.md: Include n9 settings.
25853 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
25855 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
25857 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
25858 information if necessary.
25859 (output_cond_branch_compare_zero): Likewise.
25860 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
25861 (nds32_target_alignment): Refine for alignment.
25862 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
25863 (FUNCTION_BOUNDARY): Modify.
25864 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
25866 * config/nds32/nds32.opt (malways-align, malign-functions): New.
25868 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
25870 * config/nds32/constants.md (unspec_volatile_element): Add values for
25871 TLB operation and data prefetch.
25872 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
25873 functions for TLB operation and data prefetch.
25874 * config/nds32/nds32-intrinsic.md: Likewise.
25875 * config/nds32/nds32_intrinsic.h: Likewise.
25876 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
25877 (nds32_print_operand): Likewise.
25878 * config/nds32/nds32.h (nds32_builtins): Likewise.
25880 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
25881 Andrew Pinski <pinsika@gcc.gnu.org>
25883 PR middle-end/82976
25884 * match.pd: Use constant_boolean_node of correct type instead of
25885 boolean_true_node or boolean_false_node for simplifying
25886 pointer comparisons to zero.
25888 2018-04-07 Jakub Jelinek <jakub@redhat.com>
25890 PR tree-optimization/80021
25891 * tree.c (verify_type_variant): Make error call in verify_variant_match
25892 translatable and remove final full stop.
25894 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
25896 * config/nds32/constants.md (unspec_volatile_element): Add
25897 UNSPEC_VOLATILE_EH_RETURN.
25898 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
25899 nds32_output_stack_pop): Support dwarf exception handling process.
25900 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
25901 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
25902 exception handling process.
25903 (nds32_compute_stack_frame): Likewise.
25904 (nds32_return_addr_rtx): Likewise.
25905 (nds32_initial_elimination_offset): Likewise.
25906 (nds32_expand_prologue): Likewise.
25907 (nds32_expand_epilogue): Likewise.
25908 (nds32_dynamic_chain_address): New function.
25909 * config/nds32/nds32.h (machine_function): Add fields for dwarf
25910 exception handling.
25911 (DYNAMIC_CHAIN_ADDRESS): Define.
25912 (EH_RETURN_DATA_REGNO): Define.
25913 (EH_RETURN_STACKADJ_RTX): Define.
25914 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
25915 patterns for dwarf exception handling.
25917 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
25919 * config/nds32/nds32.h: Clean up obsolete macros.
25921 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25923 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
25924 Add enum values for particular instructions.
25925 * config/nds32/nds32-intrinsic.c: Implementation of expanding
25926 particular intrinsic functions.
25927 * config/nds32/nds32-intrinsic.md: Likewise.
25928 * config/nds32/nds32_intrinsic.h: Likewise.
25929 * config/nds32/nds32.h (nds32_builtins): Likewise.
25930 * config/nds32/nds32.md (type): Add pbsad and pbsada.
25931 (btst, ave): New patterns for particular instructions.
25933 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25935 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
25936 Add enum values for atomic load/store and memory sync.
25937 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
25939 * config/nds32/nds32-intrinsic.md: Likewise.
25940 * config/nds32/nds32_intrinsic.h: Likewise.
25941 * config/nds32/nds32.h (nds32_builtins): Likewise.
25943 2018-04-07 Jakub Jelinek <jakub@redhat.com>
25945 PR tree-optimization/85257
25946 * fold-const.c (native_encode_vector): If not all elts could fit
25947 and off is -1, return 0 rather than offset.
25948 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
25949 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
25950 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
25951 adjust buffer in native_interpret_expr call.
25953 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25955 * config/nds32/constants.md (unspec_volatile_element): Add cache
25956 control enum values.
25957 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
25958 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
25959 * config/nds32/nds32.c (nds32_cctl_names): New.
25960 (nds32_print_operand): Handle cache control register names.
25961 * config/nds32/nds32.h (nds32_builtins): New enum values.
25962 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
25964 * config/nds32/nds32.md (type): Add mmu.
25965 * config/nds32/pipelines.md (simple_insn): Add mmu.
25967 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
25969 * config/nds32/nds32.md (type): Remove call.
25970 * config/nds32/pipelines.md (simple_insn): Likewise.
25972 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25974 * config/nds32/constants.md (unspec_volatile_element): Add
25975 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
25976 UNSPEC_VOLATILE_FMFCFG.
25977 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
25978 description for fmfcfg and fmfcsr.
25979 (bdesc_1arg): Add fmtcsr.
25980 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
25981 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
25982 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
25983 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
25984 unspec_fmfcfg): New patterns.
25985 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
25986 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
25987 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
25988 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
25989 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
25990 __nds32__fmfcfg): Define.
25992 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25994 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
25995 intrinsic register names.
25996 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
25997 intrinsic register enum values and macros.
25999 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
26001 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
26002 for load/store addressing form.
26003 (nds32_print_operand_address): Likewise.
26005 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
26008 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
26009 based on LABEL_REF. Remove useless assertion.
26010 (pic_address_needs_scratch): Fix formatting.
26011 (sparc_legitimize_pic_address): Minor tweaks.
26012 (sparc_delegitimize_address): Adjust assertion accordingly.
26013 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
26014 into symbolic_operand.
26015 (movsi_high_pic_label_ref): Likewise.
26016 (movsi_lo_sum_pic_label_ref): Likewise.
26017 (movdi_pic_label_ref): Likewise.
26018 (movdi_high_pic_label_ref): Likewise.
26019 (movdi_lo_sum_pic_label_ref): Likewise.
26021 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
26023 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
26024 custom LIB_SPEC setup.
26026 2018-04-06 Ruslan Bukin <br@bsdpad.com>
26027 Kito Cheng <kito.cheng@gmail.com>
26029 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
26030 * config/riscv/freebsd.h: New.
26032 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
26034 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
26035 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
26038 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
26039 Kito Cheng <kito.cheng@gmail.com>
26041 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
26042 nds32_output_call, nds32_symbol_binds_local_p): New functions.
26043 * config/nds32/nds32-protos.h (nds32_output_call,
26044 nds32_output_return): Declare.
26045 * config/nds32/nds32.md: Refine all the call and return patterns.
26047 2018-04-06 Jakub Jelinek <jakub@redhat.com>
26050 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
26051 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
26053 PR rtl-optimization/84872
26054 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
26055 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
26056 EDGE_CROSSING edge.
26058 2018-04-06 Tamar Christina <tamar.christina@arm.com>
26060 * expr.c (copy_blkmode_to_reg): Revert 254862.
26061 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
26063 2018-04-06 Richard Biener <rguenther@suse.de>
26065 PR middle-end/85244
26066 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
26067 after seeing a component reference with an adjacent field. Treat
26068 refs to arrays at struct end of external decls similar to
26069 refs to unconstrained commons.
26071 2018-04-06 Jakub Jelinek <jakub@redhat.com>
26074 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
26075 look through SAVE_EXPRs with non-side-effects argument. Adjust
26077 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
26080 2018-04-06 Richard Biener <rguenther@suse.de>
26082 PR middle-end/85180
26083 * alias.c (find_base_term): New wrapper around find_base_term
26084 unwinding CSELIB_VAL_PTR changes.
26085 (find_base_term): Do not restore CSELIB_VAL_PTR during the
26088 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
26090 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
26092 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
26093 constant definitions.
26094 ("nop"): lr 0,0 -> nopr r0
26095 ("nop_lr0", "nop_lr1"): New insn definitions.
26097 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
26099 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
26100 NDS32_V3PUSH_AVAILABLE_P macro.
26102 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
26103 Chung-Ju Wu <jasonwucj@gmail.com>
26105 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
26106 (nds32*-*-*): Add float and fpu_config into supported_defaults.
26107 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
26108 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
26109 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
26110 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
26111 * config/nds32/constraints.md: New constraints and checking for hard
26112 float configuration.
26113 * config/nds32/iterators.md: New mode iterator and attribute for hard
26114 float configuration.
26115 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
26117 * config/nds32/nds32-fpu.md: New file.
26118 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
26119 deal with hard float code generation.
26120 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
26122 (abi_type, float_reg_number): New enum type.
26123 * config/nds32/nds32-predicates.c: New predicates for hard float.
26124 * config/nds32/nds32-protos.h: Declare functions for hard float.
26125 * config/nds32/nds32.c: Implementation for hard float configuration.
26126 * config/nds32/nds32.h: Definitions for hard float configuration.
26127 * config/nds32/nds32.md: Include hard float machine description and
26128 modify patterns for hard float configuration.
26129 * config/nds32/nds32.opt: New options for hard float configuration.
26130 * config/nds32/predicates.md: New predicates for hard float
26133 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
26135 * common/config/nds32/nds32-common.c
26136 (nds32_option_optimization_table): Enable -mreleax-hint by default.
26138 2018-04-05 Jakub Jelinek <jakub@redhat.com>
26140 PR middle-end/85195
26141 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
26142 CONSTRUCTOR_ELT (ctor, ...)->value.
26144 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
26147 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
26149 2018-04-05 Tom de Vries <tom@codesourcery.com>
26152 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
26155 2018-04-05 Shiva Chen <shiva0217@gmail.com>
26156 Kito Cheng <kito.cheng@gmail.com>
26158 * config/nds32/constraints.md (U33): Fine-tune checking condition.
26159 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
26160 * config/nds32/nds32.h (nds32_16bit_address_type): Add
26161 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
26163 2018-04-05 Shiva Chen <shiva0217@gmail.com>
26164 Kito Cheng <kito.cheng@gmail.com>
26166 * config/nds32/constraints.md (Ufe): New memory constraint.
26167 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
26168 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
26169 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
26171 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
26172 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
26174 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26176 * config/nds32/nds32.md: Use optimize_size in the condition for
26177 alu-shift instructions.
26179 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26181 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
26183 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26185 * config/nds32/nds32.md (negsi2): Refine pattern.
26187 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
26188 Chung-Ju Wu <jasonwucj@gmail.com>
26190 * config/nds32/iterators.md (shift_rotate): New code iterator.
26191 (shift): New code attribute.
26192 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
26193 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
26194 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
26195 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
26196 bit-wise operations.
26197 (andsi3, *andsi3): Ditto.
26198 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
26199 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
26200 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
26201 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
26202 nds32_ior_operand, nds32_xor_operand): New predicates.
26204 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26206 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
26207 (addsi3, subsi3): ... this.
26209 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26211 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
26213 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26215 * config/nds32/nds32.md: Adjust indention.
26217 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
26219 * config/nds32/nds32.md (feature): New attribute.
26221 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26223 * config/nds32/nds32.md (subtype): New attribute.
26225 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
26228 * config/arm/arm-builtins.c (arm_expand_builtin): Change
26229 expansion to perform a bitwise AND of the argument followed by a
26230 boolean negation of the result.
26232 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
26234 PR rtl-optimization/84878
26235 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
26236 the basic block. Assert the use reference is not artificial and that
26237 it has an associated insn.
26239 2018-04-04 Michael Matz <matz@suse.de>
26241 * builtins.c (compute_objsize): Pass correct operand
26242 to array_at_struct_end_p.
26244 2018-04-04 Richard Biener <rguenther@suse.de>
26247 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
26248 from contexts for DINFO_LEVEL_TERSE and below.
26250 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
26252 * config/nds32/nds32-doubleword.md (move_<mode>): Require
26253 resiter_operand condition.
26254 * config/nds32/nds32.md (*move<mode>): Ditto.
26256 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
26257 Monk Chiang <sh.chiang04@gmail.com>
26259 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
26261 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26263 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
26265 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26266 Kito Cheng <kito.cheng@gmail.com>
26268 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
26269 nds32_cond_code_str, output_cond_branch,
26270 output_cond_branch_compare_zero, nds32_expand_cbranch,
26271 nds32_expand_cstore, nds32_expand_movcc,
26272 nds32_output_cbranchsi4_equality_zero,
26273 nds32_output_cbranchsi4_equality_reg,
26274 nds32_output_cbranchsi4_equality_reg_or_const_int,
26275 nds32_output_cbranchsi4_greater_less_zero: New functions.
26276 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
26277 nds32_expand_cstore, nds32_expand_movcc,
26278 nds32_output_cbranchsi4_equality_zero,
26279 nds32_output_cbranchsi4_equality_reg,
26280 nds32_output_cbranchsi4_equality_reg_or_const_int,
26281 nds32_output_cbranchsi4_greater_less_zero): Declare.
26282 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
26283 nds32_rimm11s_operand): New predicates.
26284 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
26285 * config/nds32/nds32.md: Rewrite all the branch and conditional move
26288 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
26290 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
26291 * config/nds32/nds32.md: Ditto.
26292 * config/nds32/pipelines.md: Ditto.
26294 2018-04-04 Richard Biener <rguenther@suse.de>
26296 PR tree-optimization/85168
26297 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
26298 propagating abnormals.
26300 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26302 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
26304 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26305 Kito Cheng <kito.cheng@gmail.com>
26307 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
26308 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
26309 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
26310 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
26311 * config/nds32/nds32.md (sibcall_internal): New.
26312 (sibcall_register): Remove.
26313 (sibcall_immediate): Remove.
26314 (sibcall_value_internal): New.
26315 (sibcall_value_register): Remove.
26316 (sibcall_value_immediate): Remove.
26317 * config/nds32/predicates.md (nds32_general_register_operand): New.
26318 (nds32_call_address_operand): New.
26320 2018-04-03 Jakub Jelinek <jakub@redhat.com>
26322 PR rtl-optimization/85167
26323 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
26324 bb_defs if *split_p, instead preinitialize it to NULL.
26326 PR tree-optimization/85156
26327 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
26328 evaluating the argument multiple times.
26330 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
26332 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
26334 (_mm_cvtpd_ps): Likewise.
26335 (_mm_cvttpd_epi32): Likewise.
26336 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
26337 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
26338 vector, pixel, and bool following altivec.h include.
26340 2018-04-03 Martin Sebor <msebor@redhat.com>
26342 * doc/extend.texi (Common Function Attributes): Clarify.
26343 (const attribute): Likewise.
26344 (pure attribute): Likewise.
26346 2018-04-03 Jakub Jelinek <jakub@redhat.com>
26349 * config/i386/i386.c (ix86_expand_vector_set): Use
26350 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
26352 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
26354 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
26355 instructions when changing rounding bits to preserve precision bits
26356 in the x87 control word.
26358 2018-04-03 Martin Liska <mliska@suse.cz>
26360 PR tree-optimization/82491
26361 * rtl.h (strip_offset_and_add): Replace += suboffset with
26362 poly_uint64 () + suboffset.
26364 2018-03-29 Martin Liska <mliska@suse.cz>
26365 Martin Jambor <mjambor@suse.cz>
26368 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
26369 param_type is not an integral or pointer type.
26371 2018-04-03 Richard Biener <rguenther@suse.de>
26373 * sese.h (recompute_all_dominators): Remove.
26375 2018-04-02 Martin Sebor <msebor@redhat.com>
26377 * doc/invoke.texi (-Wrestrict): Fix typos.
26379 2018-04-02 Jim Wilson <jimw@sifive.com>
26381 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
26382 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
26383 (<optab>di3, <optab>si3_extend): Likewise.
26384 (<optab>si3_mask, <optab>si3_mask_1): New.
26385 (<optab>di3_mask, <optab>di3_mask_1): New.
26386 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
26387 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
26388 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
26390 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
26392 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
26395 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
26397 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
26398 (nds32_canonicalize_comparison): New function.
26400 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
26401 Kito Cheng <kito.cheng@gmail.com>
26402 Kuan-Lin Chen <kuanlinchentw@gmail.com>
26404 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
26405 * config/nds32/constants.md (unspec_volatile_element): Add
26406 UNSPEC_VOLATILE_RELAX_GROUP.
26407 * config/nds32/nds32-relax-opt.c: New file.
26408 * config/nds32/nds32-predicates.c
26409 (nds32_symbol_load_store_p): New function.
26410 * config/nds32/nds32-protos.h
26411 (nds32_symbol_load_store_p): Declare function.
26412 (make_pass_nds32_relax_opt): Declare new rtl pass function.
26413 * config/nds32/nds32.c
26414 (nds32_register_pass): New function to register pass.
26415 (nds32_register_passes): New function to register passes.
26416 * config/nds32/nds32.md (relax_group): New pattern.
26417 * config/nds32/nds32.opt (mrelax-hint): New option.
26418 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
26420 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
26422 * config/nds32/t-nds32: Modify files dependency.
26424 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
26426 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
26427 (PROFILE_HOOK): Define its implementation.
26429 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
26431 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
26432 type and 32-bit size.
26434 2018-04-01 Jakub Jelinek <jakub@redhat.com>
26436 PR middle-end/85090
26437 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
26438 (V_128_256): New mode iterator.
26439 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
26440 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
26441 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
26443 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
26444 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
26446 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
26449 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
26450 NaN inputs correctly.
26452 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
26455 * config/rs6000/vsx.md (??r): New mode attribute.
26456 (*vsx_mov<mode>_64bit): Use it.
26457 (*vsx_mov<mode>_32bit): Likewise.
26459 2018-03-30 Martin Sebor <msebor@redhat.com>
26461 PR tree-optimization/84818
26462 * builtins.c (check_access): Use warning_n.
26464 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26467 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
26469 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
26472 2018-03-30 Julia Koval <julia.koval@intel.com>
26475 * x86-tune.def (movx, partial_reg_dependency): Enable for
26478 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
26480 PR inline-asm/84985
26481 * lra-constraints.c (process_alt_operands): Move setting
26482 this_alternative_matches below.
26484 2018-03-29 Martin Liska <mliska@suse.cz>
26487 * doc/invoke.texi: Document how LTO works with debug info.
26488 Describe auto-load support of binutils. Mention 'x86-64'
26489 as valid option value of -march option.
26491 2018-03-29 Jakub Jelinek <jakub@redhat.com>
26493 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
26496 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
26497 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
26498 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
26501 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
26504 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
26505 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
26506 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
26507 for RS6000_BTM_POWERPC64.
26508 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
26509 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
26510 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
26513 (DIVDEU): Likewise.
26515 2018-03-28 Carl Love <cel@us.ibm.com>
26518 2017-09-27 Carl Love <cel@us.ibm.com>
26520 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
26521 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
26522 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
26525 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26527 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
26528 instead of __vector bool.
26529 (_mm_max_pu8): Likewise.
26530 (_mm_min_pi16): Likewise.
26532 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
26535 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
26536 (DIVWEUO): Likewise.
26537 (DIVDEO): Likewise.
26538 (DIVDEUO): Likewise.
26539 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
26540 DIVWEUO and DIVDEUO.
26541 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
26542 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
26543 (div_extend): Likewise.
26544 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
26546 (__builtin_divweuo): Likewise.
26547 (__builtin_divdeo): Likewise.
26548 (__builtin_divdeuo): Likewise.
26550 2018-03-28 Jakub Jelinek <jakub@redhat.com>
26553 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
26554 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
26556 PR tree-optimization/82004
26557 * gimple-match-head.c (optimize_pow_to_exp): New function.
26558 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
26559 Don't fold to exp if optimize_pow_to_exp is false.
26561 2018-03-28 Martin Liska <mliska@suse.cz>
26564 * calls.c (initialize_argument_information): Fix trailing space.
26565 * common.opt: Fix typo and provide better explanation for
26566 -fsanitize-coverage option.
26567 * config/i386/i386.opt: Fix typo.
26569 2018-03-28 Jakub Jelinek <jakub@redhat.com>
26570 Martin Liska <mliska@suse.cz>
26573 * gimplify.c (asan_poison_variable): Don't do the check for
26574 gimplify_omp_ctxp here.
26575 (gimplify_decl_expr): Do it here.
26576 (gimplify_target_expr): Likewise.
26578 2018-03-28 Martin Liska <mliska@suse.cz>
26581 * config/i386/i386.c (ix86_function_arg_advance): Do not call
26582 chkp_type_bounds_count if MPX is not enabled.
26584 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
26586 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
26588 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
26591 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
26592 function to create the function decl for complex long double
26593 multiply and divide for -mabi=ieeelongdouble.
26594 (init_float128_ieee): Call it.
26596 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
26599 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
26600 -fcf-protection=branch -mibt.
26601 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
26603 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
26606 * config/arm/arm.c (arm_valid_symbolic_address): Handle
26607 arm_word_relocations.
26609 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
26612 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
26613 extern array declarations.
26615 2018-03-27 Richard Biener <rguenther@suse.de>
26617 PR middle-end/84067
26618 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
26619 explicit single_use checks.
26621 2018-03-27 Richard Biener <rguenther@suse.de>
26623 PR tree-optimization/85082
26624 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
26627 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
26629 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
26630 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
26631 Turn on fasynchronous-unwind-tables and funwind-tables.
26633 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
26636 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
26637 (*bmi_blsr_<mode>_ccz): Ditto.
26639 2018-03-26 Tom de Vries <tom@codesourcery.com>
26641 PR tree-optimization/85063
26642 * omp-general.c (offloading_function_p): New function. Factor out
26644 * omp-offload.c (pass_omp_target_link::gate): ... here.
26645 * omp-general.h (offloading_function_p): Declare.
26646 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
26647 with attribute omp declare target for offloading functions.
26649 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
26651 PR tree-optimization/84005
26652 * tree-data-ref.h (get_base_for_alignment): Declare.
26653 * tree-data-ref.c (get_base_for_alignment_1): New function.
26654 (get_base_for_alignment): Likewise.
26655 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26656 get_base_for_alignment to find a suitable base object, instead
26657 of always using drb->base_address.
26659 2018-03-23 Jakub Jelinek <jakub@redhat.com>
26661 PR inline-asm/85022
26662 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
26663 known size by default.
26665 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
26667 PR inline-asm/85030
26668 * lra-constraints.c (process_alt_operands): Don't match BLKmode
26669 and non BLKmode operands.
26671 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26674 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
26675 Clean up attributes.
26677 2018-03-23 Richard Biener <rguenther@suse.de>
26680 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
26681 we are going to emit early debug for LTO.
26683 2018-03-23 Jakub Jelinek <jakub@redhat.com>
26685 PR inline-asm/85034
26686 * function.c (match_asm_constraints_1): Don't optimize if input
26687 doesn't satisfy general_operand predicate for output's mode.
26689 PR inline-asm/85022
26690 * alias.c (write_dependence_p): Don't require for x_canonicalized
26691 non-VOIDmode if x has VOIDmode.
26694 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
26695 just don't try to optimize it rather than assert it never happens.
26697 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
26699 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
26700 macro expansions for definition of ST_INTERNAL_<mode> and
26701 LD_INTERNAL_<mode> builtins.
26702 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
26704 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
26706 (altivec_expand_st_builtin): Likewise.
26707 (altivec_expand_builtin): Remove calls to deleted functions.
26708 (rs6000_address_for_altivec): Delete this function.
26709 * config/rs6000/vector.md: Remove expands for
26710 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
26712 2018-03-22 Sudakshina Das <sudi.das@arm.com>
26715 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
26716 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
26717 re-computing once computed.
26718 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
26719 (arm_init_machine_status): Initialize
26720 machine->static_chain_stack_bytes.
26722 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
26725 * doc/extend.texi: Add four new prototypes for vec_ld.
26726 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
26727 definitions for more logical presentation.
26728 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
26729 entries for V1TI variants of __builtin_altivec_ld builtin.
26730 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
26731 handling of V1TI variant of LVX icode pattern.
26732 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
26733 (rs6000_gimple_fold_builtin): Likewise.
26734 (altivec_init_builtins): Add code to define
26735 __builtin_altivec_lvx_v1ti function.
26737 2018-03-22 Jakub Jelinek <jakub@redhat.com>
26739 PR inline-asm/84941
26740 * function.c (match_asm_constraints_1): Don't do the optimization
26741 if input isn't a REG, SUBREG, MEM or constant.
26743 2018-03-22 Tom de Vries <tom@codesourcery.com>
26745 PR tree-optimization/84956
26746 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
26747 bb_has_abnormal_pred.
26749 2018-03-22 Jakub Jelinek <jakub@redhat.com>
26752 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
26753 DECL_INITIAL (decl) to decl at the end.
26754 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
26755 adjust the comment.
26757 2018-03-21 Joseph Myers <joseph@codesourcery.com>
26759 * doc/extend.texi (__builtin_tgmath): Document when complex
26760 integer types are treated as _Complex _Float64.
26762 2018-03-21 Tom de Vries <tom@codesourcery.com>
26764 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
26766 2018-03-21 Jakub Jelinek <jakub@redhat.com>
26768 PR tree-optimization/84960
26769 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
26770 if it is ENTRY block, move them into single succ of ENTRY in that case.
26772 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
26774 PR tree-optimization/84811
26775 * poly-int.h (poly_span_traits): Remove the T3 parameter and
26776 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
26777 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
26778 (known_subrange_p): Update accordingly. Cast each value involved
26779 in the size comparison, rather than casting the result of the
26782 2018-03-21 Jakub Jelinek <jakub@redhat.com>
26784 PR tree-optimization/84982
26785 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
26786 by flipping the least significant bit rather than all bits from
26787 bitpos to bitpos + bitsize - 1.
26789 2018-03-21 Nathan Sidwell <nathan@acm.org>
26791 * doc/extend.texi (Deprecated Features): Remove mention of
26792 long-deleted deprecations.
26794 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26797 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
26798 * configure: Regenerate.
26800 2018-03-21 Tom de Vries <tom@codesourcery.com>
26802 PR tree-optimization/83126
26803 * tree-parloops.c (num_phis): New function.
26804 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
26806 2018-03-21 Nathan Sidwell <nathan@acm.org>
26808 * doc/extend.texi (Deprecated Features): Update deprecated flags,
26809 mention anon-struct/union members and trailing attributes.
26811 2018-03-21 Bin Cheng <bin.cheng@arm.com>
26813 PR tree-optimization/84969
26814 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
26815 builtin memset partitions if they set different rhs values.
26817 2018-03-21 Jakub Jelinek <jakub@redhat.com>
26819 PR rtl-optimization/84989
26820 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
26821 VEC_DUPLICATE with scalar result mode.
26823 2018-03-21 Martin Liska <mliska@suse.cz>
26826 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
26827 not intended return statement.
26829 2018-03-21 Martin Liska <mliska@suse.cz>
26832 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
26833 (chkp_find_bound_slots_1): Limit number of iterations.
26835 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
26838 * Minor grammar fixes for x86 options.
26840 2018-03-20 Jakub Jelinek <jakub@redhat.com>
26843 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
26844 holding REG_CFA_RESTORE notes, instead turn them into a USE.
26846 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
26849 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
26850 (altivec_lvx_<mode>_1op): Likewise.
26851 (altivec_stvx_<mode>_2op): Likewise.
26852 (altivec_stvx_<mode>_1op): Likewise.
26853 (altivec_lvx_<VM2:mode>): New define_expand.
26854 (altivec_stvx_<VM2:mode>): Likewise.
26855 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
26856 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
26857 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
26858 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
26859 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
26860 (rs6000_gen_lvx): Likewise.
26861 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
26862 (altivec_expand_stv_builtin): Likewise.
26863 (altivec_expand_builtin): Likewise.
26864 * config/rs6000/vector.md: Likewise.
26866 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26869 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
26872 2018-03-20 Richard Biener <rguenther@suse.de>
26875 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
26876 sign-conversions as zero, fall back to standard scalar_stmt
26879 2018-03-20 Martin Liska <mliska@suse.cz>
26882 * predict.c (rebuild_frequencies): Handle case when we have
26883 PROFILE_ABSENT, but flag_guess_branch_prob is false.
26885 2018-03-20 Jakub Jelinek <jakub@redhat.com>
26888 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
26889 flag_section_anchors.
26890 * varasm.c (use_blocks_for_decl_p): Remove hack for
26891 dw2_force_const_mem.
26894 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
26896 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
26897 be created, use lowpart_subreg of operands[0] rather than operands[0]
26899 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
26900 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
26901 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
26902 and n constraint instead of aarch64_shift_imm_di and Usd.
26903 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
26904 (*aarch64_<optab>_reg_minus<mode>3): ... this.
26906 2018-03-20 Sudakshina Das <sudi.das@arm.com>
26909 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
26910 to favor GPR over NEON registers.
26911 (<shift>di3_neon): Likewise.
26913 2018-03-20 Tom de Vries <tom@codesourcery.com>
26916 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
26917 (nvptx_process_pars): Emit bar.sync asap and alap.
26919 2018-03-20 Tom de Vries <tom@codesourcery.com>
26922 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
26923 seen_label if seen_label is already set.
26925 2018-03-20 Jakub Jelinek <jakub@redhat.com>
26928 * config/i386/i386.c (fold_builtin_cpu): For features above 31
26929 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
26930 Use 1U instead of 1. Formatting fixes.
26933 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
26934 instead of TREE_TYPE (s1) for the return value.
26936 2018-03-19 Jakub Jelinek <jakub@redhat.com>
26938 PR tree-optimization/84946
26939 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
26940 bitsize + bitsize in poly_uint64 rather than poly_int64.
26943 * dwarf2asm.c: Include fold-const.c.
26944 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
26945 of decl rather than decl itself.
26947 PR rtl-optimization/84643
26948 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
26950 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
26953 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
26954 calling assemble_variable.
26956 2018-03-19 Sudakshina Das <sudi.das@arm.com>
26959 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
26960 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
26962 2018-03-19 Jim Wilson <jimw@sifive.com>
26965 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
26966 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
26967 (riscv_first_stack_step): Likewise.
26968 (riscv_option_override): Use STACK_BOUNDARY instead of
26969 MIN_STACK_BOUNDARY.
26970 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
26971 MIN_STACK_BOUNDARY.
26972 (BIGGEST_ALIGNMENT): Set to 128.
26973 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
26974 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
26977 2018-03-19 Richard Biener <rguenther@suse.de>
26979 PR tree-optimization/84933
26980 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
26981 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
26983 2018-03-19 Richard Biener <rguenther@suse.de>
26985 PR tree-optimization/84859
26986 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
26987 (cond_if_else_store_replacement): Perform sinking operation on
26988 single-store BBs regardless of MAX_STORES_TO_SINK setting.
26989 Generalize what a BB with a single eligible store is.
26991 2018-03-19 Richard Biener <rguenther@suse.de>
26993 PR tree-optimization/84929
26994 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
26995 chrec_is_positive against non-chrec arg.
26997 2018-03-19 Tamar Christina <tamar.christina@arm.com>
27000 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
27002 2018-03-18 Martin Liska <mliska@suse.cz>
27004 PR rtl-optimization/84635
27005 * regrename.c (build_def_use): Use matches_mode only when
27008 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
27010 PR tree-optimization/84913
27011 * tree-vect-loop.c (vectorizable_reduction): Don't try to
27012 vectorize chains of COND_EXPRs.
27014 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
27016 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
27018 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
27020 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
27022 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
27024 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
27026 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
27027 Kito Cheng <kito.cheng@gmail.com>
27029 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
27030 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
27031 (nds32_adjust_reg_alloc_order): New function.
27032 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
27034 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
27036 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
27037 nds32_print_operand, nds32_print_operand_address): Use
27038 HOST_WIDE_INT_PRINT_DEC instead.
27040 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
27042 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
27044 2018-03-17 Jakub Jelinek <jakub@redhat.com>
27047 * config/i386/i386.c (initial_ix86_tune_features,
27048 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
27049 unsigned long long.
27050 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
27051 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
27052 rather than 1u << ix86_tune. Formatting fix.
27053 (ix86_option_override_internal): Change ix86_arch_mask from
27054 unsigned int to unsigned HOST_WIDE_INT, initialize to
27055 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
27056 (ix86_function_specific_restore): Likewise.
27058 2018-03-16 Jakub Jelinek <jakub@redhat.com>
27061 * postreload.c (reload_combine_recognize_pattern): Perform
27062 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
27063 truncate_int_for_mode the result for the destination's mode.
27066 * hsa-gen.c (mem_type_for_type): Fix comment typo.
27067 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
27069 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
27072 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
27075 * lra-assigns.c (lra_split_hard_reg_for): Don't use
27076 regno_allocno_class_array and sorted_pseudos.
27077 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
27078 insns where regno is used.
27080 2018-03-16 Martin Liska <mliska@suse.cz>
27083 * multiple_target.c (create_dispatcher_calls): Redirect
27084 reference in the symbol table.
27086 2018-03-16 Martin Liska <mliska@suse.cz>
27089 * multiple_target.c (create_dispatcher_calls): Redirect also
27092 2018-03-16 Jakub Jelinek <jakub@redhat.com>
27096 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
27097 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
27098 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
27100 2018-03-16 Julia Koval <julia.koval@intel.com>
27102 * doc/invoke.texi (Skylake Server): Add CLWB.
27103 Cannonlake): Remove CLWB.
27105 2018-03-16 Jakub Jelinek <jakub@redhat.com>
27107 PR tree-optimization/84841
27108 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
27110 (FLOAT_ONE_CONST_TYPE): Define.
27111 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
27112 (sort_by_operand_rank): Put entries with higher constant_type last
27113 rather than first to match comments.
27115 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
27117 * config/nios2/nios2.md (movsi_internal): Fix thinko in
27120 2018-03-15 Jakub Jelinek <jakub@redhat.com>
27123 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
27124 check and use address of target always.
27126 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
27129 * config/i386/i386.c (indirect_thunk_needed): Update comments.
27130 (indirect_thunk_bnd_needed): Likewise.
27131 (indirect_thunks_used): Likewise.
27132 (indirect_thunks_bnd_used): Likewise.
27133 (indirect_return_needed): New.
27134 (indirect_return_bnd_needed): Likewise.
27135 (output_indirect_thunk_function): Add a bool argument for
27137 (output_indirect_thunk_function): Don't generate alias for
27138 function return thunk.
27139 (ix86_code_end): Call output_indirect_thunk_function to generate
27140 function return thunks.
27141 (ix86_output_function_return): Set indirect_return_bnd_needed
27142 and indirect_return_needed instead of indirect_thunk_bnd_needed
27143 and indirect_thunk_needed.
27145 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
27147 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
27148 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
27149 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
27151 2018-03-15 David Malcolm <dmalcolm@redhat.com>
27152 Paul Hua <paul.hua.gm@gmail.com>
27155 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
27157 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
27159 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
27160 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
27161 resp. SFmode cases.
27163 2018-03-15 Tamar Christina <tamar.christina@arm.com>
27166 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
27167 instead of GET_MODE_SIZE when comparing Units.
27169 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
27172 * varasm.c (hash_section): Return an unchangeble hash value
27173 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
27174 Return !aarch64_can_use_per_function_literal_pools_p ().
27176 2018-03-15 Jakub Jelinek <jakub@redhat.com>
27179 * optabs.c (emit_conditional_move): Pass address of cmode's copy
27180 rather than address of cmode as last argument to prepare_cmp_insn.
27182 2018-03-15 Julia Koval <julia.koval@intel.com>
27184 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
27185 F_AVX512VNNI, F_AVX512BITALG): New.
27187 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
27190 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
27191 insn for floating-point loads and stores.
27193 2018-03-14 Carl Love <cel@us.ibm.com>
27195 * config/rs6000/rs6000-c.c: Add macro definitions for
27196 ALTIVEC_BUILTIN_VEC_PERMXOR.
27197 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
27198 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
27199 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
27200 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
27202 * config/doc/extend.texi: Add prototypes for vec_permxor.
27204 2018-03-14 David Malcolm <dmalcolm@redhat.com>
27207 * diagnostic-show-locus.c (class layout_point): Convert m_line
27208 from int to linenum_type.
27209 (line_span::comparator): Use linenum "compare" function when
27210 comparing line numbers.
27211 (test_line_span): New function.
27212 (layout_range::contains_point): Convert param "row" from int to
27214 (layout_range::intersects_line_p): Likewise.
27215 (layout::will_show_line_p): Likewise.
27216 (layout::print_source_line): Likewise.
27217 (layout::should_print_annotation_line_p): Likewise.
27218 (layout::print_annotation_line): Likewise.
27219 (layout::print_leading_fixits): Likewise.
27220 (layout::annotation_line_showed_range_p): Likewise.
27221 (struct line_corrections): Likewise for field m_row.
27222 (line_corrections::line_corrections): Likewise for param "row".
27223 (layout::print_trailing_fixits): Likewise.
27224 (layout::get_state_at_point): Likewise.
27225 (layout::get_x_bound_for_row): Likewise.
27226 (layout::print_line): Likewise.
27227 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
27228 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
27229 * input.c (selftest::test_linenum_comparisons): New function.
27230 (selftest::input_c_tests): Call it.
27231 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
27232 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
27233 * selftest.h (ASSERT_GT): New macro.
27234 (ASSERT_GT_AT): New macro.
27235 (ASSERT_LT): New macro.
27236 (ASSERT_LT_AT): New macro.
27238 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
27240 PR rtl-optimization/84780
27241 * combine.c (distribute_links): Don't make a link based on pc_rtx.
27243 2018-03-14 Martin Liska <mliska@suse.cz>
27245 * tree.c (record_node_allocation_statistics): Use
27246 get_stats_node_kind.
27247 (get_stats_node_kind): New function extracted from
27248 record_node_allocation_statistics.
27249 (free_node): Use get_stats_node_kind.
27251 2018-03-14 Richard Biener <rguenther@suse.de>
27253 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
27254 that the value-set of ANTIC_IN doesn't grow.
27257 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
27259 (BB_VISITED_WITH_VISITED_SUCCS): New define.
27260 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
27262 2018-03-14 Julia Koval <julia.koval@intel.com>
27264 * config.gcc (icelake-client, icelake-server): New.
27266 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
27267 (initial_ix86_arch_features): Ditto.
27268 (PTA_SKYLAKE): Add SGX.
27269 (PTA_ICELAKE): Remove.
27270 (PTA_ICELAKE_CLIENT): New.
27271 (PTA_ICELAKE_SERVER): New.
27272 (ix86_option_override_internal): Split up icelake on icelake client and
27274 (get_builtin_code_for_version): Ditto.
27275 (fold_builtin_cpu): Ditto.
27276 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
27277 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
27278 * config/i386/i386.h (processor_type): Ditto.
27279 * doc/invoke.texi: Ditto.
27281 2018-03-14 Jakub Jelinek <jakub@redhat.com>
27284 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
27285 INTEGER_CST offset, add it together with bitpos / 8 and
27286 sign extend based on POINTER_SIZE.
27290 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
27293 * config/i386/constraints.md (Yc): New register constraint.
27294 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
27295 Use Yc constraint for alternative 2 of operand 0. Remove
27296 preferred_for_speed attribute.
27298 2018-03-14 Richard Biener <rguenther@suse.de>
27300 PR tree-optimization/84830
27301 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
27302 with the old one to avoid oscillations.
27304 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
27307 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
27309 (assign_by_spills): Return a flag of reload assignment failure.
27310 Do not process the reload assignment failures. Do not spill other
27311 reload pseudos if they has the same reg class. Update n if
27313 (lra_assign): Add a return arg. Set up from the result of
27314 assign_by_spills call.
27315 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27316 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
27317 usage_insns if it is not NULL.
27318 (spill_hard_reg_in_range): New function.
27319 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27320 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27321 function prototypes.
27322 (lra_assign): Change prototype.
27323 * lra.c (lra): Add code to deal with fails by splitting hard reg
27326 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
27328 * config/riscv/riscv.opt (mrelax): New option.
27329 * config/riscv/riscv.c (riscv_file_start): Emit ".option
27330 "norelax" when riscv_mrelax is disabled.
27331 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
27333 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27336 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
27337 reassociation for int modes.
27339 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
27341 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
27342 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
27344 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
27345 * config/aarch64/aarch64-sve.md
27346 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
27347 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
27348 (*extend<mode><Vwide>2): Rename to...
27349 (aarch64_sve_extend<mode><Vwide>2): ...this.
27350 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
27351 renaming the old pattern to...
27352 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
27354 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
27355 define_expand, renaming the old pattern to...
27356 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
27357 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
27358 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
27359 account when deciding which SVE instruction the optab should use.
27360 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
27362 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
27364 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
27365 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
27366 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
27367 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
27368 (tlsdesc_small_<mode>): Turn a define_expand and use
27369 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
27370 (tlsdesc_small_advsimd_<mode>): ...this.
27371 (tlsdesc_small_sve_<mode>): New pattern.
27373 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
27375 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
27376 (UNSPEC_UMUL_HIGHPART): New constants.
27377 (MUL_HIGHPART): New int iteraor.
27378 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
27379 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
27381 (*<su>mul<mode>3_highpart): New define_insn.
27383 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
27386 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
27389 2018-03-13 Martin Liska <mliska@suse.cz>
27392 * (sem_item_optimizer::sem_item_optimizer): Initialize new
27394 (sem_item_optimizer::~sem_item_optimizer): Release it.
27395 (sem_item_optimizer::merge_classes): Register variable aliases.
27396 (sem_item_optimizer::fixup_pt_set): New function.
27397 (sem_item_optimizer::fixup_points_to_sets): Likewise.
27398 * ipa-icf.h: Declare new variables and functions.
27400 2018-03-13 Jakub Jelinek <jakub@redhat.com>
27402 PR middle-end/84834
27403 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
27404 integer_pow2p@2 and test integer_pow2p in condition.
27405 (A < 0 ? C : 0): Similarly for @1.
27407 PR middle-end/84831
27408 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
27409 characters starting at p contain '\0' character, don't look beyond
27413 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
27414 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
27417 * reg-stack.c (change_stack): Change update_end var from int to
27418 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
27419 also call set_block_for_insn on the newly added insns and rescan.
27422 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
27423 on the last operand.
27426 * tree.c (stabilize_reference_1): Return save_expr (e) for
27427 STATEMENT_LIST even if it doesn't have side-effects.
27429 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
27431 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
27433 2018-03-12 Renlin Li <renlin.li@arm.com>
27435 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
27436 aarch64_output_scalar_simd_mov_immediate.
27438 2018-03-12 Martin Sebor <msebor@redhat.com>
27440 PR tree-optimization/83456
27441 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
27442 for perfectly overlapping calls to memcpy.
27443 (gimple_fold_builtin_memory_chk): Same.
27444 (gimple_fold_builtin_strcpy): Handle no-warning.
27445 (gimple_fold_builtin_stxcpy_chk): Same.
27446 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
27448 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
27450 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
27451 parameter. Use it for SFmode.
27452 (rs6000_function_arg_advance_1): Adjust.
27453 (rs6000_function_arg): Adjust.
27454 (rs6000_gimplify_va_arg): Pass false for that new parameter.
27456 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
27458 PR rtl-optimization/84169
27459 PR rtl-optimization/84780
27460 * combine.c (can_combine_p): Check for a 2-insn combination whether
27461 the destination register is used between the two insns, too.
27463 2018-03-12 Richard Biener <rguenther@suse.de>
27465 PR tree-optimization/84803
27466 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
27467 for refs DR analysis didn't process.
27469 2018-03-12 Richard Biener <rguenther@suse.de>
27471 PR tree-optimization/84777
27472 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
27473 force-vectorize loops ignore whether we are optimizing for size.
27475 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
27477 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
27478 (TARGET_MD_ASM_ADJUST): Define.
27480 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
27481 Kito Cheng <kito.cheng@gmail.com>
27482 Chung-Ju Wu <jasonwucj@gmail.com>
27484 * config/nds32/nds32.c (nds32_compute_stack_frame,
27485 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
27486 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
27487 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
27488 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
27489 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
27490 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
27491 * config/nds32/nds32.md (prologue, epilogue): Use macro
27492 NDS32_V3PUSH_AVAILABLE_P to do checking.
27494 2018-03-11 Jakub Jelinek <jakub@redhat.com>
27497 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
27498 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
27499 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
27500 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
27501 addition of most attributes on !orig_type_die or the attribute not
27502 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
27504 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
27505 Chung-Ju Wu <jasonwucj@gmail.com>
27507 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
27508 __NDS32_VH__ macro.
27509 * config/nds32/nds32.opt (mvh): New option.
27511 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
27512 Chung-Ju Wu <jasonwucj@gmail.com>
27514 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
27516 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
27517 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
27520 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
27521 Chung-Ju Wu <jasonwucj@gmail.com>
27523 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
27525 * config/nds32/nds32-multiple.md (strlensi): New pattern.
27526 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
27528 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
27529 Kito Cheng <kito.cheng@gmail.com>
27530 Chung-Ju Wu <jasonwucj@gmail.com>
27532 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
27533 UNSPEC_FFMISM and UNSPEC_FLMISM.
27534 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
27535 for ffb, ffmism and flmism.
27536 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
27537 (unspec_ffmism): Ditto.
27538 (unspec_flmism): Ditto.
27539 (nds32_expand_builtin_impl): Check if string extension is available.
27540 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
27541 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
27543 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
27546 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
27549 * lra-assigns.c (assign_by_spills): Return a flag of reload
27550 assignment failure. Do not process the reload assignment
27551 failures. Do not spill other reload pseudos if they has the same
27553 (lra_assign): Add a return arg. Set up from the result of
27554 assign_by_spills call.
27555 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27556 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
27557 usage_insns if it is not NULL.
27558 (spill_hard_reg_in_range): New function.
27559 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27560 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27561 function prototypes.
27562 (lra_assign): Change prototype.
27563 * lra.c (lra): Add code to deal with fails by splitting hard reg
27566 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
27569 * config/i386/i386.opt: Replace Enforcment with Enforcement.
27571 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
27574 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
27575 (dw_val_node): Add val_symbolic_view.
27576 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
27577 (symview_upper_bound): New.
27578 (new_line_info_table): Initialize symviews_since_reset.
27579 (dwarf2out_source_line): Count symviews_since_reset and set
27580 symview_upper_bound.
27581 (dw_val_equal_p): Handle symview.
27582 (add_AT_symview): New.
27583 (print_dw_val): Handle symview.
27584 (attr_checksum, attr_checksum_ordered): Likewise.
27585 (same_dw_val_p, size_of_die): Likewise.
27586 (value_format, output_die): Likewise.
27587 (add_high_low_attributes): Use add_AT_symview for entry_view.
27588 (dwarf2out_finish): Reset symview_upper_bound, clear
27591 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
27594 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
27595 Add strict argument and use it.
27596 (rs6000_split_multireg_move): Update for new strict argument.
27597 (mem_operand_gpr): Disallow all non-offsettable addresses.
27598 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
27600 2018-03-09 Jakub Jelinek <jakub@redhat.com>
27603 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
27604 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
27605 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
27608 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
27609 decl, use remap_type if we want to use the type.
27611 2018-03-09 Martin Sebor <msebor@redhat.com>
27613 PR tree-optimization/84526
27614 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
27616 (builtin_access::generic_overlap): Be prepared to handle non-array
27619 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
27621 PR rtl-optimization/84682
27622 * lra-constraints.c (process_address_1): Check is_address flag
27623 for address constraints.
27624 (process_alt_operands): Likewise.
27625 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
27626 preprocess_constraints.
27627 * recog.h (preprocess_constraints): Add oploc parameter.
27629 * recog.c (preprocess_constraints): Test address_operand for
27630 CT_ADDRESS constraints.
27632 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
27635 * lra-assigns.c (assign_by_spills): Return a flag of reload
27636 assignment failure. Do not process the reload assignment
27637 failures. Do not spill other reload pseudos if they has the same
27639 (lra_assign): Add a return arg. Set up from the result of
27640 assign_by_spills call.
27641 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27642 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
27643 usage_insns if it is not NULL.
27644 (spill_hard_reg_in_range): New function.
27645 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27646 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27647 function prototypes.
27648 (lra_assign): Change prototype.
27649 * lra.c (lra): Add code to deal with fails by splitting hard reg
27652 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27655 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
27656 Accept complain bool parameter. Only emit errors if it is true.
27657 (arm_parse_cpu_option_name): Likewise.
27658 (arm_target_thumb_only): Adjust callers of the above.
27659 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
27660 prototype to take a default true bool parameter.
27661 (arm_parse_arch_option_name): Likewise.
27663 2018-03-09 David Malcolm <dmalcolm@redhat.com>
27664 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
27668 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
27669 * configure: Regenerate.
27670 * configure.ac ("linker --version-script option"): New.
27671 ("linker soname option"): New.
27673 2018-03-09 Richard Biener <rguenther@suse.de>
27675 PR tree-optimization/84775
27676 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
27677 immediate uses of predicate stmts and mark them modified.
27680 PR tree-optimization/84178
27681 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27683 (version_loop_for_if_conversion): Delay update_ssa call.
27684 (tree_if_conversion): Delay update_ssa until after predicate
27687 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
27690 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
27691 when the function accesses prior frames.
27693 2018-03-08 Jakub Jelinek <jakub@redhat.com>
27696 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
27697 gen_llsym, otherwise call maybe_gen_llsym.
27699 PR inline-asm/84742
27700 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
27701 has ',' character inside of it.
27703 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27706 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
27707 as clobbering CC_REGNUM.
27709 2018-03-08 Richard Biener <rguenther@suse.de>
27711 PR middle-end/84552
27712 * tree-scalar-evolution.c: Include tree-into-ssa.h.
27713 (follow_copies_to_constant): Do not follow SSA names registered
27716 2018-03-08 Richard Biener <rguenther@suse.de>
27718 PR tree-optimization/84178
27719 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27721 (version_loop_for_if_conversion): Delay update_ssa call.
27722 (tree_if_conversion): Delay update_ssa until after predicate
27725 2018-03-08 David Malcolm <dmalcolm@redhat.com>
27727 PR tree-optimization/84178
27728 * tree-if-conv.c (release_bb_predicate): Remove the
27729 the assertion that the stmts have NULL use_ops.
27730 Discard the statements, asserting that they haven't
27731 yet been added to a BB.
27733 2018-03-08 Richard Biener <rguenther@suse.de>
27735 PR tree-optimization/84746
27736 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
27737 (phi_translate): Pass in destination ANTIC_OUT set.
27738 (phi_translate_1): Likewise. For a simplified result lookup
27739 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
27740 (phi_translate_set): Adjust.
27741 (do_pre_regular_insertion): Likewise.
27742 (do_pre_partial_partial_insertion): Likewise.
27744 2018-03-08 Martin Liska <mliska@suse.cz>
27746 PR gcov-profile/84735
27747 * doc/gcov.texi: Document usage of profile files.
27748 * gcov-io.h: Document changes in the format.
27750 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
27754 * dwarf2out.c (struct dw_line_info_table): Update comments for
27756 (FORCE_RESET_NEXT_VIEW): New.
27757 (FORCE_RESETTING_VIEW_P): New.
27758 (RESETTING_VIEW_P): Check for -1 too.
27759 (ZERO_VIEW_P): Likewise.
27760 (new_line_info_table): Force-reset next view.
27761 (dwarf2out_begin_function): Likewise.
27762 (dwarf2out_source_line): Simplify zero_view_p initialization.
27763 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
27764 view directly. Omit view when omitting .loc at line 0.
27766 2018-03-08 Jakub Jelinek <jakub@redhat.com>
27768 PR tree-optimization/84740
27769 * tree-switch-conversion.c (process_switch): Call build_constructors
27770 only if info.phi_count is non-zero.
27772 PR tree-optimization/84739
27773 * tree-tailcall.c (find_tail_calls): Check call arguments against
27774 DECL_ARGUMENTS (current_function_decl) rather than
27775 DECL_ARGUMENTS (func) when checking for tail recursion.
27777 2018-03-07 Jakub Jelinek <jakub@redhat.com>
27779 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
27780 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
27781 Volker Reichelt's entry and add entries for people that perform
27782 GCC fuzzy testing and report numerous bugs.
27784 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
27787 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
27788 readonly data in sdata, if that is disabled.
27789 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
27790 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
27791 -mreadonly-in-sdata option.
27793 2018-03-07 Martin Sebor <msebor@redhat.com>
27795 PR tree-optimization/84468
27796 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
27797 basic block when looking for nul assignment.
27799 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
27802 * except.h (output_function_exception_table): Adjust prototype.
27803 * except.c (output_function_exception_table): Remove FNNAME parameter
27804 and add SECTION parameter. Ouput one part of the table at a time.
27805 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
27806 the first part of the exception table and emit unwind directives.
27807 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
27808 (i386_pe_seh_cold_init): Likewise.
27809 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
27810 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
27811 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
27812 (ix86_output_call_insn): Emit a nop in one more case for SEH.
27813 * config/i386/winnt.c: Include except.h.
27814 (struct seh_frame_state): Add reg_offset, after_prologue and
27815 in_cold_section fields.
27816 (i386_pe_seh_end_prologue): Set seh->after_prologue.
27817 (i386_pe_seh_cold_init): New function.
27818 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
27819 to seh->in_cold_section.
27820 (seh_emit_push): Record the offset of the push.
27821 (seh_emit_save): Record the offet of the save.
27822 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
27823 Test seh->after_prologue to disregard the epilogue.
27824 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
27825 (i386_pe_end_cold_function): New function.
27827 2018-03-07 Jakub Jelinek <jakub@redhat.com>
27830 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
27831 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
27834 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
27836 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
27837 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
27839 PR middle-end/84723
27840 * multiple_target.c: Include tree-inline.h and intl.h.
27841 (expand_target_clones): Diagnose and fail if node->definition and
27842 !tree_versionable_function_p (node->decl).
27844 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
27846 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
27848 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
27849 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
27850 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
27852 2018-03-06 Jakub Jelinek <jakub@redhat.com>
27855 * combine.c (try_combine): Use reg_or_subregno instead of handling
27856 just paradoxical SUBREGs and REGs.
27858 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
27860 * config/arc/arc.c (arc_finalize_pic): Remove function.
27861 (arc_must_save_register): We use single base PIC register, remove
27862 checks to save/restore the PIC register.
27863 (arc_expand_prologue): Likewise.
27864 * config/arc/arc-protos.h (arc_set_default_type_attributes):
27866 (arc_verify_short): Likewise.
27867 (arc_attr_type): Likewise.
27868 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
27869 (walk_stores): Likewise.
27870 (arc_address_cost): Make it static.
27871 (arc_verify_short): Likewise.
27872 (branch_dest): Likewise.
27873 (arc_attr_type): Likewise.
27874 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
27875 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
27876 (arc_final_prescan_insn): Remove inserting the nops due to
27877 hardware hazards. It is done in reorg step.
27878 (insn_length_variant_t): Remove.
27879 (insn_length_parameters_t): Likewise.
27880 (arc_insn_length_parameters): Likewise.
27881 (arc_get_insn_variants): Likewise.
27882 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
27884 2018-03-06 Jakub Jelinek <jakub@redhat.com>
27886 PR inline-asm/84683
27887 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
27890 PR tree-optimization/84687
27891 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
27893 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
27895 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27897 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
27898 Rename to ppc_speculation_barrier.
27899 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
27900 __builtin_ppc_speculation_barrier.
27902 2018-03-05 Jakub Jelinek <jakub@redhat.com>
27905 * combine.c (combine_simplify_rtx): Don't try to simplify if
27906 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
27909 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
27911 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
27912 to 32 bytes when compiling for POWER9.
27914 2018-03-05 Jakub Jelinek <jakub@redhat.com>
27917 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
27918 regparm >= 3 with no arg reg available also for calls with
27919 flag_force_indirect_call. Pass decl to ix86_function_regparm.
27922 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
27924 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
27926 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
27929 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
27931 2018-03-05 Richard Biener <rguenther@suse.de>
27933 PR tree-optimization/84486
27934 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
27935 When inserting a __builtin_assume_aligned call set the LHS
27936 SSA name alignment info accordingly.
27938 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
27940 PR tree-optimization/84114
27941 * config/aarch64/aarch64.c (aarch64_reassociation_width)
27942 Avoid reassociation of FLOAT_MODE addition.
27944 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
27946 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
27947 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
27948 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
27949 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
27950 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
27951 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
27952 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
27954 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
27955 __builtin_ia32_wbinvd): New builtins.
27956 (SPECIAL_ARGS2): New.
27957 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
27958 (SPECIAL_ARGS2): New.
27959 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
27960 (ix86_valid_target_attribute_inner_p): Ditto.
27961 (ix86_init_mmx_sse_builtins): Add special_args2.
27962 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
27963 TARGET_WBNOINVD_P): New.
27964 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
27965 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
27966 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
27967 * config/i386/immintrin.h (_wbinvd): New intrinsic.
27968 * config/i386/pconfigintrin.h: New file.
27969 * config/i386/wbnoinvdintrin.h: Ditto.
27970 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
27972 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
27974 2018-03-05 Richard Biener <rguenther@suse.de>
27976 PR tree-optimization/84670
27977 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
27979 (BB_VISITED_WITH_VISITED_SUCCS): New define.
27980 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
27981 (compute_antic_aux): Only assert the number of values in ANTIC_IN
27982 doesn't grow if all successors (recursively) were visited at least
27985 2018-03-05 Richard Biener <rguenther@suse.de>
27987 PR tree-optimization/84650
27988 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
27989 if executed in the loop pipeline.
27991 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
27993 * doc/configfiles.texi (Configuration Files): Move info about
27994 conditionalizing $target-protos.h to...
27995 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
27996 differs from $target-protos.h.
27998 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
27999 Chung-Ju Wu <jasonwucj@gmail.com>
28001 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
28002 * config/nds32/nds32-multiple.md (setmemsi): Define.
28003 * config/nds32/nds32-memory-manipulation.c
28004 (nds32_gen_dup_4_byte_to_word_value): New.
28005 (emit_setmem_word_loop): New.
28006 (emit_setmem_byte_loop): New.
28007 (nds32_expand_setmem_loop): New.
28008 (nds32_expand_setmem_loop_v3m): New.
28009 (nds32_expand_setmem_unroll): New.
28010 (nds32_expand_setmem): New.
28012 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
28013 Chung-Ju Wu <jasonwucj@gmail.com>
28015 * config/nds32/nds32-memory-manipulation.c
28016 (nds32_emit_load_store): New.
28017 (nds32_emit_post_inc_load_store): New.
28018 (nds32_emit_mem_move): New.
28019 (nds32_emit_mem_move_block): New.
28020 (nds32_expand_movmemsi_loop_unknown_size): New.
28021 (nds32_expand_movmemsi_loop_known_size): New.
28022 (nds32_expand_movmemsi_loop): New.
28023 (nds32_expand_movmemsi_unroll): New.
28024 (nds32_expand_movmemqi): Rename ...
28025 (nds32_expand_movmemsi): ... to this.
28026 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
28027 (movmemsi): ... to this.
28028 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
28029 (nds32_expand_movmemsi): ... to this.
28031 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
28032 Monk Chiang <sh.chiang04@gmail.com>
28033 Chung-Ju Wu <jasonwucj@gmail.com>
28035 * config/nds32/nds32-protos.h
28036 (nds32_expand_load_multiple): New arguments.
28037 (nds32_expand_store_multiple): Ditto.
28038 (nds32_valid_multiple_load_store): Rename ...
28039 (nds32_valid_multiple_load_store_p): ... to this.
28040 * config/nds32/nds32-memory-manipulation.c
28041 (nds32_expand_load_multiple): Refine implementation.
28042 (nds32_expand_store_multiple): Ditto.
28043 * config/nds32/nds32-multiple.md
28044 (load_multiple): Update nds32_expand_load_multiple interface.
28045 (store_multiple): Update nds32_expand_store_multiple interface.
28046 * config/nds32/nds32-predicates.c
28047 (nds32_valid_multiple_load_store): Rename ...
28048 (nds32_valid_multiple_load_store_p): ... to this and refine
28050 * config/nds32/predicates.md
28051 (nds32_load_multiple_and_update_address_operation): New predicate.
28052 (nds32_store_multiple_and_update_address_operation): New predicate.
28054 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
28055 Chung-Ju Wu <jasonwucj@gmail.com>
28057 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
28058 (combo): New attribute.
28059 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
28061 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
28063 * config/nds32/nds32.opt: Change -mcmodel= default value.
28065 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
28066 Monk Chiang <sh.chiang04@gmail.com>
28067 Chung-Ju Wu <jasonwucj@gmail.com>
28069 * config/nds32/constants.md (unspec_element): New enum.
28070 * config/nds32/constraints.md (Umw): New constraint.
28071 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
28072 * config/nds32/nds32-intrinsic.md: Likewise.
28073 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
28074 (nds32_valid_smw_lwm_base_p): New.
28075 (nds32_output_smw_single_word): New.
28076 (nds32_output_lmw_single_word): New.
28077 (nds32_expand_unaligned_load): New.
28078 (nds32_expand_unaligned_store): New.
28079 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
28080 (nds32_output_smw_single_word): Declare.
28081 (nds32_output_lmw_single_word): Declare.
28082 (nds32_expand_unaligned_load): Declare.
28083 (nds32_expand_unaligned_store): Declare.
28084 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
28085 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
28086 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
28087 NDS32_BUILTIN_UASTORE_DW.
28088 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
28091 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
28092 Kito Cheng <kito.cheng@gmail.com>
28093 Chung-Ju Wu <jasonwucj@gmail.com>
28095 * config/nds32/nds32-intrinsic.c
28096 (nds32_expand_builtin_null_ftype_reg): Delete.
28097 (nds32_expand_builtin_reg_ftype_imm): Ditto.
28098 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
28099 (nds32_read_argument): New.
28100 (nds32_legitimize_target): Ditto.
28101 (nds32_legitimize_argument): Ditto.
28102 (nds32_check_constant_argument): Ditto.
28103 (nds32_expand_unop_builtin): Ditto.
28104 (nds32_expand_unopimm_builtin): Ditto.
28105 (nds32_expand_binop_builtin): Ditto.
28106 (nds32_builtin_decl_impl): Ditto.
28107 (builtin_description): Ditto.
28108 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
28109 (nds32_init_builtins_impl): Ditto.
28110 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
28111 (nds32_builtin_decl): New.
28112 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
28113 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
28115 2018-03-02 Jeff Law <law@redhat.com>
28117 * reorg.c (stop_search_p): Handle DEBUG_INSN.
28118 (redundant_insn, fill_simple_delay_slots): Likewise.
28119 (fill_slots_from_thread): Likewise.
28120 * resource.c (mark_referenced_resources): Likewise.
28121 (mark_set_resources, find_dead_or_set_registers): Likewise.
28123 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28125 * substring-locations.h (format_warning_va): Formatting fix for
28126 ATTRIBUTE_GCC_DIAG.
28127 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
28129 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
28130 * substring-locations.c: Include intl.h.
28131 (format_warning_va): Turned into small wrapper around
28132 format_warning_n_va, renamed to ...
28133 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
28134 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
28136 (format_warning_at_substring_n): New function.
28137 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
28138 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
28139 format_warning_at_substring with just a shorter name instead of
28140 const function pointer.
28141 (fmtwarn_n): New function.
28142 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
28143 appropriate, get rid of all the fmtstr temporaries, move conditionals
28144 with G_() wrapped string literals directly into fmtwarn arguments,
28145 cast dir.len to (int), formatting fixes.
28147 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
28149 * doc/invoke.texi: Remove "Cilk Plus" references.
28151 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28152 Richard Biener <rguenther@suse.de>
28155 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
28156 for error or warning attributes if CALL_FROM_THUNK_P is set.
28159 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28162 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
28163 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
28166 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
28167 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
28169 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
28170 instead of -1U in last predictors element's probability member.
28172 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
28175 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
28176 arguments if they are comparable.
28178 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
28180 PR tree-optimization/84634
28181 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
28182 masks and masked_loop_p with a single loop_masks, making sure it's
28183 null for bb vectorization.
28185 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
28187 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
28188 (vect_analyze_data_ref_access): Use loop->safe_len rather than
28189 loop->force_vectorize to check whether there is no alias.
28191 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28194 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
28196 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
28198 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
28199 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
28200 instead of a loop around prev_real_insn.
28201 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
28204 PR inline-asm/84625
28205 * config/i386/i386.c (ix86_print_operand): Use conditional
28206 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
28209 2018-03-02 Richard Biener <rguenther@suse.de>
28211 PR tree-optimization/84427
28212 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
28213 (bitmap_set_subtract_values): Rewrite to handle multiple
28216 (prune_clobbered_mems): Likewise.
28217 (phi_translate): Take edge instead of pred/phiblock.
28218 (phi_translate_1): Likewise.
28219 (phi_translate_set): Likewise. Insert all translated
28220 exprs for a value into the set, keeping possibly multiple
28221 expressions per value.
28222 (compute_antic_aux): Adjust for phi_translate changes.
28223 When intersecting union the expressions and prune those
28224 not in the final value set, keeping possibly multiple
28225 expressions per value. Do not use value-insertion
28226 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
28227 all expressions. Add verification that the value-sets
28228 only shrink during iteration.
28229 (compute_partial_antic_aux): Adjust for the phi_translate changes.
28230 (do_pre_regular_insertion): Likewise.
28231 (do_pre_partial_partial_insertion): Likewise.
28233 2018-03-02 Richard Biener <rguenther@suse.de>
28236 * config/darwin.c (saved_debug_info_level): New static global.
28237 (darwin_asm_lto_start): Disable debug info generation for LTO out.
28238 (darwin_asm_lto_end): Restore debug info generation settings.
28240 2018-03-01 Martin Liska <mliska@suse.cz>
28243 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
28244 volatile arguments.
28246 2018-03-01 Richard Biener <rguenther@suse.de>
28249 * dwarf2out.c (gen_variable_die): Properly handle late VLA
28250 type annotation with LTO when debug was disabled at compile-time.
28252 2018-03-01 Matthew Fortune <mfortune@gmail.com>
28254 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
28256 (mips_final_postscan_insn): Likewise.
28258 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
28260 PR rtl-optimization/84528
28261 * alias.c (init_alias_target): Add commentary.
28262 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
28263 a unique base value if the frame pointer is not eliminated
28264 to the stack pointer.
28266 2018-03-01 Tom de Vries <tom@codesourcery.com>
28268 PR rtl-optimization/83327
28269 * lra-int.h (hard_regs_spilled_into): Declare.
28270 * lra.c (hard_regs_spilled_into): Define.
28271 (init_reg_info): Init hard_regs_spilled_into.
28272 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
28273 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
28274 (process_bb_lives): Handle hard_regs_spilled_into.
28275 (lra_create_live_ranges_1): Before doing liveness propagation, clear
28276 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
28278 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
28280 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
28281 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
28282 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
28283 * config/rs6000/aix72.h: New file.
28285 2018-02-28 Jakub Jelinek <jakub@redhat.com>
28287 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
28288 instead of warning_at with conditional singular and plural messages
28292 * stor-layout.c (update_alignment_for_field): For
28293 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
28294 && !DECL_PACKED (field), do the alignment update, just use
28295 only desired_align instead of MAX (type_align, desired_align)
28297 (place_field): Don't do known_align < desired_align handling
28298 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
28299 is non-NULL, instead do it after rli->prev_field handling and
28300 only if not within a bitfield word. For DECL_PACKED (field)
28301 use type_align of BITS_PER_UNIT.
28303 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
28305 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
28306 superfluous parentheses and trailing spaces.
28308 2018-02-28 Richard Biener <rguenther@suse.de>
28310 PR tree-optimization/84584
28311 * graphite-scop-detection.c (scop_detection::add_scop): Discard
28312 SCoPs with fake exit edge.
28314 2018-02-28 Martin Liska <mliska@suse.cz>
28317 * timevar.c (timer::print): Fix format to properly print 100%
28320 2018-02-28 Richard Biener <rguenther@suse.de>
28322 PR middle-end/84607
28323 * genmatch.c (capture_info::walk_match): Do not mark
28324 captured expressions without operands as expr_p given
28325 they act more like predicates and should be subject to
28326 "lost tail" side-effect preserving.
28328 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
28330 PR rtl-optimization/81611
28331 * auto-inc-dec.c (attempt_change): Move dead note from
28332 mem_insn if it's the next use of regno
28333 (find_address): Take address use of reg holding
28334 non-incremented value. Add parm to limit search to the named
28336 (merge_in_block): Attempt to use a mem insn that is the next
28337 use of the original regno.
28339 2018-02-27 Martin Sebor <msebor@redhat.com>
28342 * doc/invoke.texi (-Wmissing-attributes): New option.
28343 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
28345 2018-02-27 Martin Sebor <msebor@redhat.com>
28347 PR translation/84207
28348 * diagnostic-core.h (warning_n, error_n, inform_n): Change
28349 n argument to unsigned HOST_WIDE_INT.
28350 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
28351 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
28352 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
28353 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
28355 2018-02-27 Richard Biener <rguenther@suse.de>
28357 PR tree-optimization/84512
28358 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
28359 Do not use the estimate returned from record_stmt_cost for
28360 the scalar iteration cost but sum properly using add_stmt_cost.
28362 2018-02-27 Richard Biener <rguenther@suse.de>
28364 PR tree-optimization/84466
28365 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
28366 Adjust last change to less strictly validate use operands.
28368 2018-02-27 Martin Liska <mliska@suse.cz>
28370 PR gcov-profile/84548
28371 * gcov.c (process_file): Allow partial overlap and consider it
28372 also as group functions.
28373 (output_lines): Properly calculate range of lines for a group.
28375 2018-02-27 Martin Liska <mliska@suse.cz>
28377 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
28378 'ggc' suffixes. Change first column width.
28379 (timer::print): Fix formatting of the column.
28381 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
28383 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
28384 preserve inline entry blocks for the sake of debug inline
28385 entry point markers alone.
28386 (remove_unused_locals): Suggest in comments a better place to
28387 force the preservation of inline entry blocks that are
28388 otherwise unused, but do not preserve them.
28390 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
28392 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
28394 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
28397 * config/i386/constraints.md (Bs): Replace
28398 ix86_indirect_branch_register with
28399 TARGET_INDIRECT_BRANCH_REGISTER.
28401 * config/i386/i386.md (indirect_jump): Likewise.
28402 (tablejump): Likewise.
28403 (*sibcall_memory): Likewise.
28404 (*sibcall_value_memory): Likewise.
28405 Peepholes of indirect call and jump via memory: Likewise.
28406 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
28407 (*sibcall_value_GOT_32): Likewise.
28408 * config/i386/predicates.md (indirect_branch_operand): Likewise.
28409 (GOT_memory_operand): Likewise.
28410 (call_insn_operand): Likewise.
28411 (sibcall_insn_operand): Likewise.
28412 (GOT32_symbol_operand): Likewise.
28413 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
28415 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
28417 PR rtl-optimization/83496
28418 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
28419 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
28420 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
28421 redundant insn, if any.
28422 (relax_delay_slots): Likewise.
28423 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
28425 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
28427 PR tree-optimization/83965
28428 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
28429 that grouped statements are part of a reduction chain. Return
28430 true if the statement is not marked as a reduction itself but
28431 is part of a group.
28432 (vect_recog_dot_prod_pattern): Don't check whether the statement
28433 is part of a group here.
28434 (vect_recog_sad_pattern): Likewise.
28435 (vect_recog_widen_sum_pattern): Likewise.
28437 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
28440 * final.c (rest_of_clean_state): Also look for calls inside sequences.
28442 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
28445 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
28447 (ix86_output_indirect_function_return): New prototype.
28448 (ix86_split_simple_return_pop_internal): Likewise.
28449 * config/i386/i386.c (indirect_return_via_cx): New.
28450 (indirect_return_via_cx_bnd): Likewise.
28451 (indirect_thunk_name): Handle return va CX_REG.
28452 (output_indirect_thunk_function): Create alias for
28453 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
28454 (ix86_output_indirect_jmp): Remove the bool argument.
28455 (ix86_output_indirect_function_return): New function.
28456 (ix86_split_simple_return_pop_internal): Likewise.
28457 * config/i386/i386.md (*indirect_jump): Don't pass false
28458 to ix86_output_indirect_jmp.
28459 (*tablejump_1): Likewise.
28460 (simple_return_pop_internal): Change it to define_insn_and_split.
28461 Call ix86_split_simple_return_pop_internal to split it for
28462 -mfunction-return=.
28463 (simple_return_indirect_internal): Call
28464 ix86_output_indirect_function_return instead of
28465 ix86_output_indirect_jmp.
28467 2018-02-26 Jakub Jelinek <jakub@redhat.com>
28470 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
28471 memset and value initialization afterwards.
28473 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
28475 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
28477 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
28480 * common/config/aarch64/aarch64-common.c
28481 (aarch_option_optimization_table[]): Switch
28482 off fomit-frame-pointer
28484 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
28485 Chung-Ju Wu <jasonwucj@gmail.com>
28487 * config/nds32/nds32-multiple.md (load_multiple): Disallow
28489 (store_multiple): Ditto.
28491 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
28493 * config.gcc: Add --with-cpu support for nds32 target.
28494 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
28495 * config/nds32/nds32.opt: Add -mcpu= option.
28497 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
28499 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
28500 isel=yes): Warn for these deprecated options.
28502 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
28504 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
28505 ISA_2_5_MASKS_EMBEDDED.
28507 2018-02-23 Jakub Jelinek <jakub@redhat.com>
28509 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
28510 p->max as pointers rather than using iterative_hash_expr.
28512 2018-02-23 Carl Love <cel@us.ibm.com>
28514 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
28515 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
28517 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
28518 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
28519 P8V_BUILTIN_VEC_VUNSIGNED2.
28521 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
28524 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
28525 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
28527 * lra-constraints.c (curr_insn_transform): Set up
28528 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
28530 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
28531 * lra-eliminations.c (spill_pseudos): Ditto.
28532 (process_insn_for_elimination): Ditto.
28533 * lra-lives.c (reg_early_clobber_p): Use the new macros.
28534 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
28535 LRA_NON_CLOBBERED_ALT.
28537 2018-02-22 Martin Sebor <msebor@redhat.com>
28539 PR tree-optimization/84480
28540 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
28541 to maybe_diag_stxncpy_trunc. Call it.
28542 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
28543 from gimple_fold_builtin_strcpy. Print inlining stack.
28544 (handle_builtin_stxncpy): Print inlining stack.
28545 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
28547 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
28550 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
28551 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
28552 and -fcheck-pointer-bounds are used together.
28553 (indirect_thunk_prefix): New enum.
28554 (indirect_thunk_need_prefix): New function.
28555 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
28556 "_nt" instead of "_bnd" for NOTRACK prefix.
28557 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
28558 (output_indirect_thunk_function): Likewise.
28560 (ix86_code_end): Update output_indirect_thunk_function calls.
28561 (ix86_output_indirect_branch_via_reg): Replace
28562 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
28563 (ix86_output_indirect_branch_via_push): Likewise.
28564 (ix86_output_function_return): Likewise.
28565 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
28566 incompatible with -fcf-protection=branch and
28567 -fcheck-pointer-bounds.
28569 2018-02-22 Steve Ellcey <sellcey@cavium.com>
28572 * config/aarch64/aarch64.c (aarch64_print_address_internal):
28573 Change gcc_assert call to output_operand_lossage.
28575 2018-02-22 Steve Ellcey <sellcey@cavium.com>
28577 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
28579 2018-02-22 DJ Delorie <dj@redhat.com>
28580 Sebastian Perta <sebastian.perta@renesas.com>
28581 Oleg Endo <olegendo@gcc.gnu.org>
28583 * config/rx/rx.c (rx_rtx_costs): New function.
28584 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
28586 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
28588 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
28590 2018-02-22 Martin Liska <mliska@suse.cz>
28593 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
28594 Add "native" as a possible value.
28596 2018-02-22 Martin Liska <mliska@suse.cz>
28599 * config/i386/i386.c (ix86_option_override_internal):
28600 Add "native" as a possible value for -march and -mtune.
28602 2018-02-22 Jakub Jelinek <jakub@redhat.com>
28605 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
28606 to all type variants.
28608 PR tree-optimization/84503
28609 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
28610 width as info->bitpos + info->bitsize - start.
28611 (merged_store_group::merge_overlapping): Simplify width computation.
28612 (check_no_overlap): New function.
28613 (imm_store_chain_info::try_coalesce_bswap): Compute expected
28614 start + width and last_order of the group, fail if check_no_overlap
28616 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
28617 to group if check_no_overlap fails.
28619 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
28621 * config/rs6000/altivec.md: Delete contraint arguments to
28622 define_expand, define_split, and define_peephole2, and in
28623 define_insn_and_split if always unused.
28624 * config/rs6000/darwin.md: Ditto.
28625 * config/rs6000/dfp.md: Ditto.
28626 * config/rs6000/rs6000.md: Ditto.
28627 * config/rs6000/sync.md: Ditto.
28628 * config/rs6000/vector.md: Ditto.
28629 * config/rs6000/vsx.md: Ditto.
28631 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
28633 * config/rs6000/altivec.md: Write output control strings as braced
28634 blocks instead of double-quoted strings.
28635 * config/rs6000/darwin.md: Ditto.
28636 * config/rs6000/rs6000.md: Ditto.
28637 * config/rs6000/vector.md: Ditto.
28638 * config/rs6000/vsx.md: Ditto.
28640 2018-02-21 Jason Merrill <jason@redhat.com>
28642 PR c++/84314 - ICE with templates and fastcall attribute.
28643 * attribs.c (build_type_attribute_qual_variant): Remove assert.
28645 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
28647 * ipa-cp.c (determine_versionability): Fix comment typos.
28649 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
28652 * ipa-cp.c (determine_versionability): Do not version functions caling
28655 2018-02-21 Martin Liska <mliska@suse.cz>
28658 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
28659 Add "native" as a possible value.
28660 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
28661 the macro when native cpu detection is available.
28663 2018-02-21 Martin Liska <mliska@suse.cz>
28666 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
28667 Add "native" as a possible value.
28668 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
28669 when native cpu detection is available.
28671 2018-02-21 Jakub Jelinek <jakub@redhat.com>
28672 Martin Sebor <msebor@redhat.com>
28674 PR tree-optimization/84478
28675 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
28677 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
28678 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
28679 support which is conservatively correct, for 2 only stay conservative
28680 for maxlen. Formatting and comment capitalization fixes. Add STRICT
28681 argument to the 2 argument get_range_strlen, adjust 6 arg
28682 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
28684 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
28685 (gimple_fold_builtin_strlen): Pass true as last argument to
28688 2018-02-20 Martin Sebor <msebor@redhat.com>
28690 PR middle-end/84095
28691 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
28692 (builtin_memref::set_base_and_offset): Same. Handle inner references.
28693 (builtin_memref::builtin_memref): Factor out parts into
28694 set_base_and_offset and call it.
28696 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
28698 PR middle-end/84406
28699 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
28700 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
28701 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
28702 search at the associated MODE_INT.
28704 2018-02-20 Jeff Law <law@redhat.com>
28706 PR middle-end/82123
28707 PR tree-optimization/81592
28708 PR middle-end/79257
28709 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
28710 for range data rather than using global data.
28711 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
28712 range data rather than using global data.
28713 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
28714 pass it to children as needed.
28715 (struct directive::fmtresult): Similarly.
28716 (struct directive::set_width): Similarly.
28717 (struct directive::set_precision): Similarly.
28718 (format_integer, format_directive, parse_directive): Similarly.
28719 (format_none): Accept unnamed vr_values parameter.
28720 (format_percent, format_floating, format_character): Similarly.
28721 (format_string, format_plain): Similarly.
28722 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
28723 the EVRP range analyzer for range data rather than using global data.
28724 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
28725 gimple-ssa-evrp-analyze.h
28726 (class sprintf_dom_walker): Add after_dom_children member function.
28727 Add evrp_range_analyzer member.
28728 (sprintf_dom_walker::before_dom_children): Call into the EVRP
28729 range analyzer as needed.
28730 (sprintf_dom_walker::after_dom_children): New member function.
28731 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
28733 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
28734 (evrp_range_analyzer::pop_to_marker): Likewise.
28736 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
28738 PR tree-optimization/84419
28739 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
28740 with the required type if its current type is compatible but
28743 2018-02-20 Jakub Jelinek <jakub@redhat.com>
28745 PR middle-end/82004
28746 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
28747 after vectorization.
28749 2018-02-20 Martin Liska <mliska@suse.cz>
28752 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
28753 possible values if we don't have a hint.
28755 2018-02-20 Martin Liska <mliska@suse.cz>
28759 * final.c (shorten_branches): Build align_tab array with one
28761 * opts.c (finish_options): Add alignment option limit check.
28762 (MAX_CODE_ALIGN): Likewise.
28763 (MAX_CODE_ALIGN_VALUE): Likewise.
28764 * doc/invoke.texi: Document maximum allowed option value for
28765 all -falign-* options.
28767 2018-02-19 Jakub Jelinek <jakub@redhat.com>
28770 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
28771 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
28772 * var-tracking.c (emit_note_insn_var_location): Remove all references
28773 to NOTE_INSN_CALL_ARG_LOCATION.
28774 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
28775 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
28776 Use copy_rtx_if_shared.
28777 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
28778 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
28779 (dwarf2out_var_location): Remove handling of
28780 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
28782 * final.c (final_scan_insn): Remove all references to
28783 NOTE_INSN_CALL_ARG_LOCATION.
28784 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
28785 before dumping final insns.
28786 * except.c (emit_note_eh_region_end): Remove all references to
28787 NOTE_INSN_CALL_ARG_LOCATION.
28788 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
28789 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
28790 * config/arc/arc.c (hwloop_optimize): Likewise.
28791 * config/arm/arm.c (create_fix_barrier): Likewise.
28792 * config/s390/s390.c (s390_chunkify_start): Likewise.
28793 * config/sh/sh.c (find_barrier): Likewise.
28794 * config/i386/i386.c (rest_of_insert_endbranch,
28795 ix86_seh_fixup_eh_fallthru): Likewise.
28796 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
28797 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
28798 * config/frv/frv.c (frv_function_prologue): Likewise.
28799 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
28801 (note_outside_basic_block_p): Remove all references to
28802 NOTE_INSN_CALL_ARG_LOCATION.
28803 * gengtype.c (adjust_field_rtx_def): Likewise.
28804 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
28806 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
28807 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
28810 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
28813 PR tree-optimization/84452
28814 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
28815 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
28818 2018-02-19 Martin Liska <mliska@suse.cz>
28821 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
28823 2018-02-19 Martin Liska <mliska@suse.cz>
28824 Richard Sandiford <richard.sandiford@linaro.org>
28826 PR tree-optimization/82491
28827 * gimple-fold.c (get_base_constructor): Make earlier bail out
28830 2018-02-19 Carl Love <cel@us.ibm.com>
28832 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
28833 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
28835 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
28836 P8V_BUILTIN_VEC_NEG.
28838 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
28840 * config/rl78/rl78.md (movdf): New define expand.
28842 2018-02-19 Martin Liska <mliska@suse.cz>
28845 * doc/invoke.texi: Fix typo.
28846 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
28848 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
28850 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
28851 handle rs6000_single_float and rs6000_double_float specially for
28854 2018-02-16 Jeff Law <law@redhat.com>
28856 * config/rx/rx.c (add_pop_cfi_notes): New function.;
28857 (pop_regs): Use it.
28859 2018-02-16 Jakub Jelinek <jakub@redhat.com>
28862 * ipa-inline.c (inline_small_functions): Fix a typo.
28864 2018-02-16 Nathan Sidwell <nathan@acm.org>
28866 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
28868 2018-02-16 Carl Love <cel@us.ibm.com>
28870 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
28871 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
28872 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
28873 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
28874 expansion to P8V_BUILTIN_VEC_FLOAT2.
28876 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
28878 PR rtl-optimization/70023
28879 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
28880 src_regno into account.
28882 2018-02-16 Carl Love <cel@us.ibm.com>
28884 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
28885 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
28886 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
28887 * config/rs6000/rs6000.c: Remove case statements for
28888 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
28889 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
28890 and P9V_BUILTIN_VEC_VINSERT4B.
28891 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
28892 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
28893 * config/rs6000/vsx.md:
28894 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
28897 2018-02-16 Carl Love <cel@us.ibm.com>
28899 * config/rs6000/altivec.h: Add builtin names vec_extract4b
28901 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
28903 * config/rs6000/rs6000-c.c: Add the definitions for
28904 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
28905 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
28906 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
28907 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
28908 definition for insert4b and define insn *insert3b_internal.
28909 * doc/extend.texi: Add documentation for vec_extract4b.
28911 2018-02-16 Nathan Sidwell <nathan@acm.org>
28913 * doc/extend.texi (Backwards Compatibility): Mention friend
28914 injection. Note for-scope is deprecated.
28915 * doc/invoke.texi (-ffriend-injection): Deprecate.
28917 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
28919 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
28920 that moved to I2, also allow destinations that are a paradoxical
28921 subreg (instead of a normal reg).
28923 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
28926 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
28929 2018-02-16 Richard Biener <rguenther@suse.de>
28931 PR tree-optimization/84037
28932 PR tree-optimization/84016
28934 * config/i386/i386.c (ix86_builtin_vectorization_cost):
28935 Adjust vec_construct for the fact we need additional higher latency
28936 128bit inserts for AVX256 and AVX512 vector builds.
28937 (ix86_add_stmt_cost): Scale vector construction cost for
28940 2018-02-16 Richard Biener <rguenther@suse.de>
28942 PR tree-optimization/84417
28943 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
28944 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
28945 (non_rewritable_lvalue_p): Likewise, use poly-ints.
28947 2018-02-16 Martin Liska <mliska@suse.cz>
28950 * internal-fn.def (ASAN_CHECK): Set proper flags.
28951 (ASAN_MARK): Likewise.
28953 2018-02-16 Julia Koval <julia.koval@intel.com>
28955 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
28956 from PTA_CANNONLAKE.
28958 2018-02-16 Jakub Jelinek <jakub@redhat.com>
28961 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
28962 Use ++iter rather than iter++ for std::list iterators.
28963 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
28964 defer deleting them until all nodes in the forest are processed. Do
28965 free even leaf nodes. Change to_process into auto_vec.
28968 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
28969 * vec.h (vec_default_construct): Use memset instead of placement new
28970 if BROKEN_VALUE_INITIALIZATION is defined.
28971 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
28972 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
28975 PR rtl-optimization/83723
28976 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
28977 * lra.c (lra_substitute_pseudo): Likewise. If true, use
28978 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
28980 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
28982 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
28984 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
28986 PR rtl-optimization/81443
28987 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
28988 from inner REGs to paradoxical SUBREGs.
28990 2018-02-16 Richard Biener <rguenther@suse.de>
28992 PR tree-optimization/84399
28993 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
28994 For operands we can analyze at their definition make sure we can
28995 analyze them at each use as well.
28997 2018-02-16 Richard Biener <rguenther@suse.de>
28999 PR tree-optimization/84190
29000 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
29001 volatile accesses if the decl isn't volatile.
29003 2018-02-15 Jason Merrill <jason@redhat.com>
29005 PR c++/84314 - ICE with templates and fastcall attribute.
29006 * attribs.c (build_type_attribute_qual_variant): Don't clobber
29007 TYPE_CANONICAL on an existing type.
29009 2018-02-15 Jakub Jelinek <jakub@redhat.com>
29011 PR tree-optimization/84383
29012 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
29013 dstoff nor call operand_equal_p if dstbase is NULL.
29015 PR tree-optimization/84334
29016 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
29017 also a CONSTANT_CLASS_P, punt.
29019 2018-02-14 Jim Wilson <jimw@sifive.com>
29021 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
29022 first SMALL_OPERAND check. New local min_second_step. Move assert
29023 to where locals are set. Add TARGET_RVC support.
29024 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
29026 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
29028 * doc/invoke.texi: Correct -Wformat-overflow code sample.
29030 2018-02-14 Martin Sebor <msebor@redhat.com>
29032 PR tree-optimization/83698
29033 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
29034 arrays constrain the offset range to their bounds.
29035 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
29036 (builtin_access::overlap): Avoid setting the size of overlap if it's
29038 (maybe_diag_overlap): Also consider arrays when deciding what values
29039 of offsets to include in diagnostics.
29041 2018-02-14 Martin Sebor <msebor@redhat.com>
29044 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
29045 that correspond to the kind of a declaration.
29047 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
29050 * config/pa/pa.md: Load address of PIC label using the linkage table
29051 if the label is nonlocal.
29053 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
29055 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
29056 warning message if user requests -maltivec=be.
29057 * doc/invoke.texi: Document deprecation of -maltivec=be.
29059 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
29062 * config/rs6000/rs6000-c.c: Update definitions for
29063 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
29064 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
29066 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
29069 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
29070 add _get_ssp intrinsics. Remove argument from
29071 __builtin_ia32_rdssp[d|q].
29072 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
29073 * config/i386/i386-builtin.def: Remove argument from
29074 __builtin_ia32_rdssp[d|q].
29075 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
29076 ix86_expand_special_args_builtin for _rdssp[d|q].
29077 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
29078 Clear register before usage.
29079 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
29080 Add documentation for new _get_ssp and _inc_ssp intrinsics.
29082 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
29084 PR tree-optimization/84357
29085 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
29086 operand 1 of an ARRAY_REF too.
29088 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
29091 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
29092 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
29094 (set_of_reg): New struct.
29095 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
29096 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
29097 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
29099 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
29100 Split into bitclr, bitset, bitinvert patterns if appropriate.
29101 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
29102 use rx_fuse_in_memory_bitop.
29103 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
29104 to named insn, correct maximum insn length.
29106 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
29109 * machmode.def: Define a complex mode for PARTIAL_INT.
29110 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
29112 * doc/rtl.texi: Document CSPImode.
29113 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
29115 (msp430_hard_regno_nregs_with_padding): Likewise.
29117 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
29120 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
29122 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
29124 PR rtl-optimization/84169
29125 * combine.c (try_combine): New variable split_i2i3. Set it to true if
29126 we generated a parallel as new i3 and we split that to new i2 and i3
29127 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
29128 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
29129 those to i2, not i1. Partially rewrite this scan code.
29131 2018-02-13 Jakub Jelinek <jakub@redhat.com>
29134 * stor-layout.c (place_field): For variable length fields, adjust
29135 offset_align afterwards not just based on the field's alignment,
29136 but also on the size.
29138 PR middle-end/84309
29139 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
29140 of exps and logs in the use_exp2 case.
29142 2018-02-13 Jeff Law <law@redhat.com>
29144 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
29145 entry for "vector".
29147 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
29150 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
29154 * common.opt (gas-loc-support, gas-locview-support): New.
29155 (ginline-points, ginternal-reset-location-views): New.
29156 * doc/invoke.texi: Document them. Use @itemx where intended.
29157 (gvariable-location-views): Adjust.
29158 * target.def (reset_location_view): New.
29159 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
29160 (TARGET_RESET_LOCATION_VIEW): New.
29161 * doc/tm.texi: Rebuilt.
29162 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
29163 (dwarf2out_default_as_locview_support): New.
29164 (output_asm_line_debug_info): Use option variables.
29165 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
29166 (output_loc_list): Likewise.
29167 (add_high_low_attributes): Check option variables.
29168 Don't output entry view attribute in strict mode.
29169 (gen_inlined_subroutine_die): Check option variables.
29170 (dwarf2out_inline_entry): Likewise.
29171 (init_sections_and_labels): Likewise.
29172 (dwarf2out_early_finish): Likewise.
29173 (maybe_reset_location_view): New, from...
29174 (dwarf2out_var_location): ... here. Call it.
29175 * debug.h (dwarf2out_default_as_loc_support): Declare.
29176 (dwarf2out_default_as_locview_support): Declare.
29177 * hooks.c (hook_int_rtx_insn_0): New.
29178 * hooks.h (hook_int_rtx_insn_0): Declare.
29179 * toplev.c (process_options): Take -gas-loc-support and
29180 -gas-locview-support from dwarf2out. Enable
29181 -gvariable-location-views by default only with locview
29182 assembler support. Enable -ginternal-reset-location-views by
29183 default only if the target defines the corresponding hook.
29184 Enable -ginline-points by default if location views are
29185 enabled; force it disabled if statement frontiers are
29187 * tree-inline.c (expand_call_inline): Check option variables.
29188 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
29190 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
29192 PR tree-optimization/84321
29193 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
29194 handling. Also check whether the anti-range contains any values
29195 that satisfy the mask; switch to a VR_RANGE if not.
29197 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
29200 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
29202 2018-02-13 Martin Jambor <mjambor@suse.cz>
29205 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
29206 of call statements, also set location of a load to a temporary.
29208 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
29210 * config/rl78/rl78.c (add_vector_labels): New function.
29211 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
29212 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
29213 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
29214 which checks that no arguments are passed.
29215 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
29216 * doc/extend.texi: Documentation for the new attribute.
29218 2018-02-13 Andreas Schwab <schwab@suse.de>
29220 * config/riscv/linux.h (CPP_SPEC): Define.
29222 2018-02-13 Jakub Jelinek <jakub@redhat.com>
29225 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
29226 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
29227 OPTION_MASK_ISA_AES as first argument to def_builtin_const
29228 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
29229 instead of OPTION_MASK_ISA_PCLMUL as first argument to
29230 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
29231 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
29232 temporarily for AES and PCLMUL builtins.
29234 PR tree-optimization/84339
29235 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
29236 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
29239 PR middle-end/84309
29240 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
29241 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
29242 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
29244 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
29246 * omp-simd-clone.h: New file.
29247 * omp-simd-clone.c: Include omp-simd-clone.h.
29248 (expand_simd_clones): No longer static.
29249 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
29250 cgraph.h and omp-simd-clone.h.
29251 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
29252 (vect_recog_widen_shift_pattern): Formatting fix.
29253 (vect_pattern_recog_1): Don't check optab for calls.
29256 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
29257 operands[2] into a REG before using gen_lowpart on it.
29259 2018-02-12 Jeff Law <law@redhat.com>
29262 * config/sh/sh.c (find_barrier): Consider a sibling call
29265 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
29266 successfully back substituting a reg.
29268 2018-02-12 Richard Biener <rguenther@suse.de>
29270 PR tree-optimization/84037
29271 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
29272 parameter, move visited init to caller.
29273 (vect_slp_analyze_operations): Separate cost from validity
29274 check, initialize visited once for all instances.
29275 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
29277 * tree-vect-stmts.c (vect_model_simple_cost): Make early
29279 (vect_model_promotion_demotion_cost): Likewise.
29280 (vectorizable_bswap): Guard cost modeling with !slp_node
29281 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
29283 (vectorizable_call): Likewise.
29284 (vectorizable_conversion): Likewise.
29285 (vectorizable_assignment): Likewise.
29286 (vectorizable_shift): Likewise.
29287 (vectorizable_operation): Likewise.
29288 (vectorizable_store): Likewise.
29289 (vectorizable_load): Likewise.
29290 (vectorizable_condition): Likewise.
29291 (vectorizable_comparison): Likewise.
29293 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
29296 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
29297 (ASAN_MARK): Fix fnspec to account for return value, change pointer
29298 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
29300 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
29302 PR middle-end/83665
29303 * params.def (inline-min-speedup): Increase from 8 to 15.
29304 (max-inline-insns-auto): Decrease from 40 to 30.
29305 * ipa-split.c (consider_split): Add some buffer for function to
29306 be considered inlining candidate.
29307 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
29310 2018-02-12 Richard Biener <rguenther@suse.de>
29312 PR tree-optimization/84037
29313 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
29314 matched stmts if we cannot swap the non-matched ones.
29316 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
29318 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
29319 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
29320 _mm_maskz_scalef_round_ss): New intrinsics.
29321 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
29322 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
29323 __builtin_ia32_scalefss_round): Remove.
29324 (__builtin_ia32_scalefsd_mask_round,
29325 __builtin_ia32_scalefss_mask_round): New intrinsics.
29326 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
29327 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
29328 ((match_operand:VF_128 2 "<round_nimm_predicate>"
29329 "<round_constraint>")): Changed to ...
29330 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
29331 "<round_scalar_constraint>")): ... this.
29332 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
29333 %0, %1, %2<round_op3>}"): Changed to ...
29334 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
29335 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
29336 %2<round_scalar_mask_op3>}"): ... this.
29337 * config/i386/subst.md (round_scalar_nimm_predicate): New.
29339 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
29341 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
29342 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
29343 (_mm_maskz_sqrt_round_ss): New intrinsics.
29344 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
29345 (__builtin_ia32_sqrtsd_mask_round)
29346 (__builtin_ia32_sqrtss_mask_round): New builtins.
29347 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
29348 (__builtin_ia32_sqrtss_round): Remove.
29349 (__builtin_ia32_sqrtsd_mask_round)
29350 (__builtin_ia32_sqrtss_mask_round): New builtins.
29351 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
29352 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
29353 ((match_operand:VF_128 1 "vector_operand"
29354 "xBm,<round_constraint>")): Changed to ...
29355 ((match_operand:VF_128 1 "vector_operand"
29356 "xBm,<round_scalar_constraint>")): ... this.
29357 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
29358 %0, %2, %<iptr>1<round_op3>}): Changed to ...
29359 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
29360 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
29361 %<iptr>1<round_scalar_mask_op3>}): ... this.
29362 ((set_attr "prefix" "<round_prefix>")): Changed to ...
29363 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
29365 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
29368 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
29369 Cast vec_cmpeq result to correct type.
29370 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
29371 Cast vec_cmpgt result to correct type.
29373 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
29375 * final.c (final_scan_insn_1): Renamed from...
29376 (final_scan_insn): ... this. New wrapper, to recover
29377 seen from the outermost call in recursive ones.
29378 * config/sparc/sparc.c (output_return): Drop seen from call.
29379 (output_sibcall): Likewise.
29380 * config/visium/visium.c (output_branch): Likewise.
29382 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
29384 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
29387 2018-02-10 Alan Modra <amodra@gmail.com>
29390 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
29391 Specify LR as an input.
29393 2018-02-10 Jakub Jelinek <jakub@redhat.com>
29396 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
29397 remove_member_access_dummy_vars): New functions.
29398 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
29399 lower_omp_1, execute_lower_omp): Use them.
29401 PR rtl-optimization/84308
29402 * shrink-wrap.c (spread_components): Release todo vector.
29404 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
29406 PR rtl-optimization/57193
29407 * ira-color.c (struct allocno_color_data): Add member
29408 conflict_allocno_hard_prefs.
29409 (update_conflict_allocno_hard_prefs): New.
29410 (bucket_allocno_compare_func): Add a preference based on
29411 conflict_allocno_hard_prefs.
29412 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
29413 (color_allocnos): Remove a dead code. Initiate
29414 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
29416 2018-02-09 Jakub Jelinek <jakub@redhat.com>
29419 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
29420 constraint from =wa to wa. Avoid a subreg on the output operand,
29421 instead use a pseudo and subreg it in a move.
29422 (p9_xxbrd_<mode>): Changed to ...
29423 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
29424 (p9_xxbrd_v2df): New expander.
29425 (p9_xxbrw_<mode>): Changed to ...
29426 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
29427 (p9_xxbrw_v4sf): New expander.
29429 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
29431 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
29433 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
29436 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
29437 multiply in 32-bit mode.
29438 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
29439 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
29442 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
29444 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
29445 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
29446 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
29447 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
29449 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
29452 * dwarf2out.c (is_trivial_indirect_ref): New function.
29453 (dwarf2out_late_global_decl): Do not generate a location
29454 attribute for variables that have a non-trivial DECL_VALUE_EXPR
29455 and that are not defined in the current unit.
29457 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
29459 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
29460 instead of a libcall for UNORDERED.
29462 2018-02-09 Tamar Christina <tamar.christina@arm.com>
29465 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
29466 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
29468 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
29471 * config/s390/s390.c (s390_set_current_function): Invoke
29472 s390_indirect_branch_settings also if fndecl didn't change.
29474 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
29476 * config/rs6000/rs6000.md (blockage): Set length to zero.
29478 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
29480 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
29482 2018-02-09 Jakub Jelinek <jakub@redhat.com>
29485 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
29486 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
29490 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
29491 PARALLEL incoming that failed vt_get_decl_and_offset check.
29493 PR middle-end/84237
29494 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
29495 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
29497 (get_variable_section): For decls in named .bss* sections pass true as
29498 second argument to bss_initializer_p.
29500 2018-02-09 Marek Polacek <polacek@redhat.com>
29501 Jakub Jelinek <jakub@redhat.com>
29504 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
29505 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
29506 Sync some changes from cxx_fold_indirect_ref.
29508 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
29510 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
29512 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
29513 (BLOCK_INLINE_ENTRY_LABEL): New.
29514 (dwarf2out_var_location): Disregard inline entry markers.
29515 (inline_entry_data): New struct.
29516 (inline_entry_data_hasher): New hashtable type.
29517 (inline_entry_data_hasher::hash): New.
29518 (inline_entry_data_hasher::equal): New.
29519 (inline_entry_data_table): New variable.
29520 (add_high_low_attributes): Add DW_AT_entry_pc and
29521 DW_AT_GNU_entry_view attributes if a pending entry is found
29522 in inline_entry_data_table. Add old entry_pc attribute only
29523 if debug nonbinding markers are disabled.
29524 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
29525 markers are enabled.
29526 (block_within_block_p, dwarf2out_inline_entry): New.
29527 (dwarf2out_finish): Check that no entries remained in
29528 inline_entry_data_table.
29529 * final.c (reemit_insn_block_notes): Handle inline entry notes.
29530 (final_scan_insn, notice_source_line): Likewise.
29531 (rest_of_clean_state): Skip inline entry markers.
29532 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
29534 * gimple.c (gimple_build_debug_inline_entry): New.
29535 * gimple.h (enum gimple_debug_subcode): Add
29536 GIMPLE_DEBUG_INLINE_ENTRY.
29537 (gimple_build_debug_inline_entry): Declare.
29538 (gimple_debug_inline_entry_p): New.
29539 (gimple_debug_nonbind_marker_p): Adjust.
29540 * insn-notes.def (INLINE_ENTRY): New.
29541 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
29542 inline entry marker notes.
29543 (print_insn): Likewise.
29544 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
29545 (INSN_DEBUG_MARKER_KIND): Likewise.
29546 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
29547 * tree-inline.c (expand_call_inline): Build and insert
29548 debug_inline_entry stmt.
29549 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
29550 inline entry blocks early, if nonbind markers are enabled.
29551 (dump_scope_block): Dump fragment info.
29552 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
29553 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
29554 (gimple_build_debug_inline_entry): New.
29555 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
29556 Enable/disable inline entry points too.
29557 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
29558 (DEBUG_INSN): Describe inline entry markers.
29560 * common.opt (gvariable-location-views): New.
29561 (gvariable-location-views=incompat5): New.
29562 * config.in: Rebuilt.
29563 * configure: Rebuilt.
29564 * configure.ac: Test assembler for view support.
29565 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
29566 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
29567 * dwarf2out.c (var_loc_view): New typedef.
29568 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
29569 (dwarf2out_locviews_in_attribute): New.
29570 (dwarf2out_locviews_in_loclist): New.
29571 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
29572 (enum dw_line_info_opcode): Add LI_adv_address.
29573 (struct dw_line_info_table): Add view.
29574 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
29575 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
29576 (zero_view_p): New variable.
29577 (ZERO_VIEW_P): New macro.
29578 (output_asm_line_debug_info): New.
29579 (struct var_loc_node): Add view.
29580 (add_AT_view_list, AT_loc_list): New.
29581 (add_var_loc_to_decl): Add view param. Test it against last.
29582 (new_loc_list): Add view params. Record them.
29583 (AT_loc_list_ptr): Handle loc and view lists.
29584 (view_list_to_loc_list_val_node): New.
29585 (print_dw_val): Handle dw_val_class_view_list.
29586 (size_of_die): Likewise.
29587 (value_format): Likewise.
29588 (loc_list_has_views): New.
29589 (gen_llsym): Set vl_symbol too.
29590 (maybe_gen_llsym, skip_loc_list_entry): New.
29591 (dwarf2out_maybe_output_loclist_view_pair): New.
29592 (output_loc_list): Output view list or entries too.
29593 (output_view_list_offset): New.
29594 (output_die): Handle dw_val_class_view_list.
29595 (output_dwarf_version): New.
29596 (output_compilation_unit_header): Use it.
29597 (output_skeleton_debug_sections): Likewise.
29598 (output_rnglists, output_line_info): Likewise.
29599 (output_pubnames, output_aranges): Update version comments.
29600 (output_one_line_info_table): Output view numbers in asm comments.
29601 (dw_loc_list): Determine current endview, pass it to new_loc_list.
29602 Call maybe_gen_llsym.
29603 (loc_list_from_tree_1): Adjust.
29604 (add_AT_location_description): Create view list attribute if
29605 needed, check it's absent otherwise.
29606 (convert_cfa_to_fb_loc_list): Adjust.
29607 (maybe_emit_file): Call output_asm_line_debug_info for test.
29608 (dwarf2out_var_location): Reset views as needed. Precompute
29609 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
29610 attribute. Set view.
29611 (new_line_info_table): Reset next view.
29612 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
29613 (dwarf2out_source_line): Likewise. Output view resets and labels to
29614 the assembler, or select appropriate line info opcodes.
29615 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
29616 (optimize_string_length): Catch it. Adjust.
29617 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
29618 dw_val_class_view_list, and remove it if no longer needed.
29619 (hash_loc_list): Hash view numbers.
29620 (loc_list_hasher::equal): Compare them.
29621 (optimize_location_lists): Check whether a view list symbol is
29622 needed, and whether the locview attribute is present, and
29623 whether they match. Remove the locview attribute if no longer
29625 (index_location_lists): Call skip_loc_list_entry for test.
29626 (dwarf2out_finish): Call output_asm_line_debug_info for test.
29627 Use output_dwarf_version.
29628 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
29629 (struct dw_val_node): Add val_view_list.
29630 * final.c (SEEN_NEXT_VIEW): New.
29631 (set_next_view_needed): New.
29632 (clear_next_view_needed): New.
29633 (maybe_output_next_view): New.
29634 (final_start_function): Rename to...
29635 (final_start_function_1): ... this. Take pointer to FIRST,
29636 add SEEN parameter. Emit param bindings in the initial view.
29637 (final_start_function): Reintroduce SEEN-less interface.
29638 (final): Rename to...
29639 (final_1): ... this. Take SEEN parameter. Output final pending
29640 next view at the end.
29641 (final): Reintroduce seen-less interface.
29642 (final_scan_insn): Output pending next view before switching
29643 sections or ending a block. Mark the next view as needed when
29644 outputting variable locations. Notify debug backend of section
29645 changes, and of location view changes.
29646 (rest_of_handle_final): Adjust.
29647 * toplev.c (process_options): Autodetect value for debug variable
29648 location views option. Warn on incompat5 without -gdwarf-5.
29649 * doc/invoke.texi (gvariable-location-views): New.
29650 (gvariable-location-views=incompat5): New.
29651 (gno-variable-location-views): New.
29653 2018-02-08 David Malcolm <dmalcolm@redhat.com>
29655 PR tree-optimization/84136
29656 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
29657 that the result of find_edge is non-NULL.
29659 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
29662 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
29663 storing integer register in SImode. Fix cost of 256 and 512
29664 byte aligned SSE register store.
29666 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
29668 * config/i386/i386.c (ix86_multiplication_cost): Fix
29669 multiplication cost for TARGET_AVX512DQ.
29671 2018-02-08 Marek Polacek <polacek@redhat.com>
29673 PR tree-optimization/84238
29674 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
29677 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
29679 PR tree-optimization/84265
29680 * tree-vect-stmts.c (vectorizable_store): Don't treat
29681 VMAT_CONTIGUOUS accesses as grouped.
29682 (vectorizable_load): Likewise.
29684 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
29686 PR tree-optimization/81635
29687 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
29688 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
29689 (test_round_for_mask): New functions.
29690 (wide_int_cc_tests): Call test_round_for_mask.
29691 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
29692 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
29693 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
29694 range returned by get_range_info.
29696 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
29699 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
29700 * symtab.c: Include builtins.h
29701 (symtab_node::output_to_lto_symbol_table_p): Move here
29702 from lto-streamer-out.c:output_symbol_p.
29703 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
29704 (output_symbol_p): Move all logic to symtab.c
29705 (produce_symtab): Update.
29707 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
29709 * config/s390/s390-opts.h (enum indirect_branch): Define.
29710 * config/s390/s390-protos.h (s390_return_addr_from_memory)
29711 (s390_indirect_branch_via_thunk)
29712 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
29713 (enum s390_indirect_branch_type): Define.
29714 * config/s390/s390.c (struct s390_frame_layout, struct
29715 machine_function): Remove.
29716 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
29717 (indirect_branch_table_label_no, indirect_branch_table_name):
29719 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
29720 (enum s390_indirect_branch_option): Define.
29721 (s390_return_addr_from_memory): New function.
29722 (s390_handle_string_attribute): New function.
29723 (s390_attribute_table): Add new attribute handler.
29724 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
29725 (s390_indirect_branch_via_thunk): New function.
29726 (s390_indirect_branch_via_inline_thunk): New function.
29727 (s390_function_ok_for_sibcall): When jumping via thunk disallow
29728 sibling call optimization for non z10 compiles.
29729 (s390_emit_call): Force indirect branch target to be a single
29730 register. Add r1 clobber for non-z10 compiles.
29731 (s390_emit_epilogue): Emit return jump via return_use expander.
29732 (s390_reorg): Handle JUMP_INSNs as execute targets.
29733 (s390_option_override_internal): Perform validity checks for the
29734 new command line options.
29735 (s390_indirect_branch_attrvalue): New function.
29736 (s390_indirect_branch_settings): New function.
29737 (s390_set_current_function): Invoke s390_indirect_branch_settings.
29738 (s390_output_indirect_thunk_function): New function.
29739 (s390_code_end): Implement target hook.
29740 (s390_case_values_threshold): Implement target hook.
29741 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
29743 * config/s390/s390.h (struct s390_frame_layout)
29744 (struct machine_function): Move here from s390.c.
29745 (TARGET_INDIRECT_BRANCH_NOBP_RET)
29746 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
29747 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
29748 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
29749 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
29750 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
29751 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
29752 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
29753 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
29754 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
29755 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
29756 (mnemonic attribute): Add values which aren't recognized
29758 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
29759 pattern for branch conversion. Fix mnemonic attribute.
29760 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
29761 indirect branch via thunk if requested.
29762 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
29763 ("*indirect_jump"): Disable for branch conversion using out of
29765 ("indirect_jump_via_thunk<mode>_z10")
29766 ("indirect_jump_via_thunk<mode>")
29767 ("indirect_jump_via_inlinethunk<mode>_z10")
29768 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
29769 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
29770 ("casesi_jump_via_inlinethunk<mode>_z10")
29771 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
29772 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
29773 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
29774 ("*indirect2_jump"): Disable for branch conversion.
29775 ("casesi_jump"): Turn into expander and expand patterns for branch
29777 ("return_use"): New expander.
29778 ("*return"): Emit return via thunk and rename it to ...
29779 ("*return<mode>"): ... this one.
29780 * config/s390/s390.opt: Add new options and and enum for the
29783 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
29785 * lra-constraints.c (match_reload): Unconditionally use
29786 gen_lowpart_SUBREG, rather than selecting between that
29787 and equivalent gen_rtx_SUBREG code.
29789 2018-02-08 Richard Biener <rguenther@suse.de>
29791 PR tree-optimization/84233
29792 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
29793 changed flag instead of boguously re-using phi_inserted.
29795 2018-02-08 Martin Jambor <mjambor@suse.cz>
29797 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
29798 static local variables.
29800 2018-02-08 Richard Biener <rguenther@suse.de>
29802 PR tree-optimization/84278
29803 * tree-vect-stmts.c (vectorizable_store): When looking for
29804 smaller vector types to perform grouped strided loads/stores
29805 make sure the mode is supported by the target.
29806 (vectorizable_load): Likewise.
29808 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
29810 * config/aarch64/aarch64.c (aarch64_components_for_bb):
29811 Increase LDP/STP opportunities by adding adjacent callee-saves.
29813 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
29815 PR rtl-optimization/84068
29816 PR rtl-optimization/83459
29817 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
29819 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
29821 PR tree-optimization/84224
29822 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
29823 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
29824 non-zero arguments.
29826 2018-02-07 Iain Sandoe <iain@codesourcery.com>
29829 * config/rs6000/altivec.md (*restore_world): Remove LR use.
29830 * config/rs6000/predicates.md (restore_world_operation): Adjust op
29831 count, remove one USE.
29833 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
29835 * doc/install.texi (Configuration): Document the
29836 --with-long-double-format={ibm,ieee} PowerPC configuration
29840 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
29841 Convert from define_expand to be define_insn_and_split. Rework
29842 float/double/_Float128 conversions to QI/HI/SImode to work with
29843 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
29844 conversions to QI/HImode types did a store and then a load to
29845 truncate the value. For conversions to VSX registers, don't split
29846 the insn, instead emit the code directly. Use the code iterator
29847 any_fix to combine signed and unsigned conversions.
29848 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
29849 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
29850 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
29851 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
29852 (fix_<mode>di2_hw): Likewise.
29853 (fixuns_<mode>di2_hw): Likewise.
29854 (fix_<mode>si2_hw): Likewise.
29855 (fixuns_<mode>si2_hw): Likewise.
29856 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
29857 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
29858 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
29859 fix<uns>_trunc<SFDF:mode>si2_p8.
29860 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
29861 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
29862 (fix<uns>_<mode>_mem): Likewise.
29863 (fctiw<u>z_<mode>_mem): Likewise.
29864 (fix<uns>_<mode>_mem): Likewise.
29865 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
29866 the register allocator from doing a direct move to the GPRs to do
29867 a store, and instead use the ISA 3.0 store byte/half-word from
29868 vector register instruction. For IEEE 128-bit floating point,
29869 also optimize stores of 32-bit ints.
29870 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
29872 2018-02-07 Alan Hayward <alan.hayward@arm.com>
29874 * genextract.c (push_pathstr_operand): New function to support
29876 (walk_rtx): Call push_pathstr_operand.
29877 (print_path): Support [a-zA-Z].
29879 2018-02-07 Richard Biener <rguenther@suse.de>
29881 PR tree-optimization/84037
29882 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
29883 (cse_and_gimplify_to_preheader): Declare.
29884 (vect_get_place_in_interleaving_chain): Likewise.
29885 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
29887 (_loop_vec_info::~_loop_vec_info): Delete it.
29888 (cse_and_gimplify_to_preheader): New function.
29889 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
29890 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
29891 (vectorizable_load): Likewise. For grouped stores always base
29892 the IV on the first element.
29893 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
29894 condition before gimplifying.
29896 2018-02-07 Jakub Jelinek <jakub@redhat.com>
29898 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
29899 *DIV_EXPR and *MOD_EXPR.
29901 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
29904 * config/i386/i386.c (ix86_option_override_internal): Mask out
29905 the CF_SET bit when checking -fcf-protection.
29907 2018-02-07 Tom de Vries <tom@codesourcery.com>
29910 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
29913 2018-02-07 Richard Biener <rguenther@suse.de>
29915 PR tree-optimization/84204
29916 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
29919 PR tree-optimization/84205
29920 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
29921 special-case isl_ast_op_zdiv_r.
29923 PR tree-optimization/84223
29924 * graphite-scop-detection.c (gather_bbs::before_dom_children):
29925 Only add conditions from within the region.
29926 (gather_bbs::after_dom_children): Adjust.
29928 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
29931 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
29932 * config/avr/avr.md: Only post-reload split REG-REG moves if
29933 either register is GENERAL_REG_P.
29935 2018-02-07 Jakub Jelinek <jakub@redhat.com>
29937 PR tree-optimization/84235
29938 * tree-ssa-scopedtables.c
29939 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
29940 if the subtraction is performed in floating point type where NaNs are
29941 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
29942 build 1. Formatting fix.
29944 2018-02-06 Jakub Jelinek <jakub@redhat.com>
29947 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
29948 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
29949 and skip it regardless of bb boundaries. Use CALL_P macro,
29950 don't test INSN_P (insn) together with CALL_P or JUMP_P check
29951 unnecessarily, formatting fix.
29953 2018-02-06 Michael Collison <michael.collison@arm.com>
29955 * config/arm/thumb2.md:
29956 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
29957 (*thumb_mov_notscc): Ditto.
29959 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
29962 * config/rs6000/rs6000.md (su code attribute): Use "u" for
29963 unsigned_fix, not "s".
29965 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29967 * configure.ac (gcc_fn_eh_frame_ro): New function.
29968 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
29969 correct .eh_frame permissions.
29970 * configure: Regenerate.
29972 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
29974 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
29975 irrelevant options.
29977 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29979 * config/rs6000/rs6000.c (rs6000_option_override_internal):
29980 Display warning message for -mno-speculate-indirect-jumps.
29982 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
29984 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
29986 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
29988 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
29990 PR tree-optimization/84225
29991 * tree-eh.c (find_trapping_overflow): Only call
29992 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
29994 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
29997 * config/i386/i386.c: Reimplement the check of possible options
29998 -mibt/-mshstk conbination. Change error messages.
29999 * doc/invoke.texi: Fix a typo: remove extra '='.
30001 2018-02-06 Marek Polacek <polacek@redhat.com>
30003 PR tree-optimization/84228
30004 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
30006 2018-02-06 Tamar Christina <tamar.christina@arm.com>
30009 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
30010 emitted arch directives.
30011 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
30012 __ARM_FEATURE_COPROC before changing architectures.
30014 2018-02-06 Richard Biener <rguenther@suse.de>
30016 * config/i386/i386.c (print_reg): Fix typo.
30017 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
30019 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
30021 * configure: Regenerate.
30023 2018-02-05 Martin Sebor <msebor@redhat.com>
30025 PR tree-optimization/83369
30026 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
30029 2018-02-05 Martin Liska <mliska@suse.cz>
30031 * doc/invoke.texi: Cherry-pick upstream r323995.
30033 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
30035 * ira.c (ira_init_register_move_cost): Adjust comment.
30037 2018-02-05 Martin Liska <mliska@suse.cz>
30039 PR gcov-profile/84137
30040 * doc/gcov.texi: Fix typo in documentation.
30042 2018-02-05 Martin Liska <mliska@suse.cz>
30044 PR gcov-profile/83879
30045 * doc/gcov.texi: Document necessity of --dynamic-list-data when
30046 using dlopen functionality.
30048 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
30050 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
30051 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
30052 _mm_maskz_range_ss, _mm_mask_range_round_ss,
30053 _mm_maskz_range_round_ss): New intrinsics.
30054 (__builtin_ia32_rangesd128_round)
30055 (__builtin_ia32_rangess128_round): Remove.
30056 (__builtin_ia32_rangesd128_mask_round,
30057 __builtin_ia32_rangess128_mask_round): New builtins.
30058 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
30059 __builtin_ia32_rangess128_round): Remove.
30060 (__builtin_ia32_rangesd128_mask_round,
30061 __builtin_ia32_rangess128_mask_round): New builtins.
30062 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
30063 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
30064 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
30065 "<round_saeonly_constraint>")): Changed to ...
30066 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
30067 "<round_saeonly_scalar_constraint>")): ... this.
30068 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
30069 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
30070 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
30071 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
30072 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
30074 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
30076 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
30078 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
30079 Remove all values except native, 8540 and 8548.
30081 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
30083 * config/i386/i386.c (ix86_output_function_return): Pass
30084 INVALID_REGNUM, instead of -1, as invalid register number to
30085 indirect_thunk_name and output_indirect_thunk.
30087 2018-02-02 Julia Koval <julia.koval@intel.com>
30089 * config.gcc: Add -march=icelake.
30090 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
30091 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
30092 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
30093 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
30094 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
30095 (processor_target_table): Add icelake.
30096 (ix86_option_override_internal): Handle new PTAs.
30097 (get_builtin_code_for_version): Handle icelake.
30098 (M_INTEL_COREI7_ICELAKE): New.
30099 (fold_builtin_cpu): Handle icelake.
30100 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
30101 * doc/invoke.texi: Add -march=icelake.
30103 2018-02-02 Julia Koval <julia.koval@intel.com>
30105 * config/i386/i386.c (ix86_option_override_internal): Change flags type
30106 to wide_int_bitmask.
30107 * wide-int-bitmask.h: New.
30109 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
30112 * config/i386/i386.md: Replace Pmode with word_mode in
30113 builtin_setjmp_setup and builtin_longjmp to support x32.
30115 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
30119 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
30121 (rs6000_supported_cpu_names): New static variable.
30122 (linux_cpu_translation_table): Likewise.
30123 (elf_platform) <cpu>: Define new static variable and use it.
30124 Translate kernel AT_PLATFORM name to canonical name if needed.
30125 Error if platform name is unknown.
30127 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
30130 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
30132 2018-02-01 Jeff Law <law@redhat.com>
30135 * config/i386/i386.c (release_scratch_register_on_entry): Add new
30136 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
30137 the scratch if RELEASE_VIA_POP is false.
30138 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
30139 If we have to save a temporary register, decrement SIZE appropriately.
30140 Pass new arguments to release_scratch_register_on_entry.
30141 (ix86_adjust_stack_and_probe): Likewise.
30142 (ix86_emit_probe_stack_range): Pass new arguments to
30143 release_scratch_register_on_entry.
30145 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
30147 PR rtl-optimization/84157
30148 * combine.c (change_zero_ext): Use REG_P predicate in
30149 front of HARD_REGISTER_P predicate.
30151 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
30153 * config/avr/avr.c (avr_option_override): Move disabling of
30154 -fdelete-null-pointer-checks to...
30155 * common/config/avr/avr-common.c (avr_option_optimization_table):
30158 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30160 PR tree-optimization/81635
30161 * tree-data-ref.c (split_constant_offset_1): For types that
30162 wrap on overflow, try to use range info to prove that wrapping
30165 2018-02-01 Renlin Li <renlin.li@arm.com>
30168 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
30169 TAILCALL_ADDR_REGS.
30170 (aarch64_register_move_cost): Likewise.
30171 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
30172 TAILCALL_ADDR_REGS.
30173 (REG_CLASS_NAMES): Likewise.
30174 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
30175 TAILCALL_ADDR_REGS. Remove IP registers.
30176 * config/aarch64/aarch64.md (Ucs): Update register constraint.
30178 2018-02-01 Richard Biener <rguenther@suse.de>
30180 * domwalk.h (dom_walker::dom_walker): Add additional constructor
30181 for specifying RPO order and allow NULL for that.
30182 * domwalk.c (dom_walker::dom_walker): Likewise.
30183 (dom_walker::walk): Handle NULL RPO order.
30184 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
30186 (rewrite_update_dom_walker): Likewise.
30187 (mark_def_dom_walker): Likewise.
30189 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30191 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
30192 (aarch64_maybe_expand_sve_subreg_move): Declare.
30193 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
30194 * config/aarch64/predicates.md (aarch64_any_register_operand): New
30196 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
30197 that are semantically a reverse operation.
30198 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
30199 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
30200 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
30202 (aarch64_can_change_mode_class): For big-endian, forbid changes
30203 between two SVE modes if they have different element sizes.
30205 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30207 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
30208 the TImode handling for big-endian targets.
30210 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30212 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
30213 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
30215 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
30216 Remove BSWAP handing for big-endian targets and use the form of
30217 LD1RQ appropariate for the mode.
30219 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30221 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
30222 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
30223 duplicated element.
30225 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30228 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
30229 check for operands that need to go through aarch64_sve_reload_be.
30231 2018-02-01 Jakub Jelinek <jakub@redhat.com>
30233 PR tree-optimization/81661
30234 PR tree-optimization/84117
30235 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
30236 * tree-eh.c: Include gimplify.h.
30237 (find_trapping_overflow, replace_trapping_overflow,
30238 rewrite_to_non_trapping_overflow): New functions.
30239 * tree-vect-loop.c: Include tree-eh.h.
30240 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
30241 * tree-data-ref.c: Include tree-eh.h.
30242 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
30244 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
30246 PR rtl-optimization/84123
30247 * combine.c (change_zero_ext): Check if hard register satisfies
30248 can_change_dest_mode before calling gen_lowpart_SUBREG.
30250 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
30253 * ira.c (ira_init_register_move_cost): Remove assert.
30255 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
30257 PR rtl-optimization/84071
30258 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
30259 * doc/tm.texi: Regenerate.
30261 2018-01-31 Richard Biener <rguenther@suse.de>
30263 PR tree-optimization/84132
30264 * tree-data-ref.c (analyze_miv_subscript): Properly
30265 check whether evolution_function_is_affine_multivariate_p
30266 before calling gcd_of_steps_may_divide_p.
30268 2018-01-31 Julia Koval <julia.koval@intel.com>
30271 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
30272 * config/i386/i386.md (rdpid_rex64) New.
30273 (rdpid): Make 32bit only.
30275 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
30278 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
30279 an IDENTIFIER_NODE for FUNCTION_TYPE's.
30281 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
30284 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
30286 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
30288 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
30290 PR rtl-optimization/84071
30291 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
30292 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
30294 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
30296 * config/arc/arc.c (arc_handle_aux_attribute): New function.
30297 (arc_attribute_table): Add 'aux' attribute.
30298 (arc_in_small_data_p): Consider aux like variables.
30299 (arc_is_aux_reg_p): New function.
30300 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
30301 (arc_get_aux_arg): New function.
30302 (prepare_move_operands): Handle aux-register access.
30303 (arc_handle_aux_attribute): New function.
30304 * doc/extend.texi (ARC Variable attributes): Add subsection.
30306 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
30308 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
30309 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
30310 (arc_attribute_table): Add 'uncached' attribute.
30311 (arc_print_operand): Print '.di' flag for uncached memory
30313 (arc_in_small_data_p): Do not consider for small data the uncached
30315 (arc_is_uncached_mem_p): New function.
30316 * config/arc/predicates.md (compact_store_memory_operand): Check
30317 for uncached memory accesses.
30318 (nonvol_nonimm_operand): Likewise.
30319 * doc/extend.texi (ARC Type Attribute): New subsection.
30321 2018-01-31 Jakub Jelinek <jakub@redhat.com>
30324 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
30325 falign-loops=): Add Optimization flag.
30327 2018-01-30 Jeff Law <law@redhat.com>
30330 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
30331 INT_REGISTERS_SAVED. Check it prior to calling
30332 get_scratch_register_on_entry.
30333 (ix86_adjust_stack_and_probe): Similarly.
30334 (ix86_emit_probe_stack_range): Similarly.
30335 (ix86_expand_prologue): Corresponding changes.
30337 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30340 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
30341 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
30343 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
30346 * lra-constraints.c (curr_insn_transform): Process AND in the
30349 2018-01-30 Jakub Jelinek <jakub@redhat.com>
30351 PR rtl-optimization/83986
30352 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
30353 dependence against last_pending_memory_flush in addition to
30354 pending_jump_insns.
30356 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
30358 PR tree-optimization/81611
30359 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
30362 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30365 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
30368 2018-01-30 Richard Biener <rguenther@suse.de>
30369 Jakub Jelinek <jakub@redhat.com>
30371 PR tree-optimization/84111
30372 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
30373 inner loops added during recursion, as they don't have up-to-date
30376 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
30379 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
30380 (can_inline_edge_by_limits_p): ... here.
30381 (can_early_inline_edge_p, check_callers,
30382 update_caller_keys, update_callee_keys, recursive_inlining,
30383 add_new_edges_to_heap, speculation_useful_p,
30384 inline_small_functions,
30385 inline_small_functions, flatten_function,
30386 inline_to_all_callers_1): Update.
30388 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
30390 * profile-count.c (profile_count::combine_with_ipa_count): Handle
30393 2018-01-30 Richard Biener <rguenther@suse.de>
30395 PR tree-optimization/83008
30396 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
30397 invariant and constant vector uses in stmts when they need
30398 more than one stmt.
30400 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30403 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
30404 * configure: Regenerate.
30406 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
30408 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
30410 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
30411 Use gen_rtx_REG rather than gen_lowpart.
30413 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
30415 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
30416 rather than 0 when creating partial subregs.
30418 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
30420 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
30423 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
30426 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
30427 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
30428 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
30429 flags. This restores the settings used before the 2017-07-24.
30430 Turning off pre increment/decrement/modify allows IVOPTS to
30431 optimize DF/SF loops where the index is an int.
30433 2018-01-29 Richard Biener <rguenther@suse.de>
30434 Kelvin Nilsen <kelvin@gcc.gnu.org>
30437 * tree-vect-stmts.c (vectorizable_call): Don't call
30438 targetm.vectorize_builtin_md_vectorized_function if callee is
30441 2018-01-22 Carl Love <cel@us.ibm.com>
30443 * doc/extend.tex: Fix typo in second arg in
30444 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
30446 2018-01-29 Richard Biener <rguenther@suse.de>
30448 PR tree-optimization/84086
30449 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
30450 (flush_ssaname_freelist): When SSA names were released reset
30451 the SCEV hash table.
30453 2018-01-29 Richard Biener <rguenther@suse.de>
30455 PR tree-optimization/84057
30456 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
30457 removed paths when removing edges.
30459 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
30461 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
30462 -mfunction-return=@var{choice}.
30464 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
30466 PR diagnostic/84034
30467 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
30468 Handle CR like TAB.
30469 (layout::print_source_line): Likewise.
30470 (test_get_line_width_without_trailing_whitespace): Add test cases.
30472 2018-01-27 Jakub Jelinek <jakub@redhat.com>
30474 PR middle-end/84040
30475 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
30478 2018-01-26 Jim Wilson <jimw@sifive.com>
30480 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
30482 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
30485 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30487 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
30488 and CMP + SUB-immediate -> SUBS.
30490 2018-01-26 Martin Sebor <msebor@redhat.com>
30492 PR tree-optimization/83896
30493 * tree-ssa-strlen.c (get_string_len): Rename...
30494 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
30495 Avoid assuming length is constant.
30496 (handle_char_store): Use HOST_WIDE_INT for string length.
30498 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
30501 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
30502 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
30504 2018-01-26 Richard Biener <rguenther@suse.de>
30506 PR rtl-optimization/84003
30507 * dse.c (record_store): Only record redundant stores when
30508 the earlier store aliases at least all accesses the later one does.
30510 2018-01-26 Jakub Jelinek <jakub@redhat.com>
30512 PR rtl-optimization/83985
30513 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
30514 REG_CFA_RESTORE insns.
30515 (delete_unmarked_insns): Don't ignore separate shrink wrapping
30516 REG_CFA_RESTORE insns here.
30519 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
30520 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
30522 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30524 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
30525 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
30526 (arc_init): Likewise.
30527 (arc_override_options): Likewise.
30528 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
30530 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
30532 * config/arc/arc.h (TARGET_DBNZ): Define.
30533 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
30534 properly set the tune attribute.
30535 (dbnz): Use TARGET_DBNZ guard.
30536 * config/arc/arc.opt (mtune): Add core3 option.
30538 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30540 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
30541 recognize new pic like addresses.
30542 (arc_delegitimize_address): Clean up.
30544 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30546 * config/arc/arc-arches.def: Option mrf16 valid for all
30548 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
30549 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
30550 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
30551 * config/arc/arc-tables.opt: Regenerate.
30552 * config/arc/arc.c (arc_conditional_register_usage): Handle
30553 reduced register file case.
30554 (arc_file_start): Set must have build attributes.
30555 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
30556 mrf16 option value.
30557 * config/arc/arc.opt (mrf16): Add new option.
30558 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
30559 * config/arc/genmultilib.awk: Handle new mrf16 option.
30560 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
30561 * config/arc/t-multilib: Regenerate.
30562 * doc/invoke.texi (ARC Options): Document mrf16 option.
30564 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30566 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
30567 * config/arc/arc.c (arc_handle_secure_attribute): New function.
30568 (arc_attribute_table): Add 'secure_call' attribute.
30569 (arc_print_operand): Print secure call operand.
30570 (arc_function_ok_for_sibcall): Don't optimize tail calls when
30572 (arc_is_secure_call_p): New function. * config/arc/arc.md
30573 (call_i): Add support for sjli instruction.
30574 (call_value_i): Likewise.
30575 * config/arc/constraints.md (Csc): New constraint.
30577 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30578 John Eric Martin <John.Martin@emmicro-us.com>
30580 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
30581 * config/arc/arc.c (_arc_jli_section): New struct.
30582 (arc_jli_section): New type.
30583 (rc_jli_sections): New static variable.
30584 (arc_handle_jli_attribute): New function.
30585 (arc_attribute_table): Add jli_always and jli_fixed attribute.
30586 (arc_file_end): New function.
30587 (TARGET_ASM_FILE_END): Define.
30588 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
30589 (arc_add_jli_section): New function.
30590 (jli_call_scan): Likewise.
30591 (arc_reorg): Call jli_call_scan.
30592 (arc_output_addsi): Remove 'S' from printing asm operand.
30593 (arc_is_jli_call_p): New function.
30594 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
30596 (movhi_insn): Likewise.
30597 (movsi_insn): Likewise.
30598 (movsi_set_cc_insn): Likewise.
30599 (loadqi_update): Likewise.
30600 (load_zeroextendqisi_update): Likewise.
30601 (load_signextendqisi_update): Likewise.
30602 (loadhi_update): Likewise.
30603 (load_zeroextendhisi_update): Likewise.
30604 (load_signextendhisi_update): Likewise.
30605 (loadsi_update): Likewise.
30606 (loadsf_update): Likewise.
30607 (movsicc_insn): Likewise.
30608 (bset_insn): Likewise.
30609 (bxor_insn): Likewise.
30610 (bclr_insn): Likewise.
30611 (bmsk_insn): Likewise.
30612 (bicsi3_insn): Likewise.
30613 (cmpsi_cc_c_insn): Likewise.
30614 (movsi_ne): Likewise.
30615 (movsi_cond_exec): Likewise.
30616 (clrsbsi2): Likewise.
30617 (norm_f): Likewise.
30627 (call_i): Remove 'S' asm letter, add jli instruction.
30628 (call_value_i): Likewise.
30629 * config/arc/arc.op (mjli-always): New option.
30630 * config/arc/constraints.md (Cji): New constraint.
30631 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
30633 (subsf3_fpx): Likewise.
30634 (mulsf3_fpx): Likewise.
30635 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
30637 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
30638 function attrbutes.
30639 * doc/invoke.texi (ARC): Document mjli-always option.
30641 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
30643 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
30644 avoid addition with 0 and use incw and decw where possible.
30646 2018-01-26 Richard Biener <rguenther@suse.de>
30648 PR tree-optimization/81082
30649 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
30650 association if it requires casting to unsigned.
30651 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
30652 from fold_plusminus_mult_expr to catch important cases late when
30653 range info is available.
30655 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30657 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
30658 * configure.ac (hidden_linkonce): New test.
30659 * configure: Regenerate.
30660 * config.in: Regenerate.
30662 2018-01-26 Julia Koval <julia.koval@intel.com>
30664 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
30665 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
30666 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
30667 _mm_mask_bitshuffle_epi64_mask): Fix type.
30668 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
30669 USI_FTYPE_V4DI_V4DI_USI): Remove.
30670 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
30671 __builtin_ia32_vpshufbitqmb256_mask,
30672 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
30673 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
30674 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
30676 2018-01-26 Alan Modra <amodra@gmail.com>
30679 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
30680 UNSPEC_VBPERMQ. Sort other unspecs.
30682 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
30684 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
30686 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
30688 PR middle-end/83055
30689 * predict.c (drop_profile): Do not push/pop cfun; update also
30691 (handle_missing_profiles): Fix logic looking for zero profiles.
30693 2018-01-25 Jakub Jelinek <jakub@redhat.com>
30695 PR middle-end/83977
30696 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
30697 on functions with #pragma omp declare simd or functions with simd
30699 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
30700 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
30701 Remove trailing \n from warning_at calls.
30703 2018-01-25 Tom de Vries <tom@codesourcery.com>
30706 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
30707 for neutered workers.
30709 2018-01-24 Joseph Myers <joseph@codesourcery.com>
30712 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
30713 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
30715 2018-01-24 Jeff Law <law@redhat.com>
30718 * i386.c (get_probe_interval): Move to earlier point.
30719 (ix86_compute_frame_layout): If -fstack-clash-protection and
30720 the frame is larger than the probe interval, then use pushes
30721 to save registers rather than reg->mem moves.
30722 (ix86_expand_prologue): Remove conditional for int_registers_saved
30725 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
30728 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
30729 min/max for never referenced object.
30731 2018-01-24 Jakub Jelinek <jakub@redhat.com>
30733 PR middle-end/83977
30734 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
30736 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
30737 attributes from DECL_ATTRIBUTES (decl) without affecting
30738 DECL_ATTRIBUTES (current_function_decl).
30739 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
30740 functions with non-NULL DECL_ABSTRACT_ORIGIN.
30742 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
30744 PR tree-optimization/83979
30745 * fold-const.c (fold_comparison): Use constant_boolean_node
30746 instead of boolean_{true,false}_node.
30748 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
30750 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
30753 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
30755 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30756 Simplify the clause that sets the length attribute.
30757 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30758 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
30759 clause that sets the length attribute.
30760 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30762 2018-01-24 Tom de Vries <tom@codesourcery.com>
30765 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
30766 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
30767 Add strict parameter.
30768 (prevent_branch_around_nothing): Insert dummy insn between branch to
30769 label and label with no ptx insn inbetween.
30770 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
30772 2018-01-24 Tom de Vries <tom@codesourcery.com>
30775 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
30776 for neutered threads in warp.
30777 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
30779 2018-01-24 Richard Biener <rguenther@suse.de>
30781 PR tree-optimization/83176
30782 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
30785 2018-01-24 Richard Biener <rguenther@suse.de>
30787 PR tree-optimization/82819
30788 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
30789 code generating pluses that are no-ops in the target precision.
30791 2018-01-24 Richard Biener <rguenther@suse.de>
30793 PR middle-end/84000
30794 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
30796 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
30798 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
30799 to merge probabilities.
30800 * predict.c (probably_never_executed): Also mark as cold functions
30801 with global 0 profile and guessed local profile.
30802 * profile-count.c (profile_probability::combine_with_count): New
30804 * profile-count.h (profile_probability::operator*,
30805 profile_probability::operator*=, profile_probability::operator/,
30806 profile_probability::operator/=): Reduce precision to adjusted
30807 and set value to guessed on contradictory divisions.
30808 (profile_probability::combine_with_freq): Remove.
30809 (profile_probability::combine_wiht_count): Declare.
30810 (profile_count::force_nonzero):: Set to adjusted.
30811 (profile_count::probability_in):: Set quality to adjusted.
30812 * tree-ssa-tail-merge.c (replace_block_by): Use
30813 combine_with_count.
30815 2018-01-23 Andrew Waterman <andrew@sifive.com>
30816 Jim Wilson <jimw@sifive.com>
30818 * config/riscv/riscv.c (riscv_stack_boundary): New.
30819 (riscv_option_override): Set riscv_stack_boundary. Handle
30820 riscv_preferred_stack_boundary_arg.
30821 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
30822 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
30823 (STACK_BOUNDARY): Set to riscv_stack_boundary.
30824 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
30825 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
30826 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
30828 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
30831 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
30832 of struct ix86_frame.
30833 (ix86_expand_epilogue): Likewise. Add a local variable for
30834 the reg_save_offset field in struct ix86_frame.
30836 2018-01-23 Bin Cheng <bin.cheng@arm.com>
30838 PR tree-optimization/82604
30839 * tree-loop-distribution.c (enum partition_kind): New enum item
30840 PKIND_PARTIAL_MEMSET.
30841 (partition_builtin_p): Support above new enum item.
30842 (generate_code_for_partition): Ditto.
30843 (compute_access_range): Differentiate cases that equality can be
30844 proven at all loops, the innermost loops or no loops.
30845 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
30846 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
30847 (finalize_partitions, distribute_loop): Don't fuse partition of
30848 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
30849 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
30850 parloop is enabled.
30852 2018-01-23 Martin Liska <mliska@suse.cz>
30854 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
30855 order to ignore the predictor.
30856 (PRED_POLYMORPHIC_CALL): Likewise.
30857 (PRED_RECURSIVE_CALL): Likewise.
30859 2018-01-23 Martin Liska <mliska@suse.cz>
30861 * tree-profile.c (tree_profiling): Print function header to
30862 aware reader which function we are working on.
30863 * value-prof.c (gimple_find_values_to_profile): Do not print
30864 not interesting value histograms.
30866 2018-01-23 Martin Liska <mliska@suse.cz>
30868 * profile-count.h (enum profile_quality): Add
30869 profile_uninitialized as the first value. Do not number values
30870 as they are zero based.
30871 (profile_count::verify): Update sanity check.
30872 (profile_probability::verify): Likewise.
30874 2018-01-23 Nathan Sidwell <nathan@acm.org>
30876 * doc/invoke.texi (ffor-scope): Deprecate.
30878 2018-01-23 David Malcolm <dmalcolm@redhat.com>
30880 PR tree-optimization/83510
30881 * domwalk.c (set_all_edges_as_executable): New function.
30882 (dom_walker::dom_walker): Convert bool param
30883 "skip_unreachable_blocks" to enum reachability. Move setup of
30884 edge flags to set_all_edges_as_executable and only do it when
30885 reachability is REACHABLE_BLOCKS.
30886 * domwalk.h (enum dom_walker::reachability): New enum.
30887 (dom_walker::dom_walker): Convert bool param
30888 "skip_unreachable_blocks" to enum reachability.
30889 (set_all_edges_as_executable): New decl.
30890 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
30891 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
30893 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
30894 but converting true to REACHABLE_BLOCKS.
30895 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
30897 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
30898 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
30899 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
30901 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
30902 if check_all_array_refs will be called.
30904 2018-01-23 David Malcolm <dmalcolm@redhat.com>
30906 * tree.c (selftest::test_location_wrappers): Add more test
30909 2018-01-23 David Malcolm <dmalcolm@redhat.com>
30911 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
30912 (selftest::test_bit_in_range): Likewise.
30914 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
30917 * doc/sourcebuild.texi (vect_float): Say that the selector
30918 only describes the situation when -funsafe-math-optimizations is on.
30919 (vect_float_strict): Document.
30921 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
30923 PR tree-optimization/83965
30924 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
30925 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
30926 instead of checking only for a reduction.
30927 (vect_recog_widen_sum_pattern): Likewise.
30929 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
30931 * predict.c (probably_never_executed): Only use precise profile info.
30932 (compute_function_frequency): Skip after inlining hack since we now
30933 have quality checking.
30935 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
30937 * profile-count.h (profile_probability::very_unlikely,
30938 profile_probability::unlikely, profile_probability::even): Set
30939 precision to guessed.
30941 2018-01-23 Richard Biener <rguenther@suse.de>
30943 PR tree-optimization/83963
30944 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
30945 Properly terminate dominator walk when crossing the exit edge not
30946 when visiting its source block.
30948 2018-01-23 Jakub Jelinek <jakub@redhat.com>
30951 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
30952 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
30954 2018-01-22 Jakub Jelinek <jakub@redhat.com>
30956 PR tree-optimization/83957
30957 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
30958 semicolon after for body surrounded by braces.
30960 PR tree-optimization/83081
30961 * profile-count.h (profile_probability::split): New method.
30962 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
30963 Use profile_probability::split.
30964 (do_compare_rtx_and_jump): Fix adjustment of probabilities
30965 when splitting a single conditional jump into 2.
30967 2018-01-22 David Malcolm <dmalcolm@redhat.com>
30969 PR tree-optimization/69452
30970 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
30973 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30975 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
30976 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
30977 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
30979 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30981 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
30983 * config/rl78/rl78.md (movdi): New define_expand.
30984 * config/rl78/rl78.c (rl78_split_movdi): New function.
30986 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
30989 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
30991 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
30992 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
30993 128-bit to produce an UNSPEC move to get the double word with the
30994 signbit and then a shift directly to do signbit.
30995 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
30996 implementation with a new version that just does either a direct
30997 move or a regular move. Move memory interface to separate insns.
30998 Move insns so they are next to the expander.
30999 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
31000 with signbit move. Split big and little endian case.
31001 (signbit<mode>2_dm_mem_le): Likewise.
31002 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
31003 (signbit<mode>2_dm2): Likewise.
31005 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
31007 * config/rl78/rl78.md (anddi3): New define_expand.
31009 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
31011 * config/rl78/rl78.md (umindi3): New define_expand.
31013 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
31015 * config/rl78/rl78.md (smindi3): New define_expand.
31017 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
31019 * config/rl78/rl78.md (smaxdi3): New define_expand.
31021 2018-01-22 Carl Love <cel@us.ibm.com>
31023 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
31024 LVX_V1TI): Add macro expansion.
31025 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
31026 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
31027 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
31028 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
31029 Change check to determine if the instruction is a byte reversing
31030 entry. Fix typo in comment.
31031 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
31032 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
31033 Add def_builtin calls for new builtins.
31034 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
31035 Add define_insn expansion.
31037 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
31039 * config/rl78/rl78.md (umaxdi3): New define_expand.
31041 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
31043 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
31044 for non-QImode registers.
31046 2018-01-22 Richard Biener <rguenther@suse.de>
31048 PR tree-optimization/83963
31049 * graphite-scop-detection.c (scop_detection::get_sese): Delay
31050 including the loop exit block.
31051 (scop_detection::merge_sese): Likewise.
31052 (scop_detection::add_scop): Do it here instead.
31054 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31056 * doc/sourcebuild.texi (arm_softfloat): Document.
31058 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
31061 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
31062 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
31063 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
31065 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31066 David Edelsohn <dje.gcc@gmail.com>
31069 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
31070 Change "crset eq" to "crset 2".
31071 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
31072 (*call_indirect_aix<mode>_nospec): Likewise.
31073 (*call_value_indirect_aix<mode>_nospec): Likewise.
31074 (*call_indirect_elfv2<mode>_nospec): Likewise.
31075 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
31076 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
31077 change assembly output from . to $.
31078 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
31079 (indirect_jump<mode>_nospec): Change assembly output from . to $.
31080 (*tablejump<mode>_internal1_nospec): Likewise.
31082 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
31085 * config/sh/sh_optimize_sett_clrt.cc:
31086 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
31088 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
31090 PR tree-optimization/83940
31091 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
31092 offset_dt to vect_constant_def rather than vect_unknown_def_type.
31093 (vect_check_load_store_mask): Add a mask_dt_out parameter and
31094 use it to pass back the definition type.
31095 (vect_check_store_rhs): Likewise rhs_dt_out.
31096 (vect_build_gather_load_calls): Add a mask_dt argument and use
31097 it instead of a call to vect_is_simple_use.
31098 (vectorizable_store): Update calls to vect_check_load_store_mask
31099 and vect_check_store_rhs. Use the dt returned by the latter instead
31100 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
31101 instead of calls to vect_is_simple_use. Pass the scalar rather
31102 than the vector operand to vect_is_simple_use when handling
31103 second and subsequent copies of an rhs value.
31104 (vectorizable_load): Update calls to vect_check_load_store_mask
31105 and vect_build_gather_load_calls. Use the cached mask_dt and
31106 gs_info.offset_dt instead of calls to vect_is_simple_use.
31108 2018-01-20 Jakub Jelinek <jakub@redhat.com>
31110 PR middle-end/83945
31111 * tree-emutls.c: Include gimplify.h.
31112 (lower_emutls_2): New function.
31113 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
31114 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
31115 it before further processing.
31118 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
31119 UINTVAL (trueop1) instead of INTVAL (op1).
31121 2018-01-19 Jakub Jelinek <jakub@redhat.com>
31125 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
31126 INCOMING_FRAME_SP_OFFSET if not defined.
31127 (scan_trace): Add ENTRY argument. If true and
31128 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
31129 emit a note to adjust the CFA offset.
31130 (create_cfi_notes): Adjust scan_trace callers.
31131 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
31132 INCOMING_FRAME_SP_OFFSET in the CIE.
31133 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
31134 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
31136 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
31137 * doc/tm.texi: Regenerated.
31139 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
31141 PR rtl-optimization/83147
31142 * lra-constraints.c (remove_inheritance_pseudos): Use
31143 lra_substitute_pseudo_within_insn.
31145 2018-01-19 Tom de Vries <tom@codesourcery.com>
31146 Cesar Philippidis <cesar@codesourcery.com>
31149 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
31151 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
31154 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
31155 spaces for function labels.
31157 2018-01-19 Martin Liska <mliska@suse.cz>
31159 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
31160 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
31161 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
31162 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
31163 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
31164 (PRED_CONST_RETURN): Change from 69 to 65.
31165 (PRED_NULL_RETURN): Change from 91 to 71.
31166 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
31167 (PRED_LOOP_GUARD): Change from 66 to 73.
31169 2018-01-19 Martin Liska <mliska@suse.cz>
31171 * predict.c (predict_insn_def): Add new assert.
31172 (struct branch_predictor): Change type to signed integer.
31173 (test_prediction_value_range): Amend test to cover
31174 PROB_UNINITIALIZED.
31175 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
31176 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
31177 (PRED_LOOP_ITERATIONS_MAX): Likewise.
31178 (PRED_LOOP_IV_COMPARE): Likewise.
31179 * predict.h (PROB_UNINITIALIZED): Define new constant.
31181 2018-01-19 Martin Liska <mliska@suse.cz>
31183 * predict.c (dump_prediction): Add new format for
31184 analyze_brprob.py script which is enabled with -details
31186 * profile-count.h (precise_p): New function.
31188 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
31190 PR tree-optimization/83922
31191 * tree-vect-loop.c (vect_verify_full_masking): Return false if
31192 there are no statements that need masking.
31193 (vect_active_double_reduction_p): New function.
31194 (vect_analyze_loop_operations): Use it when handling phis that
31195 are not in the loop header.
31197 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
31199 PR tree-optimization/83914
31200 * tree-vect-loop.c (vectorizable_induction): Don't convert
31201 init_expr or apply the peeling adjustment for inductions
31202 that are nested within the vectorized loop.
31204 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31206 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
31209 2018-01-18 Jakub Jelinek <jakub@redhat.com>
31213 * function.h (gimplify_parameters): Add gimple_seq * argument.
31214 * function.c: Include gimple.h and options.h.
31215 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
31216 for the added local temporaries if needed.
31217 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
31218 if there are any parameter cleanups, wrap whole body into a
31219 try/finally with the cleanups.
31221 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
31224 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
31225 Use GET_MODE_CLASS for scalar floating point.
31227 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
31231 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
31232 Fix call of call_cgraph_insertion_hooks.
31234 2018-01-18 Martin Sebor <msebor@redhat.com>
31236 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
31238 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
31241 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
31244 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
31247 * common.opt: (-ffile-prefix-map): New option.
31248 * opts.c (common_handle_option): Defer it.
31249 * opts-global.c (handle_common_deferred_options): Handle it.
31250 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
31251 * file-prefix-map.h: New file.
31252 (remap_debug_filename, add_debug_prefix_map): ...here.
31253 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
31254 * final.c (debug_prefix_map, add_debug_prefix_map
31255 remap_debug_filename): Move to...
31256 * file-prefix-map.c: New file.
31257 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
31258 generalize, get rid of alloca(), use strrchr() instead of strchr().
31259 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
31260 Implement in terms of add_prefix_map().
31261 (remap_macro_filename, remap_debug_filename): Implement in term of
31263 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
31264 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
31265 * dbxout.c: Include file-prefix-map.h.
31266 * varasm.c: Likewise.
31267 * vmsdbgout.c: Likewise.
31268 * xcoffout.c: Likewise.
31269 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
31270 * doc/cppopts.texi (-fmacro-prefix-map): Document.
31271 * doc/invoke.texi (-ffile-prefix-map): Document.
31272 (-fdebug-prefix-map): Update description.
31274 2018-01-18 Martin Liska <mliska@suse.cz>
31276 * config/i386/i386.c (indirect_thunk_name): Document that also
31278 (output_indirect_thunk): Document why both instructions
31279 (pause and lfence) are generated.
31281 2018-01-18 Richard Biener <rguenther@suse.de>
31283 PR tree-optimization/83887
31284 * graphite-scop-detection.c
31285 (scop_detection::get_nearest_dom_with_single_entry): Remove.
31286 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
31287 (scop_detection::merge_sese): Re-implement with a flood-fill
31288 algorithm that properly finds a SESE region if it exists.
31290 2018-01-18 Jakub Jelinek <jakub@redhat.com>
31293 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
31294 pointer_diff optimizations use view_convert instead of convert.
31296 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31298 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
31299 Generate different code for -mno-speculate-indirect-jumps.
31300 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
31301 (*call_indirect_aix<mode>): Disable for
31302 -mno-speculate-indirect-jumps.
31303 (*call_indirect_aix<mode>_nospec): New define_insn.
31304 (*call_value_indirect_aix<mode>): Disable for
31305 -mno-speculate-indirect-jumps.
31306 (*call_value_indirect_aix<mode>_nospec): New define_insn.
31307 (*sibcall_nonlocal_sysv<mode>): Generate different code for
31308 -mno-speculate-indirect-jumps.
31309 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
31311 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
31313 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
31314 long double type, set the flags for noting the default long double
31315 type, even if we don't pass or return a long double type.
31317 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
31320 * ipa-inline.c (flatten_function): Do not overwrite final inlining
31323 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
31325 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
31326 support for merge[hl].
31327 (fold_mergehl_helper): New helper function.
31328 (tree-vector-builder.h): New #include for tree_vector_builder usage.
31329 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
31330 (altivec_vmrglw_direct): Add xxmrglw insn.
31332 2018-01-17 Andrew Waterman <andrew@sifive.com>
31334 * config/riscv/riscv.c (riscv_conditional_register_usage): If
31335 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
31337 2018-01-17 David Malcolm <dmalcolm@redhat.com>
31340 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
31341 call the lto_location_cache before reading the
31342 DECL_SOURCE_LOCATION of the types.
31344 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
31345 Richard Sandiford <richard.sandiford@linaro.org>
31347 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
31348 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
31349 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
31350 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
31351 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
31353 * config/aarch64/constraints.md (aarch64_movti_operand):
31355 * config/aarch64/predicates.md (Uti): Add new constraint.
31357 2018-01-17 Carl Love <cel@us.ibm.com>
31359 * config/rs6000/vsx.md (define_expand xl_len_r,
31360 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
31361 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
31363 (define_expand, define_insn): Move the shift left from the
31364 define_insn to the define_expand for lxvl and stxvl instructions.
31365 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
31366 and XL_LEN_R definitions to PURE.
31368 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
31370 * config/i386/i386.c (indirect_thunk_name): Declare regno
31371 as unsigned int. Compare regno with INVALID_REGNUM.
31372 (output_indirect_thunk): Ditto.
31373 (output_indirect_thunk_function): Ditto.
31374 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
31375 in the call to output_indirect_thunk_function.
31377 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
31379 PR middle-end/83884
31380 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
31381 rather than the size of inner_type to determine the stack slot size
31382 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
31384 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
31387 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
31390 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
31392 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
31393 endian Linux systems to optionally enable multilibs for selecting
31394 the long double type if the user configured an explicit type.
31395 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
31396 have no long double multilibs if not defined.
31397 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
31398 warn if the user used -mabi={ieee,ibm}longdouble and we built
31399 multilibs for long double.
31400 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
31401 appropriate multilib option.
31402 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
31404 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
31405 for building long double multilibs.
31406 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
31408 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
31410 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
31413 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
31415 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
31418 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
31421 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
31424 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
31426 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
31427 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
31429 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
31431 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
31432 different rtl trees depending on TARGET_64BIT.
31433 (rs6000_gen_lvx): Likewise.
31435 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
31437 * config/visium/visium.md (nop): Tweak comment.
31438 (hazard_nop): Likewise.
31440 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31442 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
31443 -mspeculate-indirect-jumps.
31444 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
31445 for -mno-speculate-indirect-jumps.
31446 (*call_indirect_elfv2<mode>_nospec): New define_insn.
31447 (*call_value_indirect_elfv2<mode>): Disable for
31448 -mno-speculate-indirect-jumps.
31449 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
31450 (indirect_jump): Emit different RTL for
31451 -mno-speculate-indirect-jumps.
31452 (*indirect_jump<mode>): Disable for
31453 -mno-speculate-indirect-jumps.
31454 (*indirect_jump<mode>_nospec): New define_insn.
31455 (tablejump): Emit different RTL for
31456 -mno-speculate-indirect-jumps.
31457 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
31458 (tablejumpsi_nospec): New define_expand.
31459 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
31460 (tablejumpdi_nospec): New define_expand.
31461 (*tablejump<mode>_internal1): Disable for
31462 -mno-speculate-indirect-jumps.
31463 (*tablejump<mode>_internal1_nospec): New define_insn.
31464 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
31467 2018-01-16 Artyom Skrobov tyomitch@gmail.com
31469 * caller-save.c (insert_save): Drop unnecessary parameter. All
31472 2018-01-16 Jakub Jelinek <jakub@redhat.com>
31473 Richard Biener <rguenth@suse.de>
31476 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
31477 return early, inline manually is_gimple_sizepos. Make sure if we
31478 call gimplify_expr we don't end up with a gimple constant.
31479 * tree.c (variably_modified_type_p): Don't return true for
31480 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
31481 * gimplify.h (is_gimple_sizepos): Remove.
31483 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31485 PR tree-optimization/83857
31486 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
31487 vectorizable_live_operation for pure SLP statements.
31488 (vectorizable_live_operation): Handle PHIs.
31490 2018-01-16 Richard Biener <rguenther@suse.de>
31492 PR tree-optimization/83867
31493 * tree-vect-stmts.c (vect_transform_stmt): Precompute
31494 nested_in_vect_loop_p since the scalar stmt may get invalidated.
31496 2018-01-16 Jakub Jelinek <jakub@redhat.com>
31499 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
31500 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
31501 If off is not INTEGER_CST, issue a may not be aligned warning
31502 rather than isn't aligned. Use isn%'t rather than isn't.
31503 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
31505 <case MULT_EXPR>: Improve the case when bottom and one of the
31506 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
31507 operand, in that case check if the other operand is multiple of
31508 bottom divided by the INTEGER_CST operand.
31510 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31513 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
31514 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
31515 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
31516 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
31517 * config/pa/pa.c (pa_function_arg_advance): Likewise.
31518 (pa_function_arg, pa_arg_partial_bytes): Likewise.
31519 (pa_function_arg_size): New function.
31521 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31523 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
31524 in a separate statement.
31526 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31528 PR tree-optimization/83847
31529 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
31530 group gathers and scatters.
31532 2018-01-16 Jakub Jelinek <jakub@redhat.com>
31534 PR rtl-optimization/86620
31535 * params.def (max-sched-ready-insns): Bump minimum value to 1.
31537 PR rtl-optimization/83213
31538 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
31539 to last if both are JUMP_INSNs.
31541 PR tree-optimization/83843
31542 * gimple-ssa-store-merging.c
31543 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
31544 store_immediate_info for bswap/nop orig_stores.
31546 2018-01-15 Andrew Waterman <andrew@sifive.com>
31548 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
31550 <UDIV>: Increase cost if !TARGET_DIV.
31552 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
31554 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
31555 (define_attr "cr_logical_3op"): New.
31556 (cceq_ior_compare): Adjust.
31557 (cceq_ior_compare_complement): Adjust.
31558 (*cceq_rev_compare): Adjust.
31559 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
31560 (is_cracked_insn): Adjust.
31561 (insn_must_be_first_in_group): Adjust.
31562 * config/rs6000/40x.md: Adjust.
31563 * config/rs6000/440.md: Adjust.
31564 * config/rs6000/476.md: Adjust.
31565 * config/rs6000/601.md: Adjust.
31566 * config/rs6000/603.md: Adjust.
31567 * config/rs6000/6xx.md: Adjust.
31568 * config/rs6000/7450.md: Adjust.
31569 * config/rs6000/7xx.md: Adjust.
31570 * config/rs6000/8540.md: Adjust.
31571 * config/rs6000/cell.md: Adjust.
31572 * config/rs6000/e300c2c3.md: Adjust.
31573 * config/rs6000/e500mc.md: Adjust.
31574 * config/rs6000/e500mc64.md: Adjust.
31575 * config/rs6000/e5500.md: Adjust.
31576 * config/rs6000/e6500.md: Adjust.
31577 * config/rs6000/mpc.md: Adjust.
31578 * config/rs6000/power4.md: Adjust.
31579 * config/rs6000/power5.md: Adjust.
31580 * config/rs6000/power6.md: Adjust.
31581 * config/rs6000/power7.md: Adjust.
31582 * config/rs6000/power8.md: Adjust.
31583 * config/rs6000/power9.md: Adjust.
31584 * config/rs6000/rs64.md: Adjust.
31585 * config/rs6000/titan.md: Adjust.
31587 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31589 * config/i386/predicates.md (indirect_branch_operand): Rewrite
31590 ix86_indirect_branch_register logic.
31592 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31594 * config/i386/constraints.md (Bs): Update
31595 ix86_indirect_branch_register check. Don't check
31596 ix86_indirect_branch_register with GOT_memory_operand.
31598 * config/i386/predicates.md (GOT_memory_operand): Don't check
31599 ix86_indirect_branch_register here.
31600 (GOT32_symbol_operand): Likewise.
31602 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31604 * config/i386/predicates.md (constant_call_address_operand):
31605 Rewrite ix86_indirect_branch_register logic.
31606 (sibcall_insn_operand): Likewise.
31608 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31610 * config/i386/constraints.md (Bs): Replace
31611 ix86_indirect_branch_thunk_register with
31612 ix86_indirect_branch_register.
31614 * config/i386/i386.md (indirect_jump): Likewise.
31615 (tablejump): Likewise.
31616 (*sibcall_memory): Likewise.
31617 (*sibcall_value_memory): Likewise.
31618 Peepholes of indirect call and jump via memory: Likewise.
31619 * config/i386/i386.opt: Likewise.
31620 * config/i386/predicates.md (indirect_branch_operand): Likewise.
31621 (GOT_memory_operand): Likewise.
31622 (call_insn_operand): Likewise.
31623 (sibcall_insn_operand): Likewise.
31624 (GOT32_symbol_operand): Likewise.
31626 2018-01-15 Jakub Jelinek <jakub@redhat.com>
31628 PR middle-end/83837
31629 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
31630 type rather than type addr's type points to.
31631 (expand_omp_atomic_mutex): Likewise.
31632 (expand_omp_atomic): Likewise.
31634 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31637 * config/i386/i386.c (output_indirect_thunk_function): Use
31638 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
31639 for __x86_return_thunk.
31641 2018-01-15 Richard Biener <rguenther@suse.de>
31643 PR middle-end/83850
31644 * expmed.c (extract_bit_field_1): Fix typo.
31646 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31649 * config/arm/iterators.md (VF): New mode iterator.
31650 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
31651 Remove integer-related logic from pattern.
31652 (neon_vabd<mode>_3): Likewise.
31654 2018-01-15 Jakub Jelinek <jakub@redhat.com>
31656 PR middle-end/82694
31657 * common.opt (fstrict-overflow): No longer an alias.
31658 (fwrapv-pointer): New option.
31659 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
31660 also for pointer types based on flag_wrapv_pointer.
31661 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
31662 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
31663 opts->x_flag_wrapv got set.
31664 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
31665 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
31666 POINTER_TYPE_OVERFLOW_UNDEFINED.
31667 * match.pd: Likewise in address comparison pattern.
31668 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
31670 2018-01-15 Richard Biener <rguenther@suse.de>
31673 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
31674 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
31675 Reset type names to their identifier if their TYPE_DECL doesn't
31676 have linkage (and thus is used for ODR and devirt).
31677 (save_debug_info_for_decl): Remove.
31678 (save_debug_info_for_type): Likewise.
31679 (add_tree_to_fld_list): Adjust.
31680 * tree-pretty-print.c (dump_generic_node): Make dumping of
31681 type names more robust.
31683 2018-01-15 Richard Biener <rguenther@suse.de>
31685 * BASE-VER: Bump to 8.0.1.
31687 2018-01-14 Martin Sebor <msebor@redhat.com>
31690 * builtins.c (check_access): Avoid warning when the no-warning bit
31693 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
31695 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
31696 * ira-color (allocno_hard_regs_compare): Likewise.
31698 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
31701 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
31702 Use .pushsection/.popsection.
31704 2018-01-14 Martin Sebor <msebor@redhat.com>
31707 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
31709 2018-01-14 Jakub Jelinek <jakub@redhat.com>
31711 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
31712 entry from extra_headers.
31713 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
31714 extra_headers, make the list bitwise identical to the i?86-*-* one.
31716 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31718 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
31719 -mcmodel=large with -mindirect-branch=thunk,
31720 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
31721 -mfunction-return=thunk-extern.
31722 * doc/invoke.texi: Document -mcmodel=large is incompatible with
31723 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
31724 -mfunction-return=thunk and -mfunction-return=thunk-extern.
31726 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31728 * config/i386/i386.c (print_reg): Print the name of the full
31729 integer register without '%'.
31730 (ix86_print_operand): Handle 'V'.
31731 * doc/extend.texi: Document 'V' modifier.
31733 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31735 * config/i386/constraints.md (Bs): Disallow memory operand for
31736 -mindirect-branch-register.
31738 * config/i386/predicates.md (indirect_branch_operand): Likewise.
31739 (GOT_memory_operand): Likewise.
31740 (call_insn_operand): Likewise.
31741 (sibcall_insn_operand): Likewise.
31742 (GOT32_symbol_operand): Likewise.
31743 * config/i386/i386.md (indirect_jump): Call convert_memory_address
31744 for -mindirect-branch-register.
31745 (tablejump): Likewise.
31746 (*sibcall_memory): Likewise.
31747 (*sibcall_value_memory): Likewise.
31748 Disallow peepholes of indirect call and jump via memory for
31749 -mindirect-branch-register.
31750 (*call_pop): Replace m with Bw.
31751 (*call_value_pop): Likewise.
31752 (*sibcall_pop_memory): Replace m with Bs.
31753 * config/i386/i386.opt (mindirect-branch-register): New option.
31754 * doc/invoke.texi: Document -mindirect-branch-register option.
31756 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31758 * config/i386/i386-protos.h (ix86_output_function_return): New.
31759 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
31760 set function_return_type.
31761 (indirect_thunk_name): Add ret_p to indicate thunk for function
31763 (output_indirect_thunk_function): Pass false to
31764 indirect_thunk_name.
31765 (ix86_output_indirect_branch_via_reg): Likewise.
31766 (ix86_output_indirect_branch_via_push): Likewise.
31767 (output_indirect_thunk_function): Create alias for function
31768 return thunk if regno < 0.
31769 (ix86_output_function_return): New function.
31770 (ix86_handle_fndecl_attribute): Handle function_return.
31771 (ix86_attribute_table): Add function_return.
31772 * config/i386/i386.h (machine_function): Add
31773 function_return_type.
31774 * config/i386/i386.md (simple_return_internal): Use
31775 ix86_output_function_return.
31776 (simple_return_internal_long): Likewise.
31777 * config/i386/i386.opt (mfunction-return=): New option.
31778 (indirect_branch): Mention -mfunction-return=.
31779 * doc/extend.texi: Document function_return function attribute.
31780 * doc/invoke.texi: Document -mfunction-return= option.
31782 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31784 * config/i386/i386-opts.h (indirect_branch): New.
31785 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
31786 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
31787 with local indirect jump when converting indirect call and jump.
31788 (ix86_set_indirect_branch_type): New.
31789 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
31790 (indirectlabelno): New.
31791 (indirect_thunk_needed): Likewise.
31792 (indirect_thunk_bnd_needed): Likewise.
31793 (indirect_thunks_used): Likewise.
31794 (indirect_thunks_bnd_used): Likewise.
31795 (INDIRECT_LABEL): Likewise.
31796 (indirect_thunk_name): Likewise.
31797 (output_indirect_thunk): Likewise.
31798 (output_indirect_thunk_function): Likewise.
31799 (ix86_output_indirect_branch_via_reg): Likewise.
31800 (ix86_output_indirect_branch_via_push): Likewise.
31801 (ix86_output_indirect_branch): Likewise.
31802 (ix86_output_indirect_jmp): Likewise.
31803 (ix86_code_end): Call output_indirect_thunk_function if needed.
31804 (ix86_output_call_insn): Call ix86_output_indirect_branch if
31806 (ix86_handle_fndecl_attribute): Handle indirect_branch.
31807 (ix86_attribute_table): Add indirect_branch.
31808 * config/i386/i386.h (machine_function): Add indirect_branch_type
31809 and has_local_indirect_jump.
31810 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
31812 (tablejump): Likewise.
31813 (*indirect_jump): Use ix86_output_indirect_jmp.
31814 (*tablejump_1): Likewise.
31815 (simple_return_indirect_internal): Likewise.
31816 * config/i386/i386.opt (mindirect-branch=): New option.
31817 (indirect_branch): New.
31820 (thunk-inline): Likewise.
31821 (thunk-extern): Likewise.
31822 * doc/extend.texi: Document indirect_branch function attribute.
31823 * doc/invoke.texi: Document -mindirect-branch= option.
31825 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
31828 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
31830 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
31832 * ipa-inline.c (want_inline_small_function_p): Return false if
31833 inlining has already failed with CIF_FINAL_ERROR.
31834 (update_caller_keys): Call want_inline_small_function_p before
31836 (update_callee_keys): Likewise.
31838 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
31840 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
31842 (rs6000_quadword_masked_address_p): Likewise.
31843 (quad_aligned_load_p): Likewise.
31844 (quad_aligned_store_p): Likewise.
31845 (const_load_sequence_p): Add comment to describe the outer-most loop.
31846 (mimic_memory_attributes_and_flags): New function.
31847 (rs6000_gen_stvx): Likewise.
31848 (replace_swapped_aligned_store): Likewise.
31849 (rs6000_gen_lvx): Likewise.
31850 (replace_swapped_aligned_load): Likewise.
31851 (replace_swapped_load_constant): Capitalize argument name in
31852 comment describing this function.
31853 (rs6000_analyze_swaps): Add a third pass to search for vector loads
31854 and stores that access quad-word aligned addresses and replace
31855 with stvx or lvx instructions when appropriate.
31856 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
31857 New function prototype.
31858 (rs6000_quadword_masked_address_p): Likewise.
31859 (rs6000_gen_lvx): Likewise.
31860 (rs6000_gen_stvx): Likewise.
31861 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
31862 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
31863 when memory address is aligned.
31864 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
31865 this split to select lvx instruction when memory address is aligned.
31866 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
31867 instruction when memory address is aligned.
31868 (*vsx_le_perm_load_v16qi): Likewise.
31869 (four unnamed splitters): Modify to select the stvx instruction
31870 when memory is aligned.
31872 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
31874 * predict.c (determine_unlikely_bbs): Handle correctly BBs
31875 which appears in the queue multiple times.
31877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31878 Alan Hayward <alan.hayward@arm.com>
31879 David Sherwood <david.sherwood@arm.com>
31881 * tree-vectorizer.h (vec_lower_bound): New structure.
31882 (_loop_vec_info): Add check_nonzero and lower_bounds.
31883 (LOOP_VINFO_CHECK_NONZERO): New macro.
31884 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
31885 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
31886 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
31887 fields. Make seg_len the distance travelled, not including the
31889 (dr_direction_indicator): Declare.
31890 (dr_zero_step_indicator): Likewise.
31891 (dr_known_forward_stride_p): Likewise.
31892 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
31894 (runtime_alias_check_p): Allow runtime alias checks with
31896 (operator ==): Compare access_size and align.
31897 (prune_runtime_alias_test_list): Rework for new distinction between
31898 the access_size and seg_len.
31899 (create_intersect_range_checks_index): Likewise. Cope with polynomial
31901 (get_segment_min_max): New function.
31902 (create_intersect_range_checks): Use it.
31903 (dr_step_indicator): New function.
31904 (dr_direction_indicator): Likewise.
31905 (dr_zero_step_indicator): Likewise.
31906 (dr_known_forward_stride_p): Likewise.
31907 * tree-loop-distribution.c (data_ref_segment_size): Return
31908 DR_STEP * (niters - 1).
31909 (compute_alias_check_pairs): Update call to the dr_with_seg_len
31911 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
31912 (vect_preserves_scalar_order_p): New function, split out from...
31913 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
31914 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
31915 (vect_vfa_access_size): New function.
31916 (vect_vfa_align): Likewise.
31917 (vect_compile_time_alias): Take access_size_a and access_b arguments.
31918 (dump_lower_bound): New function.
31919 (vect_check_lower_bound): Likewise.
31920 (vect_small_gap_p): Likewise.
31921 (vectorizable_with_step_bound_p): Likewise.
31922 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
31923 depencies if the vectorization factor is 1. Convert the checks
31924 for nonzero steps into checks on the bounds of DR_STEP. Try using
31925 a bunds check for variable steps if the minimum required step is
31926 relatively small. Update calls to the dr_with_seg_len
31927 constructor and to vect_compile_time_alias.
31928 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
31930 (vect_loop_versioning): Call it.
31931 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
31933 (vect_estimate_min_profitable_iters): Account for any bounds checks.
31935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31936 Alan Hayward <alan.hayward@arm.com>
31937 David Sherwood <david.sherwood@arm.com>
31939 * doc/sourcebuild.texi (vect_scatter_store): Document.
31940 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
31942 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
31944 * genopinit.c (main): Add supports_vec_scatter_store and
31945 supports_vec_scatter_store_cached to target_optabs.
31946 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
31947 IFN_MASK_SCATTER_STORE.
31948 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
31950 * internal-fn.h (internal_store_fn_p): Declare.
31951 (internal_fn_stored_value_index): Likewise.
31952 * internal-fn.c (scatter_store_direct): New macro.
31953 (expand_scatter_store_optab_fn): New function.
31954 (direct_scatter_store_optab_supported_p): New macro.
31955 (internal_store_fn_p): New function.
31956 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
31957 IFN_MASK_SCATTER_STORE.
31958 (internal_fn_mask_index): Likewise.
31959 (internal_fn_stored_value_index): New function.
31960 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
31961 for scatter stores.
31962 * optabs-query.h (supports_vec_scatter_store_p): Declare.
31963 * optabs-query.c (supports_vec_scatter_store_p): New function.
31964 * tree-vectorizer.h (vect_get_store_rhs): Declare.
31965 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
31966 true for scatter stores.
31967 (vect_gather_scatter_fn_p): Handle scatter stores too.
31968 (vect_check_gather_scatter): Consider using scatter stores if
31969 supports_vec_scatter_store_p.
31970 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
31971 scatter stores too.
31972 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
31973 internal_fn_stored_value_index.
31974 (check_load_store_masking): Handle scatter stores too.
31975 (vect_get_store_rhs): Make public.
31976 (vectorizable_call): Use internal_store_fn_p.
31977 (vectorizable_store): Handle scatter store internal functions.
31978 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
31979 when deciding whether the end of the group has been reached.
31980 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
31981 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
31982 (mask_scatter_store<mode>): New insns.
31984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31985 Alan Hayward <alan.hayward@arm.com>
31986 David Sherwood <david.sherwood@arm.com>
31988 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
31989 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
31990 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
31992 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
31993 Use vect_truncate_gather_scatter_offset if we can't treat the
31994 operation as a normal gather load or scatter store.
31995 (get_group_load_store_type): Take the gather_scatter_info
31996 as argument. Try using a gather load or scatter store for
31997 single-element groups.
31998 (get_load_store_type): Update calls to get_group_load_store_type
31999 and vect_use_strided_gather_scatters_p.
32001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32002 Alan Hayward <alan.hayward@arm.com>
32003 David Sherwood <david.sherwood@arm.com>
32005 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
32006 optional tree argument.
32007 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
32009 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
32010 but continue to use the current value as a fallback.
32011 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
32012 to compare the updates.
32013 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
32014 (get_load_store_type): Use it when handling a strided access.
32015 (vect_get_strided_load_store_ops): New function.
32016 (vect_get_data_ptr_increment): Likewise.
32017 (vectorizable_load): Handle strided gather loads. Always pass
32018 a step to vect_create_data_ref_ptr and bump_vector_ptr.
32020 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32021 Alan Hayward <alan.hayward@arm.com>
32022 David Sherwood <david.sherwood@arm.com>
32024 * doc/md.texi (gather_load@var{m}): Document.
32025 (mask_gather_load@var{m}): Likewise.
32026 * genopinit.c (main): Add supports_vec_gather_load and
32027 supports_vec_gather_load_cached to target_optabs.
32028 * optabs-tree.c (init_tree_optimization_optabs): Use
32029 ggc_cleared_alloc to allocate target_optabs.
32030 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
32031 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
32033 * internal-fn.h (internal_load_fn_p): Declare.
32034 (internal_gather_scatter_fn_p): Likewise.
32035 (internal_fn_mask_index): Likewise.
32036 (internal_gather_scatter_fn_supported_p): Likewise.
32037 * internal-fn.c (gather_load_direct): New macro.
32038 (expand_gather_load_optab_fn): New function.
32039 (direct_gather_load_optab_supported_p): New macro.
32040 (direct_internal_fn_optab): New function.
32041 (internal_load_fn_p): Likewise.
32042 (internal_gather_scatter_fn_p): Likewise.
32043 (internal_fn_mask_index): Likewise.
32044 (internal_gather_scatter_fn_supported_p): Likewise.
32045 * optabs-query.c (supports_at_least_one_mode_p): New function.
32046 (supports_vec_gather_load_p): Likewise.
32047 * optabs-query.h (supports_vec_gather_load_p): Declare.
32048 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
32049 and memory_type field.
32050 (NUM_PATTERNS): Bump to 15.
32051 * tree-vect-data-refs.c: Include internal-fn.h.
32052 (vect_gather_scatter_fn_p): New function.
32053 (vect_describe_gather_scatter_call): Likewise.
32054 (vect_check_gather_scatter): Try using internal functions for
32055 gather loads. Recognize existing calls to a gather load function.
32056 (vect_analyze_data_refs): Consider using gather loads if
32057 supports_vec_gather_load_p.
32058 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
32059 (vect_get_gather_scatter_offset_type): Likewise.
32060 (vect_convert_mask_for_vectype): Likewise.
32061 (vect_add_conversion_to_patterm): Likewise.
32062 (vect_try_gather_scatter_pattern): Likewise.
32063 (vect_recog_gather_scatter_pattern): New pattern recognizer.
32064 (vect_vect_recog_func_ptrs): Add it.
32065 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
32066 internal_fn_mask_index and internal_gather_scatter_fn_p.
32067 (check_load_store_masking): Take the gather_scatter_info as an
32068 argument and handle gather loads.
32069 (vect_get_gather_scatter_ops): New function.
32070 (vectorizable_call): Check internal_load_fn_p.
32071 (vectorizable_load): Likewise. Handle gather load internal
32073 (vectorizable_store): Update call to check_load_store_masking.
32074 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
32075 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
32076 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
32077 (aarch64_gather_scale_operand_d): New predicates.
32078 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
32079 (mask_gather_load<mode>): New insns.
32081 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32082 Alan Hayward <alan.hayward@arm.com>
32083 David Sherwood <david.sherwood@arm.com>
32085 * optabs.def (fold_left_plus_optab): New optab.
32086 * doc/md.texi (fold_left_plus_@var{m}): Document.
32087 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
32088 * internal-fn.c (fold_left_direct): Define.
32089 (expand_fold_left_optab_fn): Likewise.
32090 (direct_fold_left_optab_supported_p): Likewise.
32091 * fold-const-call.c (fold_const_fold_left): New function.
32092 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
32093 * tree-parloops.c (valid_reduction_p): New function.
32094 (gather_scalar_reductions): Use it.
32095 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
32096 (vect_finish_replace_stmt): Declare.
32097 * tree-vect-loop.c (fold_left_reduction_fn): New function.
32098 (needs_fold_left_reduction_p): New function, split out from...
32099 (vect_is_simple_reduction): ...here. Accept reductions that
32100 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
32101 (vect_force_simple_reduction): Also store the reduction type in
32102 the assignment's STMT_VINFO_REDUC_TYPE.
32103 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
32104 (merge_with_identity): New function.
32105 (vect_expand_fold_left): Likewise.
32106 (vectorize_fold_left_reduction): Likewise.
32107 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
32108 scalar phi in place for it. Check for target support and reject
32109 cases that would reassociate the operation. Defer the transform
32110 phase to vectorize_fold_left_reduction.
32111 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
32112 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
32113 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
32115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32117 * tree-if-conv.c (predicate_mem_writes): Remove redundant
32118 call to ifc_temp_var.
32120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32121 Alan Hayward <alan.hayward@arm.com>
32122 David Sherwood <david.sherwood@arm.com>
32124 * target.def (legitimize_address_displacement): Take the original
32125 offset as a poly_int.
32126 * targhooks.h (default_legitimize_address_displacement): Update
32128 * targhooks.c (default_legitimize_address_displacement): Likewise.
32129 * doc/tm.texi: Regenerate.
32130 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
32131 as an argument, moving assert of ad->disp == ad->disp_term to...
32132 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
32133 Try calling targetm.legitimize_address_displacement before expanding
32134 the address rather than afterwards, and adjust for the new interface.
32135 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
32136 Match the new hook interface. Handle SVE addresses.
32137 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
32138 new hook interface.
32140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32142 * Makefile.in (OBJS): Add early-remat.o.
32143 * target.def (select_early_remat_modes): New hook.
32144 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
32145 * doc/tm.texi: Regenerate.
32146 * targhooks.h (default_select_early_remat_modes): Declare.
32147 * targhooks.c (default_select_early_remat_modes): New function.
32148 * timevar.def (TV_EARLY_REMAT): New timevar.
32149 * passes.def (pass_early_remat): New pass.
32150 * tree-pass.h (make_pass_early_remat): Declare.
32151 * early-remat.c: New file.
32152 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
32154 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
32156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32157 Alan Hayward <alan.hayward@arm.com>
32158 David Sherwood <david.sherwood@arm.com>
32160 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
32161 vfm1 with a bound_epilog parameter.
32162 (vect_do_peeling): Update calls accordingly, and move the prologue
32163 call earlier in the function. Treat the base bound_epilog as 0 for
32164 fully-masked loops and retain vf - 1 for other loops. Add 1 to
32165 this base when peeling for gaps.
32166 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
32167 with fully-masked loops.
32168 (vect_estimate_min_profitable_iters): Handle the single peeled
32169 iteration in that case.
32171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32172 Alan Hayward <alan.hayward@arm.com>
32173 David Sherwood <david.sherwood@arm.com>
32175 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
32176 single-element interleaving even if the size is not a power of 2.
32177 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
32178 accesses for single-element interleaving if the group size is
32181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32182 Alan Hayward <alan.hayward@arm.com>
32183 David Sherwood <david.sherwood@arm.com>
32185 * doc/md.texi (fold_extract_last_@var{m}): Document.
32186 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
32187 * optabs.def (fold_extract_last_optab): New optab.
32188 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
32189 * internal-fn.c (fold_extract_direct): New macro.
32190 (expand_fold_extract_optab_fn): Likewise.
32191 (direct_fold_extract_optab_supported_p): Likewise.
32192 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
32193 * tree-vect-loop.c (vect_model_reduction_cost): Handle
32194 EXTRACT_LAST_REDUCTION.
32195 (get_initial_def_for_reduction): Do not create an initial vector
32196 for EXTRACT_LAST_REDUCTION reductions.
32197 (vectorizable_reduction): Leave the scalar phi in place for
32198 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
32199 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
32200 epilogue code for EXTRACT_LAST_REDUCTION and defer the
32201 transform phase to vectorizable_condition.
32202 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
32204 (vect_finish_stmt_generation): ...here.
32205 (vect_finish_replace_stmt): New function.
32206 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
32207 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
32209 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
32211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32212 Alan Hayward <alan.hayward@arm.com>
32213 David Sherwood <david.sherwood@arm.com>
32215 * doc/md.texi (extract_last_@var{m}): Document.
32216 * optabs.def (extract_last_optab): New optab.
32217 * internal-fn.def (EXTRACT_LAST): New internal function.
32218 * internal-fn.c (cond_unary_direct): New macro.
32219 (expand_cond_unary_optab_fn): Likewise.
32220 (direct_cond_unary_optab_supported_p): Likewise.
32221 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
32222 loops using EXTRACT_LAST.
32223 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
32224 (extract_last_<mode>): ...this optab.
32225 (vec_extract<mode><Vel>): Update accordingly.
32227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32228 Alan Hayward <alan.hayward@arm.com>
32229 David Sherwood <david.sherwood@arm.com>
32231 * target.def (empty_mask_is_expensive): New hook.
32232 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
32233 * doc/tm.texi: Regenerate.
32234 * targhooks.h (default_empty_mask_is_expensive): Declare.
32235 * targhooks.c (default_empty_mask_is_expensive): New function.
32236 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
32237 if the target says that empty masks are expensive.
32238 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
32240 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
32242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32243 Alan Hayward <alan.hayward@arm.com>
32244 David Sherwood <david.sherwood@arm.com>
32246 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
32247 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
32248 (vect_use_loop_mask_for_alignment_p): New function.
32249 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
32250 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
32251 niters_skip argument. Make sure that the first niters_skip elements
32252 of the first iteration are inactive.
32253 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
32254 Update call to vect_set_loop_masks_directly.
32255 (get_misalign_in_elems): New function, split out from...
32256 (vect_gen_prolog_loop_niters): ...here.
32257 (vect_update_init_of_dr): Take a code argument that specifies whether
32258 the adjustment should be added or subtracted.
32259 (vect_update_init_of_drs): Likewise.
32260 (vect_prepare_for_masked_peels): New function.
32261 (vect_do_peeling): Skip prologue peeling if we're using a mask
32262 instead. Update call to vect_update_inits_of_drs.
32263 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
32265 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
32266 alignment. Do not include the number of peeled iterations in
32267 the minimum threshold in that case.
32268 (vectorizable_induction): Adjust the start value down by
32269 LOOP_VINFO_MASK_SKIP_NITERS iterations.
32270 (vect_transform_loop): Call vect_prepare_for_masked_peels.
32271 Take the number of skipped iterations into account when calculating
32273 * tree-vect-stmts.c (vect_gen_while_not): New function.
32275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32276 Alan Hayward <alan.hayward@arm.com>
32277 David Sherwood <david.sherwood@arm.com>
32279 * doc/sourcebuild.texi (vect_fully_masked): Document.
32280 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
32281 default value to 0.
32282 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
32284 (vect_analyze_loop_2): ...here. Don't check the vectorization
32285 factor against the number of loop iterations if the loop is
32288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32289 Alan Hayward <alan.hayward@arm.com>
32290 David Sherwood <david.sherwood@arm.com>
32292 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
32293 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
32294 (dump_groups): Update accordingly.
32295 (iv_use::mem_type): New member variable.
32296 (address_p): New function.
32297 (record_use): Add a mem_type argument and initialize the new
32299 (record_group_use): Add a mem_type argument. Use address_p.
32300 Remove obsolete null checks of base_object. Update call to record_use.
32301 (find_interesting_uses_op): Update call to record_group_use.
32302 (find_interesting_uses_cond): Likewise.
32303 (find_interesting_uses_address): Likewise.
32304 (get_mem_type_for_internal_fn): New function.
32305 (find_address_like_use): Likewise.
32306 (find_interesting_uses_stmt): Try find_address_like_use before
32307 calling find_interesting_uses_op.
32308 (addr_offset_valid_p): Use the iv mem_type field as the type
32309 of the addressed memory.
32310 (add_autoinc_candidates): Likewise.
32311 (get_address_cost): Likewise.
32312 (split_small_address_groups_p): Use address_p.
32313 (split_address_groups): Likewise.
32314 (add_iv_candidate_for_use): Likewise.
32315 (autoinc_possible_for_pair): Likewise.
32316 (rewrite_groups): Likewise.
32317 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
32318 (determine_group_iv_cost): Update after split of USE_ADDRESS.
32319 (get_alias_ptr_type_for_ptr_address): New function.
32320 (rewrite_use_address): Rewrite address uses in calls that were
32321 identified by find_address_like_use.
32323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32324 Alan Hayward <alan.hayward@arm.com>
32325 David Sherwood <david.sherwood@arm.com>
32327 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
32329 * gimple-expr.h (is_gimple_addressable: Likewise.
32330 * gimple-expr.c (is_gimple_address): Likewise.
32331 * internal-fn.c (expand_call_mem_ref): New function.
32332 (expand_mask_load_optab_fn): Use it.
32333 (expand_mask_store_optab_fn): Likewise.
32335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32336 Alan Hayward <alan.hayward@arm.com>
32337 David Sherwood <david.sherwood@arm.com>
32339 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
32340 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
32341 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
32342 (cond_umax@var{mode}): Document.
32343 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
32344 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
32345 (cond_umin_optab, cond_umax_optab): New optabs.
32346 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
32347 (COND_IOR, COND_XOR): New internal functions.
32348 * internal-fn.h (get_conditional_internal_fn): Declare.
32349 * internal-fn.c (cond_binary_direct): New macro.
32350 (expand_cond_binary_optab_fn): Likewise.
32351 (direct_cond_binary_optab_supported_p): Likewise.
32352 (get_conditional_internal_fn): New function.
32353 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
32354 Cope with reduction statements that are vectorized as calls rather
32356 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
32357 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
32358 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
32359 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
32360 (UNSPEC_COND_EOR): New unspecs.
32361 (optab): Add mappings for them.
32362 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
32363 (sve_int_op, sve_fp_op): New int attributes.
32365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32366 Alan Hayward <alan.hayward@arm.com>
32367 David Sherwood <david.sherwood@arm.com>
32369 * optabs.def (while_ult_optab): New optab.
32370 * doc/md.texi (while_ult@var{m}@var{n}): Document.
32371 * internal-fn.def (WHILE_ULT): New internal function.
32372 * internal-fn.h (direct_internal_fn_supported_p): New override
32373 that takes two types as argument.
32374 * internal-fn.c (while_direct): New macro.
32375 (expand_while_optab_fn): New function.
32376 (convert_optab_supported_p): Likewise.
32377 (direct_while_optab_supported_p): New macro.
32378 * wide-int.h (wi::udiv_ceil): New function.
32379 * tree-vectorizer.h (rgroup_masks): New structure.
32380 (vec_loop_masks): New typedef.
32381 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
32382 and fully_masked_p.
32383 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
32384 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
32385 (vect_max_vf): New function.
32386 (slpeel_make_loop_iterate_ntimes): Delete.
32387 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
32388 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
32389 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
32390 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
32391 internal-fn.h, stor-layout.h and optabs-query.h.
32392 (vect_set_loop_mask): New function.
32393 (add_preheader_seq): Likewise.
32394 (add_header_seq): Likewise.
32395 (interleave_supported_p): Likewise.
32396 (vect_maybe_permute_loop_masks): Likewise.
32397 (vect_set_loop_masks_directly): Likewise.
32398 (vect_set_loop_condition_masked): Likewise.
32399 (vect_set_loop_condition_unmasked): New function, split out from
32400 slpeel_make_loop_iterate_ntimes.
32401 (slpeel_make_loop_iterate_ntimes): Rename to..
32402 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
32403 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
32404 (vect_do_peeling): Update call accordingly.
32405 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
32407 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
32408 mask_compare_type, can_fully_mask_p and fully_masked_p.
32409 (release_vec_loop_masks): New function.
32410 (_loop_vec_info): Use it to free the loop masks.
32411 (can_produce_all_loop_masks_p): New function.
32412 (vect_get_max_nscalars_per_iter): Likewise.
32413 (vect_verify_full_masking): Likewise.
32414 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
32415 retries, and free the mask rgroups before retrying. Check loop-wide
32416 reasons for disallowing fully-masked loops. Make the final decision
32417 about whether use a fully-masked loop or not.
32418 (vect_estimate_min_profitable_iters): Do not assume that peeling
32419 for the number of iterations will be needed for fully-masked loops.
32420 (vectorizable_reduction): Disable fully-masked loops.
32421 (vectorizable_live_operation): Likewise.
32422 (vect_halve_mask_nunits): New function.
32423 (vect_double_mask_nunits): Likewise.
32424 (vect_record_loop_mask): Likewise.
32425 (vect_get_loop_mask): Likewise.
32426 (vect_transform_loop): Handle the case in which the final loop
32427 iteration might handle a partial vector. Call vect_set_loop_condition
32428 instead of slpeel_make_loop_iterate_ntimes.
32429 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
32430 (check_load_store_masking): New function.
32431 (prepare_load_store_mask): Likewise.
32432 (vectorizable_store): Handle fully-masked loops.
32433 (vectorizable_load): Likewise.
32434 (supportable_widening_operation): Use vect_halve_mask_nunits for
32436 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
32437 (vect_gen_while): New function.
32438 * config/aarch64/aarch64.md (umax<mode>3): New expander.
32439 (aarch64_uqdec<mode>): New insn.
32441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32442 Alan Hayward <alan.hayward@arm.com>
32443 David Sherwood <david.sherwood@arm.com>
32445 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
32446 (reduc_xor_scal_optab): New optabs.
32447 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
32448 (reduc_xor_scal_@var{m}): Document.
32449 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
32450 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
32451 internal functions.
32452 * fold-const-call.c (fold_const_call): Handle them.
32453 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
32454 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
32455 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
32456 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
32457 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
32458 (UNSPEC_XORV): New unspecs.
32459 (optab): Add entries for them.
32460 (BITWISEV): New int iterator.
32461 (bit_reduc_op): New int attributes.
32463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32464 Alan Hayward <alan.hayward@arm.com>
32465 David Sherwood <david.sherwood@arm.com>
32467 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
32468 * internal-fn.def (VEC_SHL_INSERT): New internal function.
32469 * optabs.def (vec_shl_insert_optab): New optab.
32470 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
32471 (duplicate_and_interleave): Likewise.
32472 * tree-vect-loop.c: Include internal-fn.h.
32473 (neutral_op_for_slp_reduction): New function, split out from
32474 get_initial_defs_for_reduction.
32475 (get_initial_def_for_reduction): Handle option 2 for variable-length
32476 vectors by loading the neutral value into a vector and then shifting
32477 the initial value into element 0.
32478 (get_initial_defs_for_reduction): Replace the code argument with
32479 the neutral value calculated by neutral_op_for_slp_reduction.
32480 Use gimple_build_vector for constant-length vectors.
32481 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
32482 but the first group_size elements have a neutral value.
32483 Use duplicate_and_interleave otherwise.
32484 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
32485 Update call to get_initial_defs_for_reduction. Handle SLP
32486 reductions for variable-length vectors by creating one vector
32487 result for each scalar result, with the elements associated
32488 with other scalar results stubbed out with the neutral value.
32489 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
32490 Require IFN_VEC_SHL_INSERT for double reductions on
32491 variable-length vectors, or SLP reductions that have
32492 a neutral value. Require can_duplicate_and_interleave_p
32493 support for variable-length unchained SLP reductions if there
32494 is no neutral value, such as for MIN/MAX reductions. Also require
32495 the number of vector elements to be a multiple of the number of
32496 SLP statements when doing variable-length unchained SLP reductions.
32497 Update call to vect_create_epilog_for_reduction.
32498 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
32499 and remove initial values.
32500 (duplicate_and_interleave): Make public.
32501 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
32502 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
32504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32505 Alan Hayward <alan.hayward@arm.com>
32506 David Sherwood <david.sherwood@arm.com>
32508 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
32509 (can_duplicate_and_interleave_p): New function.
32510 (vect_get_and_check_slp_defs): Take the vector of statements
32511 rather than just the current one. Remove excess parentheses.
32512 Restriction rejectinon of vect_constant_def and vect_external_def
32513 for variable-length vectors to boolean types, or types for which
32514 can_duplicate_and_interleave_p is false.
32515 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
32516 (duplicate_and_interleave): New function.
32517 (vect_get_constant_vectors): Use gimple_build_vector for
32518 constant-length vectors and suitable variable-length constant
32519 vectors. Use duplicate_and_interleave for other variable-length
32520 vectors. Don't defer the update when inserting new statements.
32522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32523 Alan Hayward <alan.hayward@arm.com>
32524 David Sherwood <david.sherwood@arm.com>
32526 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
32527 min_profitable_iters doesn't go negative.
32529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32530 Alan Hayward <alan.hayward@arm.com>
32531 David Sherwood <david.sherwood@arm.com>
32533 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
32534 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
32535 * optabs.def (vec_mask_load_lanes_optab): New optab.
32536 (vec_mask_store_lanes_optab): Likewise.
32537 * internal-fn.def (MASK_LOAD_LANES): New internal function.
32538 (MASK_STORE_LANES): Likewise.
32539 * internal-fn.c (mask_load_lanes_direct): New macro.
32540 (mask_store_lanes_direct): Likewise.
32541 (expand_mask_load_optab_fn): Handle masked operations.
32542 (expand_mask_load_lanes_optab_fn): New macro.
32543 (expand_mask_store_optab_fn): Handle masked operations.
32544 (expand_mask_store_lanes_optab_fn): New macro.
32545 (direct_mask_load_lanes_optab_supported_p): Likewise.
32546 (direct_mask_store_lanes_optab_supported_p): Likewise.
32547 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
32549 (vect_load_lanes_supported): Likewise.
32550 * tree-vect-data-refs.c (strip_conversion): New function.
32551 (can_group_stmts_p): Likewise.
32552 (vect_analyze_data_ref_accesses): Use it instead of checking
32553 for a pair of assignments.
32554 (vect_store_lanes_supported): Take a masked_p parameter.
32555 (vect_load_lanes_supported): Likewise.
32556 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
32557 vect_store_lanes_supported and vect_load_lanes_supported.
32558 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
32559 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
32560 parameter. Don't allow gaps for masked accesses.
32561 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
32562 and vect_load_lanes_supported.
32563 (get_load_store_type): Take a masked_p parameter and update
32564 call to get_group_load_store_type.
32565 (vectorizable_store): Update call to get_load_store_type.
32566 Handle IFN_MASK_STORE_LANES.
32567 (vectorizable_load): Update call to get_load_store_type.
32568 Handle IFN_MASK_LOAD_LANES.
32570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32571 Alan Hayward <alan.hayward@arm.com>
32572 David Sherwood <david.sherwood@arm.com>
32574 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
32576 * config/aarch64/aarch64-protos.h
32577 (aarch64_sve_struct_memory_operand_p): Declare.
32578 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
32579 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
32580 (VPRED, vpred): Handle SVE structure modes.
32581 * config/aarch64/constraints.md (Utx): New constraint.
32582 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
32583 (aarch64_sve_struct_nonimmediate_operand): New predicates.
32584 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
32585 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
32586 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
32587 structure modes. Split into pieces after RA.
32588 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
32589 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
32591 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
32592 SVE structure modes.
32593 (aarch64_classify_address): Likewise.
32594 (sizetochar): Move earlier in file.
32595 (aarch64_print_operand): Handle SVE register lists.
32596 (aarch64_array_mode): New function.
32597 (aarch64_sve_struct_memory_operand_p): Likewise.
32598 (TARGET_ARRAY_MODE): Redefine.
32600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32601 Alan Hayward <alan.hayward@arm.com>
32602 David Sherwood <david.sherwood@arm.com>
32604 * target.def (array_mode): New target hook.
32605 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
32606 * doc/tm.texi: Regenerate.
32607 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
32608 * hooks.c (hook_optmode_mode_uhwi_none): New function.
32609 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
32610 targetm.array_mode.
32611 * stor-layout.c (mode_for_array): Likewise. Support polynomial
32614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32615 Alan Hayward <alan.hayward@arm.com>
32616 David Sherwood <david.sherwood@arm.com>
32618 * fold-const.c (fold_binary_loc): Check the argument types
32619 rather than the result type when testing for a vector operation.
32621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32623 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
32624 * doc/tm.texi: Regenerate.
32626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32627 Alan Hayward <alan.hayward@arm.com>
32628 David Sherwood <david.sherwood@arm.com>
32630 * doc/invoke.texi (-msve-vector-bits=): Document new option.
32631 (sve): Document new AArch64 extension.
32632 * doc/md.texi (w): Extend the description of the AArch64
32633 constraint to include SVE vectors.
32634 (Upl, Upa): Document new AArch64 predicate constraints.
32635 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
32637 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
32638 (msve-vector-bits=): New option.
32639 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
32640 SVE when these are disabled.
32641 (sve): New extension.
32642 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
32643 modes. Adjust their number of units based on aarch64_sve_vg.
32644 (MAX_BITSIZE_MODE_ANY_MODE): Define.
32645 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
32646 aarch64_addr_query_type.
32647 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
32648 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
32649 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
32650 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
32651 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
32652 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
32653 (aarch64_simd_imm_zero_p): Delete.
32654 (aarch64_check_zero_based_sve_index_immediate): Declare.
32655 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
32656 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
32657 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
32658 (aarch64_sve_float_mul_immediate_p): Likewise.
32659 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32660 rather than an rtx.
32661 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
32662 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
32663 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
32664 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
32665 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
32666 (aarch64_regmode_natural_size): Likewise.
32667 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
32668 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
32670 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
32671 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
32672 for VG and the SVE predicate registers.
32673 (V_ALIASES): Add a "z"-prefixed alias.
32674 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
32675 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
32676 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
32677 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
32678 (REG_CLASS_NAMES): Add entries for them.
32679 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
32680 and the predicate registers.
32681 (aarch64_sve_vg): Declare.
32682 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
32683 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
32684 (REGMODE_NATURAL_SIZE): Define.
32685 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
32687 * config/aarch64/aarch64.c: Include cfgrtl.h.
32688 (simd_immediate_info): Add a constructor for series vectors,
32689 and an associated step field.
32690 (aarch64_sve_vg): New variable.
32691 (aarch64_dbx_register_number): Handle VG and the predicate registers.
32692 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
32693 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
32694 (VEC_ANY_DATA, VEC_STRUCT): New constants.
32695 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
32696 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
32697 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
32698 (aarch64_get_mask_mode): New functions.
32699 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
32700 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
32701 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
32702 predicate modes and predicate registers. Explicitly restrict
32703 GPRs to modes of 16 bytes or smaller. Only allow FP registers
32704 to store a vector mode if it is recognized by
32705 aarch64_classify_vector_mode.
32706 (aarch64_regmode_natural_size): New function.
32707 (aarch64_hard_regno_caller_save_mode): Return the original mode
32709 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
32710 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
32711 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
32712 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
32714 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
32715 does not overlap dest if the function is frame-related. Handle
32717 (aarch64_split_add_offset): New function.
32718 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
32719 them aarch64_add_offset.
32720 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
32721 and update call to aarch64_sub_sp.
32722 (aarch64_add_cfa_expression): New function.
32723 (aarch64_expand_prologue): Pass extra temporary registers to the
32724 functions above. Handle the case in which we need to emit new
32725 DW_CFA_expressions for registers that were originally saved
32726 relative to the stack pointer, but now have to be expressed
32727 relative to the frame pointer.
32728 (aarch64_output_mi_thunk): Pass extra temporary registers to the
32730 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
32731 IP0 and IP1 values for SVE frames.
32732 (aarch64_expand_vec_series): New function.
32733 (aarch64_expand_sve_widened_duplicate): Likewise.
32734 (aarch64_expand_sve_const_vector): Likewise.
32735 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
32736 Handle SVE constants. Use emit_move_insn to move a force_const_mem
32737 into the register, rather than emitting a SET directly.
32738 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
32739 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
32740 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
32741 (offset_9bit_signed_scaled_p): New functions.
32742 (aarch64_replicate_bitmask_imm): New function.
32743 (aarch64_bitmask_imm): Use it.
32744 (aarch64_cannot_force_const_mem): Reject expressions involving
32745 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
32746 (aarch64_classify_index): Handle SVE indices, by requiring
32747 a plain register index with a scale that matches the element size.
32748 (aarch64_classify_address): Handle SVE addresses. Assert that
32749 the mode of the address is VOIDmode or an integer mode.
32750 Update call to aarch64_classify_symbol.
32751 (aarch64_classify_symbolic_expression): Update call to
32752 aarch64_classify_symbol.
32753 (aarch64_const_vec_all_in_range_p): New function.
32754 (aarch64_print_vector_float_operand): Likewise.
32755 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
32756 "vN" for FP registers with SVE modes. Handle (const ...) vectors
32757 and the FP immediates 1.0 and 0.5.
32758 (aarch64_print_address_internal): Handle SVE addresses.
32759 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
32760 (aarch64_regno_regclass): Handle predicate registers.
32761 (aarch64_secondary_reload): Handle big-endian reloads of SVE
32763 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
32764 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
32765 (aarch64_convert_sve_vector_bits): New function.
32766 (aarch64_override_options): Use it to handle -msve-vector-bits=.
32767 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32768 rather than an rtx.
32769 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
32770 Handle SVE vector and predicate modes. Accept VL-based constants
32771 that need only one temporary register, and VL offsets that require
32772 no temporary registers.
32773 (aarch64_conditional_register_usage): Mark the predicate registers
32774 as fixed if SVE isn't available.
32775 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
32776 Return true for SVE vector and predicate modes.
32777 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
32778 rather than an unsigned int. Handle SVE modes.
32779 (aarch64_preferred_simd_mode): Update call accordingly. Handle
32781 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
32783 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
32784 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
32785 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
32786 (aarch64_sve_float_mul_immediate_p): New functions.
32787 (aarch64_sve_valid_immediate): New function.
32788 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
32789 Explicitly reject structure modes. Check for INDEX constants.
32790 Handle PTRUE and PFALSE constants.
32791 (aarch64_check_zero_based_sve_index_immediate): New function.
32792 (aarch64_simd_imm_zero_p): Delete.
32793 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
32794 vector modes. Accept constants in the range of CNT[BHWD].
32795 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
32796 ask for an Advanced SIMD mode.
32797 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
32798 (aarch64_simd_vector_alignment): Handle SVE predicates.
32799 (aarch64_vectorize_preferred_vector_alignment): New function.
32800 (aarch64_simd_vector_alignment_reachable): Use it instead of
32802 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
32803 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
32805 (MAX_VECT_LEN): Delete.
32806 (expand_vec_perm_d): Add a vec_flags field.
32807 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
32808 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
32809 (aarch64_evpc_ext): Don't apply a big-endian lane correction
32811 (aarch64_evpc_rev): Rename to...
32812 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
32813 (aarch64_evpc_rev_global): New function.
32814 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
32815 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
32817 (aarch64_evpc_sve_tbl): New function.
32818 (aarch64_expand_vec_perm_const_1): Update after rename of
32819 aarch64_evpc_rev. Handle SVE permutes too, trying
32820 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
32821 than aarch64_evpc_tbl.
32822 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
32823 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
32824 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
32825 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
32826 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
32827 (aarch64_expand_sve_vcond): New functions.
32828 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
32829 of aarch64_vector_mode_p.
32830 (aarch64_dwarf_poly_indeterminate_value): New function.
32831 (aarch64_compute_pressure_classes): Likewise.
32832 (aarch64_can_change_mode_class): Likewise.
32833 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
32834 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
32835 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
32836 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
32837 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
32838 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
32839 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
32840 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
32842 (Dn, Dl, Dr): Accept const as well as const_vector.
32843 (Dz): Likewise. Compare against CONST0_RTX.
32844 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
32845 of "vector" where appropriate.
32846 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
32847 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
32848 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
32849 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
32850 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
32851 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
32852 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
32853 (v_int_equiv): Extend to SVE modes.
32854 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
32856 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
32857 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
32858 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
32859 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
32860 (SVE_COND_FP_CMP): New int iterators.
32861 (perm_hilo): Handle the new unpack unspecs.
32862 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
32864 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
32865 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
32866 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
32867 (aarch64_equality_operator, aarch64_constant_vector_operand)
32868 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
32869 (aarch64_sve_nonimmediate_operand): Likewise.
32870 (aarch64_sve_general_operand): Likewise.
32871 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
32872 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
32873 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
32874 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
32875 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
32876 (aarch64_sve_float_arith_immediate): Likewise.
32877 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
32878 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
32879 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
32880 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
32881 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
32882 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
32883 (aarch64_sve_float_arith_operand): Likewise.
32884 (aarch64_sve_float_arith_with_sub_operand): Likewise.
32885 (aarch64_sve_float_mul_operand): Likewise.
32886 (aarch64_sve_vec_perm_operand): Likewise.
32887 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
32888 (aarch64_mov_operand): Accept const_poly_int and const_vector.
32889 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
32890 as well as const_vector.
32891 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
32892 in file. Use CONST0_RTX and CONSTM1_RTX.
32893 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
32894 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
32895 Use aarch64_simd_imm_zero.
32896 * config/aarch64/aarch64-sve.md: New file.
32897 * config/aarch64/aarch64.md: Include it.
32898 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
32899 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
32900 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
32901 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
32902 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
32903 (sve): New attribute.
32904 (enabled): Disable instructions with the sve attribute unless
32906 (movqi, movhi): Pass CONST_POLY_INT operaneds through
32907 aarch64_expand_mov_immediate.
32908 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
32909 CNT[BHSD] immediates.
32910 (movti): Split CONST_POLY_INT moves into two halves.
32911 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
32912 Split additions that need a temporary here if the destination
32913 is the stack pointer.
32914 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
32915 (*add<mode>3_poly_1): New instruction.
32916 (set_clobber_cc): New expander.
32918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32920 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
32921 parameter and use it instead of GET_MODE_SIZE (innermode). Use
32922 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
32923 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
32924 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
32925 Change innermode from fixed_mode_size to machine_mode.
32926 (simplify_subreg): Update call accordingly. Handle a constant-sized
32927 subreg of a variable-length CONST_VECTOR.
32929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32930 Alan Hayward <alan.hayward@arm.com>
32931 David Sherwood <david.sherwood@arm.com>
32933 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
32934 (add_offset_to_base): New function, split out from...
32935 (create_mem_ref): ...here. When handling a scale other than 1,
32936 check first whether the address is valid without the offset.
32937 Add it into the base if so, leaving the index and scale as-is.
32939 2018-01-12 Jakub Jelinek <jakub@redhat.com>
32942 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
32943 fold_for_warn before checking if arg2 is INTEGER_CST.
32945 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
32947 * config/rs6000/predicates.md (load_multiple_operation): Delete.
32948 (store_multiple_operation): Delete.
32949 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
32950 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
32951 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
32952 guarded by TARGET_STRING.
32953 (rs6000_output_load_multiple): Delete.
32954 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
32955 OPTION_MASK_STRING / TARGET_STRING handling.
32956 (print_operand) <'N', 'O'>: Add comment that these are unused now.
32957 (const rs6000_opt_masks) <"string">: Change mask to 0.
32958 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
32959 (MASK_STRING): Delete.
32960 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
32962 (load_multiple): Delete.
32969 (store_multiple): Delete.
32976 (movmemsi_8reg): Delete.
32977 (corresponding unnamed define_insn): Delete.
32978 (movmemsi_6reg): Delete.
32979 (corresponding unnamed define_insn): Delete.
32980 (movmemsi_4reg): Delete.
32981 (corresponding unnamed define_insn): Delete.
32982 (movmemsi_2reg): Delete.
32983 (corresponding unnamed define_insn): Delete.
32984 (movmemsi_1reg): Delete.
32985 (corresponding unnamed define_insn): Delete.
32986 * config/rs6000/rs6000.opt (mno-string): New.
32987 (mstring): Replace by deprecation warning stub.
32988 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
32990 2018-01-12 Jakub Jelinek <jakub@redhat.com>
32992 * regrename.c (regrename_do_replace): If replacing the same
32993 reg multiple times, try to reuse last created gen_raw_REG.
32996 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
32997 main to workaround a bug in GDB.
32999 2018-01-12 Tom de Vries <tom@codesourcery.com>
33002 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
33004 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
33006 PR rtl-optimization/80481
33007 * ira-color.c (get_cap_member): New function.
33008 (allocnos_conflict_by_live_ranges_p): Use it.
33009 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
33010 (setup_slot_coalesced_allocno_live_ranges): Ditto.
33012 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
33015 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
33016 (*saddl_se_1): Ditto.
33017 (*ssubsi_1): Ditto.
33018 (*ssubl_se_1): Ditto.
33020 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
33022 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
33023 rather than wi::to_widest for DR_INITs.
33024 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
33025 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
33026 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
33028 (vect_analyze_group_access_1): Note that here.
33030 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
33032 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
33033 polynomial type sizes.
33035 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
33037 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
33038 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
33039 (gimple_add_tmp_var): Likewise.
33041 2018-01-12 Martin Liska <mliska@suse.cz>
33043 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
33044 (gimple_alloc_sizes): Likewise.
33045 (dump_gimple_statistics): Use PRIu64 in printf format.
33046 * gimple.h: Change uint64_t to int.
33048 2018-01-12 Martin Liska <mliska@suse.cz>
33050 * tree-core.h: Use uint64_t instead of int.
33051 * tree.c (tree_node_counts): Likewise.
33052 (tree_node_sizes): Likewise.
33053 (dump_tree_statistics): Use PRIu64 in printf format.
33055 2018-01-12 Martin Liska <mliska@suse.cz>
33057 * Makefile.in: As qsort_chk is implemented in vec.c, add
33058 vec.o to linkage of gencfn-macros.
33059 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
33060 passing the info to record_node_allocation_statistics.
33061 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
33063 * ggc-common.c (struct ggc_usage): Add operator== and use
33064 it in operator< and compare function.
33065 * mem-stats.h (struct mem_usage): Likewise.
33066 * vec.c (struct vec_usage): Remove operator< and compare
33067 function. Can be simply inherited.
33069 2018-01-12 Martin Jambor <mjambor@suse.cz>
33072 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
33073 * tree-ssa-math-opts.c: Include domwalk.h.
33074 (convert_mult_to_fma_1): New function.
33075 (fma_transformation_info): New type.
33076 (fma_deferring_state): Likewise.
33077 (cancel_fma_deferring): New function.
33078 (result_of_phi): Likewise.
33079 (last_fma_candidate_feeds_initial_phi): Likewise.
33080 (convert_mult_to_fma): Added deferring logic, split actual
33081 transformation to convert_mult_to_fma_1.
33082 (math_opts_dom_walker): New type.
33083 (math_opts_dom_walker::after_dom_children): New method, body moved
33084 here from pass_optimize_widening_mul::execute, added deferring logic
33086 (pass_optimize_widening_mul::execute): Moved most of code to
33087 math_opts_dom_walker::after_dom_children.
33088 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
33089 * config/i386/i386.c (ix86_option_override_internal): Added
33090 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
33092 2018-01-12 Richard Biener <rguenther@suse.de>
33095 * dwarf2out.c (gen_variable_die): Do not reset old_die for
33096 inline instance vars.
33098 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
33101 * config/rx/rx.c (rx_is_restricted_memory_address):
33102 Handle SUBREG case.
33104 2018-01-12 Richard Biener <rguenther@suse.de>
33106 PR tree-optimization/80846
33107 * target.def (split_reduction): New target hook.
33108 * targhooks.c (default_split_reduction): New function.
33109 * targhooks.h (default_split_reduction): Declare.
33110 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
33111 target requests first reduce vectors by combining low and high
33113 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
33114 (get_vectype_for_scalar_type_and_size): Export.
33115 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
33116 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
33117 * doc/tm.texi: Regenerate.
33118 * config/i386/i386.c (ix86_split_reduction): Implement
33119 TARGET_VECTORIZE_SPLIT_REDUCTION.
33121 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
33124 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
33125 in PIC mode except for TARGET_VXWORKS_RTP.
33126 * config/sparc/sparc.c: Include cfgrtl.h.
33127 (TARGET_INIT_PIC_REG): Define.
33128 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
33129 (sparc_pic_register_p): New predicate.
33130 (sparc_legitimate_address_p): Use it.
33131 (sparc_legitimize_pic_address): Likewise.
33132 (sparc_delegitimize_address): Likewise.
33133 (sparc_mode_dependent_address_p): Likewise.
33134 (gen_load_pcrel_sym): Remove 4th parameter.
33135 (load_got_register): Adjust call to above. Remove obsolete stuff.
33136 (sparc_expand_prologue): Do not call load_got_register here.
33137 (sparc_flat_expand_prologue): Likewise.
33138 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
33139 (sparc_use_pseudo_pic_reg): New function.
33140 (sparc_init_pic_reg): Likewise.
33141 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
33142 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
33144 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
33146 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
33147 Add item for branch_cost.
33149 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
33151 PR rtl-optimization/83565
33152 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
33153 not extend the result to a larger mode for rotate operations.
33154 (num_sign_bit_copies1): Likewise.
33156 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
33159 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
33161 Use values-Xc.o for -pedantic.
33162 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
33164 2018-01-12 Martin Liska <mliska@suse.cz>
33167 * ipa-devirt.c (final_warning_record::grow_type_warnings):
33169 (possible_polymorphic_call_targets): Use it.
33170 (ipa_devirt): Likewise.
33172 2018-01-12 Martin Liska <mliska@suse.cz>
33174 * profile-count.h (enum profile_quality): Use 0 as invalid
33175 enum value of profile_quality.
33177 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
33179 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
33180 -mext-string options.
33182 2018-01-12 Richard Biener <rguenther@suse.de>
33184 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
33185 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
33186 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
33188 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
33190 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
33192 * configure.ac (--with-long-double-format): Add support for the
33193 configuration option to change the default long double format on
33195 * config.gcc (powerpc*-linux*-*): Likewise.
33196 * configure: Regenerate.
33197 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
33198 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
33199 used without modification.
33201 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
33203 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
33204 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
33205 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
33206 MISC_BUILTIN_SPEC_BARRIER.
33207 (rs6000_init_builtins): Likewise.
33208 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
33210 (speculation_barrier): New define_insn.
33211 * doc/extend.texi: Document __builtin_speculation_barrier.
33213 2018-01-11 Jakub Jelinek <jakub@redhat.com>
33216 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
33217 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
33218 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
33220 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
33221 integral modes instead of "ss" and "sd".
33222 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
33223 vectors with 32-bit and 64-bit elements.
33224 (vecdupssescalarmodesuffix): New mode attribute.
33225 (vec_dup<mode>): Use it.
33227 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
33230 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
33231 frame if argument is passed on stack.
33233 2018-01-11 Jakub Jelinek <jakub@redhat.com>
33236 * ree.c (combine_reaching_defs): Optimize also
33237 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
33238 reg2=any_extend(exp); reg1=reg2;, formatting fix.
33240 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
33242 PR middle-end/83189
33243 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
33245 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
33247 PR middle-end/83718
33248 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
33249 after they are computed.
33251 2018-01-11 Bin Cheng <bin.cheng@arm.com>
33253 PR tree-optimization/83695
33254 * gimple-loop-linterchange.cc
33255 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
33256 reset cached scev information after interchange.
33257 (pass_linterchange::execute): Remove call to scev_reset_htab.
33259 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33261 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
33262 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
33263 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
33264 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
33265 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
33266 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
33267 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
33268 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
33269 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
33270 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
33271 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
33272 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
33273 (V_lane_reg): Likewise.
33274 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
33276 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
33277 (vfmal_lane_low<mode>_intrinsic,
33278 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
33279 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
33280 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
33281 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
33282 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
33283 vfmsl_lane_high<mode>_intrinsic): New define_insns.
33285 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33287 * config/arm/arm-cpus.in (fp16fml): New feature.
33288 (ALL_SIMD): Add fp16fml.
33289 (armv8.2-a): Add fp16fml as an option.
33290 (armv8.3-a): Likewise.
33291 (armv8.4-a): Add fp16fml as part of fp16.
33292 * config/arm/arm.h (TARGET_FP16FML): Define.
33293 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
33295 * config/arm/arm-modes.def (V2HF): Define.
33296 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
33297 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
33298 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
33299 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
33300 vfmsl_low, vfmsl_high): New set of builtins.
33301 * config/arm/iterators.md (PLUSMINUS): New code iterator.
33302 (vfml_op): New code attribute.
33303 (VFMLHALVES): New int iterator.
33304 (VFML, VFMLSEL): New mode attributes.
33305 (V_reg): Define mapping for V2HF.
33306 (V_hi, V_lo): New mode attributes.
33307 (VF_constraint): Likewise.
33308 (vfml_half, vfml_half_selector): New int attributes.
33309 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
33311 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
33312 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
33314 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
33315 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
33316 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
33317 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
33319 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
33320 Document new effective target and option set.
33322 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33324 * config/arm/arm-cpus.in (armv8_4): New feature.
33325 (ARMv8_4a): New fgroup.
33326 (armv8.4-a): New arch.
33327 * config/arm/arm-tables.opt: Regenerate.
33328 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
33329 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
33330 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
33331 Add matching rules for -march=armv8.4-a and extensions.
33332 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
33334 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
33337 * config/rx/rx.md (BW): New mode attribute.
33338 (sync_lock_test_and_setsi): Add mode suffix to insn output.
33340 2018-01-11 Richard Biener <rguenther@suse.de>
33342 PR tree-optimization/83435
33343 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
33344 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
33345 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
33347 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33348 Alan Hayward <alan.hayward@arm.com>
33349 David Sherwood <david.sherwood@arm.com>
33351 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
33353 (aarch64_classify_address): Initialize it. Track polynomial offsets.
33354 (aarch64_print_address_internal): Use it to check for a zero offset.
33356 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33357 Alan Hayward <alan.hayward@arm.com>
33358 David Sherwood <david.sherwood@arm.com>
33360 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
33361 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
33362 Return a poly_int64 rather than a HOST_WIDE_INT.
33363 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
33364 rather than a HOST_WIDE_INT.
33365 * config/aarch64/aarch64.h (aarch64_frame): Protect with
33366 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
33367 hard_fp_offset, frame_size, initial_adjust, callee_offset and
33368 final_offset from HOST_WIDE_INT to poly_int64.
33369 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
33370 to_constant when getting the number of units in an Advanced SIMD
33372 (aarch64_builtin_vectorized_function): Check for a constant number
33374 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
33376 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
33377 attribute instead of GET_MODE_NUNITS.
33378 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
33379 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
33380 GET_MODE_SIZE for fixed-size registers.
33381 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
33382 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
33383 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
33384 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
33385 (aarch64_print_operand, aarch64_print_address_internal)
33386 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
33387 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
33388 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
33389 Handle polynomial GET_MODE_SIZE.
33390 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
33391 wider than SImode without modification.
33392 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
33393 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
33394 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
33395 passing and returning SVE modes.
33396 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
33397 rather than GEN_INT.
33398 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
33399 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
33400 (aarch64_allocate_and_probe_stack_space): Likewise.
33401 (aarch64_layout_frame): Cope with polynomial offsets.
33402 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
33403 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
33404 polynomial offsets.
33405 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
33406 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
33407 poly_int64 rather than a HOST_WIDE_INT.
33408 (aarch64_get_separate_components, aarch64_process_components)
33409 (aarch64_expand_prologue, aarch64_expand_epilogue)
33410 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
33411 (aarch64_anchor_offset): New function, split out from...
33412 (aarch64_legitimize_address): ...here.
33413 (aarch64_builtin_vectorization_cost): Handle polynomial
33414 TYPE_VECTOR_SUBPARTS.
33415 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
33417 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
33418 number of elements from the PARALLEL rather than the mode.
33419 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
33420 rather than GET_MODE_BITSIZE.
33421 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
33422 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
33423 (aarch64_expand_vec_perm_const_1): Handle polynomial
33424 d->perm.length () and d->perm elements.
33425 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
33426 Apply to_constant to d->perm elements.
33427 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
33428 polynomial CONST_VECTOR_NUNITS.
33429 (aarch64_move_pointer): Take amount as a poly_int64 rather
33431 (aarch64_progress_pointer): Avoid temporary variable.
33432 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
33433 the mode attribute instead of GET_MODE.
33435 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33436 Alan Hayward <alan.hayward@arm.com>
33437 David Sherwood <david.sherwood@arm.com>
33439 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
33440 x exists before using it.
33441 (aarch64_add_constant_internal): Rename to...
33442 (aarch64_add_offset_1): ...this. Replace regnum with separate
33443 src and dest rtxes. Handle the case in which they're different,
33444 including when the offset is zero. Replace scratchreg with an rtx.
33445 Use 2 additions if there is no spare register into which we can
33446 move a 16-bit constant.
33447 (aarch64_add_constant): Delete.
33448 (aarch64_add_offset): Replace reg with separate src and dest
33449 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
33450 Use aarch64_add_offset_1.
33451 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
33452 an rtx rather than an int. Take the delta as a poly_int64
33453 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
33454 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
33455 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
33456 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
33457 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
33458 and aarch64_add_sp.
33459 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
33460 aarch64_add_constant.
33462 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33464 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
33465 Use scalar_float_mode.
33467 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33469 * config/aarch64/aarch64-simd.md
33470 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
33471 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
33472 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
33473 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
33474 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
33475 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
33476 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
33477 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
33478 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
33479 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
33481 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33484 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
33485 targ_options->x_arm_arch_string is non NULL.
33487 2018-01-11 Tamar Christina <tamar.christina@arm.com>
33489 * config/aarch64/aarch64.h
33490 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
33492 2018-01-11 Sudakshina Das <sudi.das@arm.com>
33495 * expmed.c (emit_store_flag_force): Swap if const op0
33496 and change VOIDmode to mode of op0.
33498 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33500 PR rtl-optimization/83761
33501 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
33502 than bytes to mode_for_size.
33504 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
33506 PR middle-end/83189
33507 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
33508 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
33511 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
33513 PR middle-end/83575
33514 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
33515 when in layout mode.
33516 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
33517 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
33520 2018-01-10 Michael Collison <michael.collison@arm.com>
33522 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
33523 * config/aarch64/aarch64-option-extension.def: Add
33524 AARCH64_OPT_EXTENSION of 'fp16fml'.
33525 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33526 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
33527 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
33528 * config/aarch64/constraints.md (Ui7): New constraint.
33529 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
33530 (VFMLA_SEL_W): Ditto.
33533 (VFMLA16_LOW): New int iterator.
33534 (VFMLA16_HIGH): Ditto.
33535 (UNSPEC_FMLAL): New unspec.
33536 (UNSPEC_FMLSL): Ditto.
33537 (UNSPEC_FMLAL2): Ditto.
33538 (UNSPEC_FMLSL2): Ditto.
33539 (f16mac): New code attribute.
33540 * config/aarch64/aarch64-simd-builtins.def
33541 (aarch64_fmlal_lowv2sf): Ditto.
33542 (aarch64_fmlsl_lowv2sf): Ditto.
33543 (aarch64_fmlalq_lowv4sf): Ditto.
33544 (aarch64_fmlslq_lowv4sf): Ditto.
33545 (aarch64_fmlal_highv2sf): Ditto.
33546 (aarch64_fmlsl_highv2sf): Ditto.
33547 (aarch64_fmlalq_highv4sf): Ditto.
33548 (aarch64_fmlslq_highv4sf): Ditto.
33549 (aarch64_fmlal_lane_lowv2sf): Ditto.
33550 (aarch64_fmlsl_lane_lowv2sf): Ditto.
33551 (aarch64_fmlal_laneq_lowv2sf): Ditto.
33552 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
33553 (aarch64_fmlalq_lane_lowv4sf): Ditto.
33554 (aarch64_fmlsl_lane_lowv4sf): Ditto.
33555 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
33556 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
33557 (aarch64_fmlal_lane_highv2sf): Ditto.
33558 (aarch64_fmlsl_lane_highv2sf): Ditto.
33559 (aarch64_fmlal_laneq_highv2sf): Ditto.
33560 (aarch64_fmlsl_laneq_highv2sf): Ditto.
33561 (aarch64_fmlalq_lane_highv4sf): Ditto.
33562 (aarch64_fmlsl_lane_highv4sf): Ditto.
33563 (aarch64_fmlalq_laneq_highv4sf): Ditto.
33564 (aarch64_fmlsl_laneq_highv4sf): Ditto.
33565 * config/aarch64/aarch64-simd.md:
33566 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
33567 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
33568 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
33569 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
33570 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
33571 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
33572 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
33573 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
33574 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
33575 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
33576 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
33577 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
33578 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
33579 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
33580 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
33581 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
33582 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
33583 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
33584 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
33585 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
33586 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
33587 (vfmlsl_low_u32): Ditto.
33588 (vfmlalq_low_u32): Ditto.
33589 (vfmlslq_low_u32): Ditto.
33590 (vfmlal_high_u32): Ditto.
33591 (vfmlsl_high_u32): Ditto.
33592 (vfmlalq_high_u32): Ditto.
33593 (vfmlslq_high_u32): Ditto.
33594 (vfmlal_lane_low_u32): Ditto.
33595 (vfmlsl_lane_low_u32): Ditto.
33596 (vfmlal_laneq_low_u32): Ditto.
33597 (vfmlsl_laneq_low_u32): Ditto.
33598 (vfmlalq_lane_low_u32): Ditto.
33599 (vfmlslq_lane_low_u32): Ditto.
33600 (vfmlalq_laneq_low_u32): Ditto.
33601 (vfmlslq_laneq_low_u32): Ditto.
33602 (vfmlal_lane_high_u32): Ditto.
33603 (vfmlsl_lane_high_u32): Ditto.
33604 (vfmlal_laneq_high_u32): Ditto.
33605 (vfmlsl_laneq_high_u32): Ditto.
33606 (vfmlalq_lane_high_u32): Ditto.
33607 (vfmlslq_lane_high_u32): Ditto.
33608 (vfmlalq_laneq_high_u32): Ditto.
33609 (vfmlslq_laneq_high_u32): Ditto.
33610 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
33611 (AARCH64_FL_FOR_ARCH8_4): New.
33612 (AARCH64_ISA_F16FML): New ISA flag.
33613 (TARGET_F16FML): New feature flag for fp16fml.
33614 (doc/invoke.texi): Document new fp16fml option.
33616 2018-01-10 Michael Collison <michael.collison@arm.com>
33618 * config/aarch64/aarch64-builtins.c:
33619 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
33620 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33621 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
33622 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
33623 (AARCH64_ISA_SHA3): New ISA flag.
33624 (TARGET_SHA3): New feature flag for sha3.
33625 * config/aarch64/iterators.md (sha512_op): New int attribute.
33626 (CRYPTO_SHA512): New int iterator.
33627 (UNSPEC_SHA512H): New unspec.
33628 (UNSPEC_SHA512H2): Ditto.
33629 (UNSPEC_SHA512SU0): Ditto.
33630 (UNSPEC_SHA512SU1): Ditto.
33631 * config/aarch64/aarch64-simd-builtins.def
33632 (aarch64_crypto_sha512hqv2di): New builtin.
33633 (aarch64_crypto_sha512h2qv2di): Ditto.
33634 (aarch64_crypto_sha512su0qv2di): Ditto.
33635 (aarch64_crypto_sha512su1qv2di): Ditto.
33636 (aarch64_eor3qv8hi): Ditto.
33637 (aarch64_rax1qv2di): Ditto.
33638 (aarch64_xarqv2di): Ditto.
33639 (aarch64_bcaxqv8hi): Ditto.
33640 * config/aarch64/aarch64-simd.md:
33641 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
33642 (aarch64_crypto_sha512su0qv2di): Ditto.
33643 (aarch64_crypto_sha512su1qv2di): Ditto.
33644 (aarch64_eor3qv8hi): Ditto.
33645 (aarch64_rax1qv2di): Ditto.
33646 (aarch64_xarqv2di): Ditto.
33647 (aarch64_bcaxqv8hi): Ditto.
33648 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
33649 (vsha512h2q_u64): Ditto.
33650 (vsha512su0q_u64): Ditto.
33651 (vsha512su1q_u64): Ditto.
33652 (veor3q_u16): Ditto.
33653 (vrax1q_u64): Ditto.
33654 (vxarq_u64): Ditto.
33655 (vbcaxq_u16): Ditto.
33656 * config/arm/types.md (crypto_sha512): New type attribute.
33657 (crypto_sha3): Ditto.
33658 (doc/invoke.texi): Document new sha3 option.
33660 2018-01-10 Michael Collison <michael.collison@arm.com>
33662 * config/aarch64/aarch64-builtins.c:
33663 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
33664 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33665 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
33666 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
33667 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
33668 (AARCH64_ISA_SM4): New ISA flag.
33669 (TARGET_SM4): New feature flag for sm4.
33670 * config/aarch64/aarch64-simd-builtins.def
33671 (aarch64_sm3ss1qv4si): Ditto.
33672 (aarch64_sm3tt1aq4si): Ditto.
33673 (aarch64_sm3tt1bq4si): Ditto.
33674 (aarch64_sm3tt2aq4si): Ditto.
33675 (aarch64_sm3tt2bq4si): Ditto.
33676 (aarch64_sm3partw1qv4si): Ditto.
33677 (aarch64_sm3partw2qv4si): Ditto.
33678 (aarch64_sm4eqv4si): Ditto.
33679 (aarch64_sm4ekeyqv4si): Ditto.
33680 * config/aarch64/aarch64-simd.md:
33681 (aarch64_sm3ss1qv4si): Ditto.
33682 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
33683 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
33684 (aarch64_sm4eqv4si): Ditto.
33685 (aarch64_sm4ekeyqv4si): Ditto.
33686 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
33687 (sm3part_op): Ditto.
33688 (CRYPTO_SM3TT): Ditto.
33689 (CRYPTO_SM3PART): Ditto.
33690 (UNSPEC_SM3SS1): New unspec.
33691 (UNSPEC_SM3TT1A): Ditto.
33692 (UNSPEC_SM3TT1B): Ditto.
33693 (UNSPEC_SM3TT2A): Ditto.
33694 (UNSPEC_SM3TT2B): Ditto.
33695 (UNSPEC_SM3PARTW1): Ditto.
33696 (UNSPEC_SM3PARTW2): Ditto.
33697 (UNSPEC_SM4E): Ditto.
33698 (UNSPEC_SM4EKEY): Ditto.
33699 * config/aarch64/constraints.md (Ui2): New constraint.
33700 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
33701 * config/arm/types.md (crypto_sm3): New type attribute.
33702 (crypto_sm4): Ditto.
33703 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
33704 (vsm3tt1aq_u32): Ditto.
33705 (vsm3tt1bq_u32): Ditto.
33706 (vsm3tt2aq_u32): Ditto.
33707 (vsm3tt2bq_u32): Ditto.
33708 (vsm3partw1q_u32): Ditto.
33709 (vsm3partw2q_u32): Ditto.
33710 (vsm4eq_u32): Ditto.
33711 (vsm4ekeyq_u32): Ditto.
33712 (doc/invoke.texi): Document new sm4 option.
33714 2018-01-10 Michael Collison <michael.collison@arm.com>
33716 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
33717 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
33718 (AARCH64_FL_FOR_ARCH8_4): New.
33719 (AARCH64_FL_V8_4): New flag.
33720 (doc/invoke.texi): Document new armv8.4-a option.
33722 2018-01-10 Michael Collison <michael.collison@arm.com>
33724 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33725 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
33726 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
33727 * config/aarch64/aarch64-option-extension.def: Add
33728 AARCH64_OPT_EXTENSION of 'sha2'.
33729 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
33730 (crypto): Disable sha2 and aes if crypto disabled.
33731 (crypto): Enable aes and sha2 if enabled.
33732 (simd): Disable sha2 and aes if simd disabled.
33733 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
33735 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
33736 (TARGET_SHA2): New feature flag for sha2.
33737 (TARGET_AES): New feature flag for aes.
33738 * config/aarch64/aarch64-simd.md:
33739 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
33740 conditional on TARGET_AES.
33741 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
33742 (aarch64_crypto_sha1hsi): Make pattern conditional
33744 (aarch64_crypto_sha1hv4si): Ditto.
33745 (aarch64_be_crypto_sha1hv4si): Ditto.
33746 (aarch64_crypto_sha1su1v4si): Ditto.
33747 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
33748 (aarch64_crypto_sha1su0v4si): Ditto.
33749 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
33750 (aarch64_crypto_sha256su0v4si): Ditto.
33751 (aarch64_crypto_sha256su1v4si): Ditto.
33752 (doc/invoke.texi): Document new aes and sha2 options.
33754 2018-01-10 Martin Sebor <msebor@redhat.com>
33756 PR tree-optimization/83781
33757 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
33760 2018-01-11 Martin Sebor <msebor@gmail.com>
33761 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33763 PR tree-optimization/83501
33764 PR tree-optimization/81703
33766 * tree-ssa-strlen.c (get_string_cst): Rename...
33767 (get_string_len): ...to this. Handle global constants.
33768 (handle_char_store): Adjust.
33770 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
33771 Jim Wilson <jimw@sifive.com>
33773 * config/riscv/riscv-protos.h (riscv_output_return): New.
33774 * config/riscv/riscv.c (struct machine_function): New naked_p field.
33775 (riscv_attribute_table, riscv_output_return),
33776 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
33777 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
33778 (riscv_compute_frame_info): Only compute frame->mask if not a naked
33780 (riscv_expand_prologue): Add early return for naked function.
33781 (riscv_expand_epilogue): Likewise.
33782 (riscv_function_ok_for_sibcall): Return false for naked function.
33783 (riscv_set_current_function): New.
33784 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
33785 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
33786 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
33787 * doc/extend.texi (RISC-V Function Attributes): New.
33789 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
33791 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
33792 check for 128-bit long double before checking TCmode.
33793 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
33794 128-bit long doubles before checking TFmode or TCmode.
33795 (FLOAT128_IBM_P): Likewise.
33797 2018-01-10 Martin Sebor <msebor@redhat.com>
33799 PR tree-optimization/83671
33800 * builtins.c (c_strlen): Unconditionally return zero for the empty
33802 Use -Warray-bounds for warnings.
33803 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
33804 for non-constant array indices with COMPONENT_REF, arrays of
33805 arrays, and pointers to arrays.
33806 (gimple_fold_builtin_strlen): Determine and set length range for
33807 non-constant character arrays.
33809 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
33811 PR middle-end/81897
33812 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
33815 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
33817 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
33819 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
33822 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
33823 VECTOR_MEM_ALTIVEC_OR_VSX_P.
33824 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
33825 indexed_or_indirect_operand predicate.
33826 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
33827 (*vsx_le_perm_load_v8hi): Likewise.
33828 (*vsx_le_perm_load_v16qi): Likewise.
33829 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
33830 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
33831 (*vsx_le_perm_store_v8hi): Likewise.
33832 (*vsx_le_perm_store_v16qi): Likewise.
33833 (eight unnamed splitters): Likewise.
33835 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
33837 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
33838 * config/rs6000/emmintrin.h: Likewise.
33839 * config/rs6000/mmintrin.h: Likewise.
33840 * config/rs6000/xmmintrin.h: Likewise.
33842 2018-01-10 David Malcolm <dmalcolm@redhat.com>
33845 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
33847 * tree.c (tree_nop_conversion): Return true for location wrapper
33849 (maybe_wrap_with_location): New function.
33850 (selftest::check_strip_nops): New function.
33851 (selftest::test_location_wrappers): New function.
33852 (selftest::tree_c_tests): Call it.
33853 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
33854 (maybe_wrap_with_location): New decl.
33855 (EXPR_LOCATION_WRAPPER_P): New macro.
33856 (location_wrapper_p): New inline function.
33857 (tree_strip_any_location_wrapper): New inline function.
33859 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
33862 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
33863 stack_realign_offset for the largest alignment of stack slot
33865 (ix86_find_max_used_stack_alignment): New function.
33866 (ix86_finalize_stack_frame_flags): Use it. Set
33867 max_used_stack_alignment if we don't realign stack.
33868 * config/i386/i386.h (machine_function): Add
33869 max_used_stack_alignment.
33871 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
33873 * config/arm/arm.opt (-mbranch-cost): New option.
33874 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
33877 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
33880 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
33881 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
33883 2018-01-10 Richard Biener <rguenther@suse.de>
33886 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
33887 early out so it also covers the case where we have a non-NULL
33890 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
33892 PR tree-optimization/83753
33893 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
33894 for non-strided grouped accesses if the number of elements is 1.
33896 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
33899 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
33900 * i386.h (TARGET_USE_GATHER): Define.
33901 * x86-tune.def (X86_TUNE_USE_GATHER): New.
33903 2018-01-10 Martin Liska <mliska@suse.cz>
33906 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
33907 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
33909 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
33910 CLEANUP_NO_PARTITIONING is not set.
33912 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
33914 * doc/rtl.texi: Remove documentation of (const ...) wrappers
33915 for vectors, as a partial revert of r254296.
33916 * rtl.h (const_vec_p): Delete.
33917 (const_vec_duplicate_p): Don't test for vector CONSTs.
33918 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
33919 * expmed.c (make_tree): Likewise.
33922 * common.md (E, F): Use CONSTANT_P instead of checking for
33924 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
33925 checking for CONST_VECTOR.
33927 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
33929 PR middle-end/83575
33930 * predict.c (force_edge_cold): Handle in more sane way edges
33931 with no prediction.
33933 2018-01-09 Carl Love <cel@us.ibm.com>
33935 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
33937 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
33938 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
33939 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
33940 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
33941 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
33942 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
33943 * config/rs6000/rs6000-protos.h: Add extern defition for
33944 rs6000_generate_float2_double_code.
33945 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
33947 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
33948 (float2_v2df): Add define_expand.
33950 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
33953 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
33954 op_mode in the force_to_mode call.
33956 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
33958 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
33959 instead of checking each element individually.
33960 (aarch64_evpc_uzp): Likewise.
33961 (aarch64_evpc_zip): Likewise.
33962 (aarch64_evpc_ext): Likewise.
33963 (aarch64_evpc_rev): Likewise.
33964 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
33965 instead of checking each element individually. Return true without
33967 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
33968 whether all selected elements come from the same input, instead of
33969 checking each element individually. Remove calls to gen_rtx_REG,
33970 start_sequence and end_sequence and instead assert that no rtl is
33973 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
33975 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
33976 order of HIGH and CONST checks.
33978 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
33980 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
33981 if the destination isn't an SSA_NAME.
33983 2018-01-09 Richard Biener <rguenther@suse.de>
33985 PR tree-optimization/83668
33986 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
33988 (canonicalize_loop_form): ... here, renamed from ...
33989 (canonicalize_loop_closed_ssa_form): ... this and amended to
33990 swap successor edges for loop exit blocks to make us use
33991 the RPO order we need for initial schedule generation.
33993 2018-01-09 Joseph Myers <joseph@codesourcery.com>
33995 PR tree-optimization/64811
33996 * match.pd: When optimizing comparisons with Inf, avoid
33997 introducing or losing exceptions from comparisons with NaN.
33999 2018-01-09 Martin Liska <mliska@suse.cz>
34002 * asan.c (shadow_mem_size): Add gcc_assert.
34004 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
34006 Don't save registers in main().
34009 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
34010 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
34011 * config/avr/avr.c (avr_set_current_function): Don't error if
34012 naked, OS_task or OS_main are specified at the same time.
34013 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
34015 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
34017 * common/config/avr/avr-common.c (avr_option_optimization_table):
34018 Switch on -mmain-is-OS_task for optimizing compilations.
34020 2018-01-09 Richard Biener <rguenther@suse.de>
34022 PR tree-optimization/83572
34023 * graphite.c: Include cfganal.h.
34024 (graphite_transform_loops): Connect infinite loops to exit
34025 and remove fake edges at the end.
34027 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
34029 * ipa-inline.c (edge_badness): Revert accidental checkin.
34031 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
34034 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
34035 symbols; not inline clones.
34037 2018-01-09 Jakub Jelinek <jakub@redhat.com>
34040 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
34041 hard registers. Formatting fixes.
34043 PR preprocessor/83722
34044 * gcc.c (try_generate_repro): Pass
34045 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
34046 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
34049 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
34050 Kito Cheng <kito.cheng@gmail.com>
34052 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
34053 (riscv_leaf_function_p): Delete.
34054 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
34056 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
34058 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
34060 (do_ifelse): New function.
34061 (do_isel): New function.
34062 (do_sub3): New function.
34063 (do_add3): New function.
34064 (do_load_mask_compare): New function.
34065 (do_overlap_load_compare): New function.
34066 (expand_compare_loop): New function.
34067 (expand_block_compare): Call expand_compare_loop() when appropriate.
34068 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
34069 option description.
34070 (-mblock-compare-inline-loop-limit): New option.
34072 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
34075 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
34076 Reverse order of second and third operands in first alternative.
34077 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
34078 of first and second elements in UNSPEC_VPERMR vector.
34079 (altivec_expand_vec_perm_le): Likewise.
34081 2018-01-08 Jeff Law <law@redhat.com>
34083 PR rtl-optimizatin/81308
34084 * tree-switch-conversion.c (cfg_altered): New file scoped static.
34085 (process_switch): If group_case_labels makes a change, then set
34087 (pass_convert_switch::execute): If a switch is converted, then
34088 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
34090 PR rtl-optimization/81308
34091 * recog.c (split_all_insns): Conditionally cleanup the CFG after
34094 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
34096 PR target/83663 - Revert r255946
34097 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
34098 generation for cases where splatting a value is not useful.
34099 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
34100 across a vec_duplicate and a paradoxical subreg forming a vector
34101 mode to a vec_concat.
34103 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34105 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
34106 -march=armv8.3-a variants.
34107 * config/arm/t-multilib: Likewise.
34108 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
34110 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
34112 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
34114 (cceq_ior_compare_complement): Give it a name so I can use it, and
34115 change boolean_or_operator predicate to boolean_operator so it can
34116 be used to generate a crand.
34117 (eqne): New code iterator.
34118 (bd/bd_neg): New code_attrs.
34119 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
34120 a single define_insn.
34121 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
34122 decrement (bdnzt/bdnzf/bdzt/bdzf).
34123 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
34124 with the new names of the branch decrement patterns, and added the
34125 names of the branch decrement conditional patterns.
34127 2018-01-08 Richard Biener <rguenther@suse.de>
34129 PR tree-optimization/83563
34130 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
34133 2018-01-08 Richard Biener <rguenther@suse.de>
34135 PR middle-end/83713
34136 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
34138 2018-01-08 Richard Biener <rguenther@suse.de>
34140 PR tree-optimization/83685
34141 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
34142 references to abnormals.
34144 2018-01-08 Richard Biener <rguenther@suse.de>
34147 * dwarf2out.c (output_indirect_strings): Handle empty
34148 skeleton_debug_str_hash.
34149 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
34151 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
34153 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
34154 (emit_store_direct): Likewise.
34155 (arc_trampoline_adjust_address): Likewise.
34156 (arc_asm_trampoline_template): New function.
34157 (arc_initialize_trampoline): Use asm_trampoline_template.
34158 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
34159 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
34160 * config/arc/arc.md (flush_icache): Delete pattern.
34162 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
34164 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
34165 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
34168 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
34171 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
34172 by not USED_FOR_TARGET.
34173 (make_pass_resolve_sw_modes): Likewise.
34175 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
34177 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
34180 2018-01-08 Richard Biener <rguenther@suse.de>
34182 PR middle-end/83580
34183 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
34185 2018-01-08 Richard Biener <rguenther@suse.de>
34187 PR middle-end/83517
34188 * match.pd ((t * 2) / 2) -> t): Add missing :c.
34190 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
34192 PR middle-end/81897
34193 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
34194 basic blocks with a small number of successors.
34195 (convert_control_dep_chain_into_preds): Improve handling of
34197 (dump_predicates): Split apart into...
34198 (dump_pred_chain): ...here...
34199 (dump_pred_info): ...and here.
34200 (can_one_predicate_be_invalidated_p): Add debugging printfs.
34201 (can_chain_union_be_invalidated_p): Improve check for invalidation
34203 (uninit_uses_cannot_happen): Avoid unnecessary if
34204 convert_control_dep_chain_into_preds yielded nothing.
34206 2018-01-06 Martin Sebor <msebor@redhat.com>
34208 PR tree-optimization/83640
34209 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
34210 subtracting negative offset from size.
34211 (builtin_access::overlap): Adjust offset bounds of the access to fall
34212 within the size of the object if possible.
34214 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
34216 PR rtl-optimization/83699
34217 * expmed.c (extract_bit_field_1): Restrict the vector usage of
34218 extract_bit_field_as_subreg to cases in which the extracted
34219 value is also a vector.
34221 * lra-constraints.c (process_alt_operands): Test for the equivalence
34222 substitutions when detecting a possible reload cycle.
34224 2018-01-06 Jakub Jelinek <jakub@redhat.com>
34227 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
34228 by default if flag_selective_schedling{,2}. Formatting fixes.
34230 PR rtl-optimization/83682
34231 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
34232 if it has non-VECTOR_MODE element mode.
34233 (vec_duplicate_p): Likewise.
34235 PR middle-end/83694
34236 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
34237 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
34239 2018-01-05 Jakub Jelinek <jakub@redhat.com>
34242 * config/i386/i386-builtin.def
34243 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
34244 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
34245 Require also OPTION_MASK_ISA_AVX512F in addition to
34246 OPTION_MASK_ISA_GFNI.
34247 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
34248 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
34249 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
34250 to OPTION_MASK_ISA_GFNI.
34251 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
34252 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
34253 OPTION_MASK_ISA_AVX512BW.
34254 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
34255 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
34256 addition to OPTION_MASK_ISA_GFNI.
34257 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
34258 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
34259 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
34260 to OPTION_MASK_ISA_GFNI.
34261 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
34262 a requirement for all ISAs rather than any of them with a few
34264 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
34266 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
34267 bitmasks to be enabled with 3 exceptions, instead of requiring any
34268 enabled ISA with lots of exceptions.
34269 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
34270 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
34271 Change avx512bw in isa attribute to avx512f.
34272 * config/i386/sgxintrin.h: Add license boilerplate.
34273 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
34274 to __AVX512F__ and __AVX512VL to __AVX512VL__.
34275 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
34276 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
34278 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
34279 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
34280 temporarily sse2 rather than sse if not enabled already.
34283 * config/i386/sse.md (VI248_VLBW): Rename to ...
34284 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
34285 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
34286 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
34287 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
34288 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
34289 mode iterator instead of VI248_VLBW.
34291 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
34293 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
34294 (record_modified): Skip clobbers; add debug output.
34295 (param_change_prob): Use sreal frequencies.
34297 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
34299 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
34300 punt for user-aligned variables.
34302 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
34304 * tree-chrec.c (chrec_contains_symbols): Return true for
34307 2018-01-05 Sudakshina Das <sudi.das@arm.com>
34310 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
34311 of (x|y) == x for BICS pattern.
34313 2018-01-05 Jakub Jelinek <jakub@redhat.com>
34315 PR tree-optimization/83605
34316 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
34317 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
34320 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
34322 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
34323 * config/epiphany/rtems.h: New file.
34325 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34326 Uros Bizjak <ubizjak@gmail.com>
34329 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
34330 QIreg_operand instead of register_operand predicate.
34331 * config/i386/i386.c (ix86_rop_should_change_byte_p,
34332 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
34333 comments instead of -fmitigate[-_]rop.
34335 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
34338 * cgraphunit.c (symbol_table::compile): Switch to text_section
34339 before calling assembly_start debug hook.
34340 * run-rtl-passes.c (run_rtl_passes): Likewise.
34343 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34345 * tree-vrp.c (extract_range_from_binary_expr_1): Check
34346 range_int_cst_p rather than !symbolic_range_p before calling
34347 extract_range_from_multiplicative_op_1.
34349 2018-01-04 Jeff Law <law@redhat.com>
34351 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
34352 redundant test in assertion.
34354 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34356 * doc/rtl.texi: Document machine_mode wrapper classes.
34358 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34360 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
34361 using tree_to_uhwi.
34363 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34365 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
34366 the VEC_PERM_EXPR fold to fail.
34368 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34371 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
34372 to switched_sections.
34374 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34377 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
34378 test for d.testing.
34380 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
34383 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
34384 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
34386 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34389 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
34390 is BLKmode and bitpos not zero or mode change is needed.
34392 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34395 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
34398 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
34401 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
34402 instead of MULT rtx. Update all corresponding splitters.
34403 (*saddl_se): Ditto.
34404 (*ssub<modesuffix>): Ditto.
34405 (*ssubl_se): Ditto.
34406 (*cmp_sadd_di): Update split patterns.
34407 (*cmp_sadd_si): Ditto.
34408 (*cmp_sadd_sidi): Ditto.
34409 (*cmp_ssub_di): Ditto.
34410 (*cmp_ssub_si): Ditto.
34411 (*cmp_ssub_sidi): Ditto.
34412 * config/alpha/predicates.md (const23_operand): New predicate.
34413 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
34414 Look for ASHIFT, not MULT inner operand.
34415 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
34417 2018-01-04 Martin Liska <mliska@suse.cz>
34419 PR gcov-profile/83669
34420 * gcov.c (output_intermediate_file): Add version to intermediate
34422 * doc/gcov.texi: Document new field 'version' in intermediate
34423 file format. Fix location of '-k' option of gcov command.
34425 2018-01-04 Martin Liska <mliska@suse.cz>
34428 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
34430 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34432 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
34434 2018-01-03 Martin Sebor <msebor@redhat.com>
34436 PR tree-optimization/83655
34437 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
34438 checking calls with invalid arguments.
34440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34442 * tree-vect-stmts.c (vect_get_store_rhs): New function.
34443 (vectorizable_mask_load_store): Delete.
34444 (vectorizable_call): Return false for masked loads and stores.
34445 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
34446 instead of gimple_assign_rhs1.
34447 (vectorizable_load): Handle IFN_MASK_LOAD.
34448 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
34450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34452 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
34454 (vectorizable_mask_load_store): ...here.
34455 (vectorizable_load): ...and here.
34457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34459 * tree-vect-stmts.c (vect_build_all_ones_mask)
34460 (vect_build_zero_merge_argument): New functions, split out from...
34461 (vectorizable_load): ...here.
34463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34465 * tree-vect-stmts.c (vect_check_store_rhs): New function,
34467 (vectorizable_mask_load_store): ...here.
34468 (vectorizable_store): ...and here.
34470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34472 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
34474 (vectorizable_mask_load_store): ...here.
34476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34478 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
34479 (vect_model_store_cost): Take a vec_load_store_type instead of a
34481 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
34482 (vect_model_store_cost): Take a vec_load_store_type instead of a
34484 (vectorizable_mask_load_store): Update accordingly.
34485 (vectorizable_store): Likewise.
34486 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
34488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34490 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
34491 IFN_MASK_LOAD calls here rather than...
34492 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
34494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34495 Alan Hayward <alan.hayward@arm.com>
34496 David Sherwood <david.sherwood@arm.com>
34498 * expmed.c (extract_bit_field_1): For vector extracts,
34499 fall back to extract_bit_field_as_subreg if vec_extract
34502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34503 Alan Hayward <alan.hayward@arm.com>
34504 David Sherwood <david.sherwood@arm.com>
34506 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
34507 they are variable or constant sized.
34508 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
34509 slots for constant-sized data.
34511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34512 Alan Hayward <alan.hayward@arm.com>
34513 David Sherwood <david.sherwood@arm.com>
34515 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
34516 handling COND_EXPRs with boolean comparisons, try to find a better
34517 basis for the mask type than the boolean itself.
34519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34521 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
34522 is calculated and how it can be overridden.
34523 * genmodes.c (max_bitsize_mode_any_mode): New variable.
34524 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
34526 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
34529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34530 Alan Hayward <alan.hayward@arm.com>
34531 David Sherwood <david.sherwood@arm.com>
34533 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
34534 Remove the mode argument.
34535 (aarch64_simd_valid_immediate): Remove the mode and inverse
34537 * config/aarch64/iterators.md (bitsize): New iterator.
34538 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
34539 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
34540 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
34541 aarch64_simd_valid_immediate.
34542 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
34543 (aarch64_reg_or_bic_imm): Likewise.
34544 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
34545 with an insn_type enum and msl with a modifier_type enum.
34546 Replace element_width with a scalar_mode. Change the shift
34547 to unsigned int. Add constructors for scalar_float_mode and
34548 scalar_int_mode elements.
34549 (aarch64_vect_float_const_representable_p): Delete.
34550 (aarch64_can_const_movi_rtx_p)
34551 (aarch64_simd_scalar_immediate_valid_for_move)
34552 (aarch64_simd_make_constant): Update call to
34553 aarch64_simd_valid_immediate.
34554 (aarch64_advsimd_valid_immediate_hs): New function.
34555 (aarch64_advsimd_valid_immediate): Likewise.
34556 (aarch64_simd_valid_immediate): Remove mode and inverse
34557 arguments. Rewrite to use the above. Use const_vec_duplicate_p
34558 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
34559 and aarch64_float_const_representable_p on the result.
34560 (aarch64_output_simd_mov_immediate): Remove mode argument.
34561 Update call to aarch64_simd_valid_immediate and use of
34562 simd_immediate_info.
34563 (aarch64_output_scalar_simd_mov_immediate): Update call
34566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34567 Alan Hayward <alan.hayward@arm.com>
34568 David Sherwood <david.sherwood@arm.com>
34570 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
34571 (mode_nunits): Likewise CONST_MODE_NUNITS.
34572 * machmode.def (ADJUST_NUNITS): Document.
34573 * genmodes.c (mode_data::need_nunits_adj): New field.
34574 (blank_mode): Update accordingly.
34575 (adj_nunits): New variable.
34576 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
34578 (emit_mode_size_inline): Set need_bytesize_adj for all modes
34579 listed in adj_nunits.
34580 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
34581 listed in adj_nunits. Don't emit case statements for such modes.
34582 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
34583 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
34584 nothing if adj_nunits is nonnull.
34585 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
34586 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
34587 (emit_mode_fbit): Update use of print_maybe_const_decl.
34588 (emit_move_size): Likewise. Treat the array as non-const
34590 (emit_mode_adjustments): Handle adj_nunits.
34592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34594 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
34595 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
34596 (VECTOR_MODES): Use it.
34597 (make_vector_modes): Take the prefix as an argument.
34599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34600 Alan Hayward <alan.hayward@arm.com>
34601 David Sherwood <david.sherwood@arm.com>
34603 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
34604 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
34605 for MODE_VECTOR_BOOL.
34606 * machmode.def (VECTOR_BOOL_MODE): Document.
34607 * genmodes.c (VECTOR_BOOL_MODE): New macro.
34608 (make_vector_bool_mode): New function.
34609 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
34611 * lto-streamer-in.c (lto_input_mode_table): Likewise.
34612 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
34614 * stor-layout.c (int_mode_for_mode): Likewise.
34615 * tree.c (build_vector_type_for_mode): Likewise.
34616 * varasm.c (output_constant_pool_2): Likewise.
34617 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
34618 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
34619 for MODE_VECTOR_BOOL.
34620 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
34621 of mode class checks.
34622 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
34623 instead of a list of mode class checks.
34624 (expand_vector_scalar_condition): Likewise.
34625 (type_for_widest_vector_mode): Handle BImode as an inner mode.
34627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34628 Alan Hayward <alan.hayward@arm.com>
34629 David Sherwood <david.sherwood@arm.com>
34631 * machmode.h (mode_size): Change from unsigned short to
34633 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
34634 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34635 or if measurement_type is not polynomial.
34636 (fixed_size_mode::includes_p): Check for constant-sized modes.
34637 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
34638 return a poly_uint16 rather than an unsigned short.
34639 (emit_mode_size): Change the type of mode_size from unsigned short
34640 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
34641 (emit_mode_adjustments): Cope with polynomial vector sizes.
34642 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34644 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34646 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
34647 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
34648 * caller-save.c (setup_save_areas): Likewise.
34649 (replace_reg_with_saved_mem): Likewise.
34650 * calls.c (emit_library_call_value_1): Likewise.
34651 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
34652 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
34653 (gen_lowpart_for_combine): Likewise.
34654 * convert.c (convert_to_integer_1): Likewise.
34655 * cse.c (equiv_constant, cse_insn): Likewise.
34656 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
34657 (cselib_subst_to_values): Likewise.
34658 * dce.c (word_dce_process_block): Likewise.
34659 * df-problems.c (df_word_lr_mark_ref): Likewise.
34660 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
34661 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
34662 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
34663 (rtl_for_decl_location): Likewise.
34664 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
34665 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
34666 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
34667 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
34668 (expand_expr_real_1): Likewise.
34669 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
34670 (pad_below): Likewise.
34671 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34672 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
34673 * ira.c (get_subreg_tracking_sizes): Likewise.
34674 * ira-build.c (ira_create_allocno_objects): Likewise.
34675 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
34676 (ira_sort_regnos_for_alter_reg): Likewise.
34677 * ira-costs.c (record_operand_costs): Likewise.
34678 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
34679 (resolve_simple_move): Likewise.
34680 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
34681 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
34682 (lra_constraints): Likewise.
34683 (CONST_POOL_OK_P): Reject variable-sized modes.
34684 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
34685 (add_pseudo_to_slot, lra_spill): Likewise.
34686 * omp-low.c (omp_clause_aligned_alignment): Likewise.
34687 * optabs-query.c (get_best_extraction_insn): Likewise.
34688 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
34689 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
34690 (expand_mult_highpart, valid_multiword_target_p): Likewise.
34691 * recog.c (offsettable_address_addr_space_p): Likewise.
34692 * regcprop.c (maybe_mode_change): Likewise.
34693 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
34694 * regrename.c (build_def_use): Likewise.
34695 * regstat.c (dump_reg_info): Likewise.
34696 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
34697 (find_reloads, find_reloads_subreg_address): Likewise.
34698 * reload1.c (eliminate_regs_1): Likewise.
34699 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
34700 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
34701 (simplify_binary_operation_1, simplify_subreg): Likewise.
34702 * targhooks.c (default_function_arg_padding): Likewise.
34703 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
34704 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
34705 (verify_gimple_assign_ternary): Likewise.
34706 * tree-inline.c (estimate_move_cost): Likewise.
34707 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34708 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
34709 (get_address_cost_ainc): Likewise.
34710 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
34711 (vect_supportable_dr_alignment): Likewise.
34712 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
34713 (vectorizable_reduction): Likewise.
34714 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
34715 (vectorizable_operation, vectorizable_load): Likewise.
34716 * tree.c (build_same_sized_truth_vector_type): Likewise.
34717 * valtrack.c (cleanup_auto_inc_dec): Likewise.
34718 * var-tracking.c (emit_note_insn_var_location): Likewise.
34719 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
34720 (ADDR_VEC_ALIGN): Likewise.
34722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34723 Alan Hayward <alan.hayward@arm.com>
34724 David Sherwood <david.sherwood@arm.com>
34726 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
34728 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34729 or if measurement_type is polynomial.
34730 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
34731 * combine.c (make_extraction): Likewise.
34732 * dse.c (find_shift_sequence): Likewise.
34733 * dwarf2out.c (mem_loc_descriptor): Likewise.
34734 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
34735 (extract_bit_field, extract_low_bits): Likewise.
34736 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
34737 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
34738 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
34739 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
34740 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34741 * reload.c (find_reloads): Likewise.
34742 * reload1.c (alter_reg): Likewise.
34743 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
34744 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
34745 * tree-if-conv.c (predicate_mem_writes): Likewise.
34746 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
34747 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
34748 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
34749 * valtrack.c (dead_debug_insert_temp): Likewise.
34750 * varasm.c (mergeable_constant_section): Likewise.
34751 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
34753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34754 Alan Hayward <alan.hayward@arm.com>
34755 David Sherwood <david.sherwood@arm.com>
34757 * expr.c (expand_assignment): Cope with polynomial mode sizes
34758 when assigning to a CONCAT.
34760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34761 Alan Hayward <alan.hayward@arm.com>
34762 David Sherwood <david.sherwood@arm.com>
34764 * machmode.h (mode_precision): Change from unsigned short to
34766 (mode_to_precision): Return a poly_uint16 rather than an unsigned
34768 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
34769 or if measurement_type is not polynomial.
34770 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
34771 in which the mode is already known to be a scalar_int_mode.
34772 * genmodes.c (emit_mode_precision): Change the type of mode_precision
34773 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
34775 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34776 for GET_MODE_PRECISION.
34777 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34778 for GET_MODE_PRECISION.
34779 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
34781 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
34782 (expand_field_assignment, make_extraction): Likewise.
34783 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
34784 (get_last_value): Likewise.
34785 * convert.c (convert_to_integer_1): Likewise.
34786 * cse.c (cse_insn): Likewise.
34787 * expr.c (expand_expr_real_1): Likewise.
34788 * lra-constraints.c (simplify_operand_subreg): Likewise.
34789 * optabs-query.c (can_atomic_load_p): Likewise.
34790 * optabs.c (expand_atomic_load): Likewise.
34791 (expand_atomic_store): Likewise.
34792 * ree.c (combine_reaching_defs): Likewise.
34793 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
34794 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
34795 * tree.h (type_has_mode_precision_p): Likewise.
34796 * ubsan.c (instrument_si_overflow): Likewise.
34798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34799 Alan Hayward <alan.hayward@arm.com>
34800 David Sherwood <david.sherwood@arm.com>
34802 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
34803 polynomial numbers of units.
34804 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
34805 (valid_vector_subparts_p): New function.
34806 (build_vector_type): Remove temporary shim and take the number
34807 of units as a poly_uint64 rather than an int.
34808 (build_opaque_vector_type): Take the number of units as a
34809 poly_uint64 rather than an int.
34810 * tree.c (build_vector_from_ctor): Handle polynomial
34811 TYPE_VECTOR_SUBPARTS.
34812 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
34813 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
34814 (build_vector_from_val): If the number of units is variable,
34815 use build_vec_duplicate_cst for constant operands and
34816 VEC_DUPLICATE_EXPR otherwise.
34817 (make_vector_type): Remove temporary is_constant ().
34818 (build_vector_type, build_opaque_vector_type): Take the number of
34819 units as a poly_uint64 rather than an int.
34820 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
34822 * cfgexpand.c (expand_debug_expr): Likewise.
34823 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
34824 (store_constructor, expand_expr_real_1): Likewise.
34825 (const_scalar_mask_from_tree): Likewise.
34826 * fold-const-call.c (fold_const_reduction): Likewise.
34827 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
34828 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
34829 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
34830 (fold_relational_const): Likewise.
34831 (native_interpret_vector): Likewise. Change the size from an
34832 int to an unsigned int.
34833 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
34834 TYPE_VECTOR_SUBPARTS.
34835 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
34836 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
34837 duplicating a non-constant operand into a variable-length vector.
34838 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
34839 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
34840 * ipa-icf.c (sem_variable::equals): Likewise.
34841 * match.pd: Likewise.
34842 * omp-simd-clone.c (simd_clone_subparts): Likewise.
34843 * print-tree.c (print_node): Likewise.
34844 * stor-layout.c (layout_type): Likewise.
34845 * targhooks.c (default_builtin_vectorization_cost): Likewise.
34846 * tree-cfg.c (verify_gimple_comparison): Likewise.
34847 (verify_gimple_assign_binary): Likewise.
34848 (verify_gimple_assign_ternary): Likewise.
34849 (verify_gimple_assign_single): Likewise.
34850 * tree-pretty-print.c (dump_generic_node): Likewise.
34851 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34852 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
34853 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
34854 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
34855 (vect_shift_permute_load_chain): Likewise.
34856 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
34857 (expand_vector_condition, optimize_vector_constructor): Likewise.
34858 (lower_vec_perm, get_compute_type): Likewise.
34859 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
34860 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
34861 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
34862 (vect_recog_mask_conversion_pattern): Likewise.
34863 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
34864 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
34865 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34866 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
34867 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
34868 (vectorizable_shift, vectorizable_operation, vectorizable_store)
34869 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
34870 (supportable_widening_operation): Likewise.
34871 (supportable_narrowing_operation): Likewise.
34872 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
34874 * varasm.c (output_constant): Likewise.
34876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34877 Alan Hayward <alan.hayward@arm.com>
34878 David Sherwood <david.sherwood@arm.com>
34880 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
34881 so that both the length == 3 and length != 3 cases set up their
34882 own permute vectors. Add comments explaining why we know the
34883 number of elements is constant.
34884 (vect_permute_load_chain): Likewise.
34886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34887 Alan Hayward <alan.hayward@arm.com>
34888 David Sherwood <david.sherwood@arm.com>
34890 * machmode.h (mode_nunits): Change from unsigned char to
34892 (ONLY_FIXED_SIZE_MODES): New macro.
34893 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
34894 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
34895 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
34897 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
34898 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
34899 or if measurement_type is not polynomial.
34900 * genmodes.c (ZERO_COEFFS): New macro.
34901 (emit_mode_nunits_inline): Make mode_nunits_inline return a
34903 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
34904 Use ZERO_COEFFS when emitting initializers.
34905 * data-streamer.h (bp_pack_poly_value): New function.
34906 (bp_unpack_poly_value): Likewise.
34907 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34908 for GET_MODE_NUNITS.
34909 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34910 for GET_MODE_NUNITS.
34911 * tree.c (make_vector_type): Remove temporary shim and make
34912 the real function take the number of units as a poly_uint64
34913 rather than an int.
34914 (build_vector_type_for_mode): Handle polynomial nunits.
34915 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
34916 * emit-rtl.c (const_vec_series_p_1): Likewise.
34917 (gen_rtx_CONST_VECTOR): Likewise.
34918 * fold-const.c (test_vec_duplicate_folding): Likewise.
34919 * genrecog.c (validate_pattern): Likewise.
34920 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
34921 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
34922 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
34923 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
34924 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
34925 * rtlanal.c (subreg_get_info): Likewise.
34926 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
34927 (vect_grouped_load_supported): Likewise.
34928 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
34929 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
34930 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
34931 (simplify_const_unary_operation, simplify_binary_operation_1)
34932 (simplify_const_binary_operation, simplify_ternary_operation)
34933 (test_vector_ops_duplicate, test_vector_ops): Likewise.
34934 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
34935 instead of CONST_VECTOR_NUNITS.
34936 * varasm.c (output_constant_pool_2): Likewise.
34937 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
34938 explicit-encoded elements in the XVEC for variable-length vectors.
34940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34942 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
34944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34945 Alan Hayward <alan.hayward@arm.com>
34946 David Sherwood <david.sherwood@arm.com>
34948 * coretypes.h (fixed_size_mode): Declare.
34949 (fixed_size_mode_pod): New typedef.
34950 * builtins.h (target_builtins::x_apply_args_mode)
34951 (target_builtins::x_apply_result_mode): Change type to
34952 fixed_size_mode_pod.
34953 * builtins.c (apply_args_size, apply_result_size, result_vector)
34954 (expand_builtin_apply_args_1, expand_builtin_apply)
34955 (expand_builtin_return): Update accordingly.
34957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34959 * cse.c (hash_rtx_cb): Hash only the encoded elements.
34960 * cselib.c (cselib_hash_rtx): Likewise.
34961 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
34962 CONST_VECTOR encoding.
34964 2018-01-03 Jakub Jelinek <jakub@redhat.com>
34965 Jeff Law <law@redhat.com>
34968 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
34969 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
34970 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
34971 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
34974 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
34975 explicitly probe *sp in a noreturn function if there were any callee
34976 register saves or frame pointer is needed.
34978 2018-01-03 Jakub Jelinek <jakub@redhat.com>
34981 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
34982 BLKmode for ternary, binary or unary expressions.
34985 * var-tracking.c (delete_vta_debug_insn): New inline function.
34986 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
34987 insns from get_insns () to NULL instead of each bb separately.
34988 Use delete_vta_debug_insn. No longer static.
34989 (vt_debug_insns_local, variable_tracking_main_1): Adjust
34990 delete_vta_debug_insns callers.
34991 * rtl.h (delete_vta_debug_insns): Declare.
34992 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
34993 instead of variable_tracking_main.
34995 2018-01-03 Martin Sebor <msebor@redhat.com>
34997 PR tree-optimization/83603
34998 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
34999 arguments past the endof the argument list in functions declared
35000 without a prototype.
35001 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
35002 Avoid checking when arguments are null.
35004 2018-01-03 Martin Sebor <msebor@redhat.com>
35007 * doc/extend.texi (attribute const): Fix a typo.
35008 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
35009 issuing -Wsuggest-attribute for void functions.
35011 2018-01-03 Martin Sebor <msebor@redhat.com>
35013 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
35014 offset_int::from instead of wide_int::to_shwi.
35015 (maybe_diag_overlap): Remove assertion.
35016 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
35017 * gimple-ssa-sprintf.c (format_directive): Same.
35018 (parse_directive): Same.
35019 (sprintf_dom_walker::compute_format_length): Same.
35020 (try_substitute_return_value): Same.
35022 2018-01-03 Jeff Law <law@redhat.com>
35024 PR middle-end/83654
35025 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
35026 non-constant residual for zero at runtime and avoid probing in
35027 that case. Reorganize code for trailing problem to mirror handling
35030 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
35032 PR tree-optimization/83501
35033 * tree-ssa-strlen.c (get_string_cst): New.
35034 (handle_char_store): Call get_string_cst.
35036 2018-01-03 Martin Liska <mliska@suse.cz>
35038 PR tree-optimization/83593
35039 * tree-ssa-strlen.c: Include tree-cfg.h.
35040 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
35041 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
35042 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
35044 (strlen_dom_walker::before_dom_children): Call
35045 gimple_purge_dead_eh_edges. Dump tranformation with details
35047 (strlen_dom_walker::before_dom_children): Update call by adding
35048 new argument cleanup_eh.
35049 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
35051 2018-01-03 Martin Liska <mliska@suse.cz>
35054 * cif-code.def (VARIADIC_THUNK): New enum value.
35055 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
35058 2018-01-03 Jan Beulich <jbeulich@suse.com>
35060 * sse.md (mov<mode>_internal): Tighten condition for when to use
35061 vmovdqu<ssescalarsize> for TI and OI modes.
35063 2018-01-03 Jakub Jelinek <jakub@redhat.com>
35065 Update copyright years.
35067 2018-01-03 Martin Liska <mliska@suse.cz>
35070 * ipa-visibility.c (function_and_variable_visibility): Skip
35071 functions with noipa attribure.
35073 2018-01-03 Jakub Jelinek <jakub@redhat.com>
35075 * gcc.c (process_command): Update copyright notice dates.
35076 * gcov-dump.c (print_version): Ditto.
35077 * gcov.c (print_version): Ditto.
35078 * gcov-tool.c (print_version): Ditto.
35079 * gengtype.c (create_file): Ditto.
35080 * doc/cpp.texi: Bump @copying's copyright year.
35081 * doc/cppinternals.texi: Ditto.
35082 * doc/gcc.texi: Ditto.
35083 * doc/gccint.texi: Ditto.
35084 * doc/gcov.texi: Ditto.
35085 * doc/install.texi: Ditto.
35086 * doc/invoke.texi: Ditto.
35088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35090 * vector-builder.h (vector_builder::m_full_nelts): Change from
35091 unsigned int to poly_uint64.
35092 (vector_builder::full_nelts): Update prototype accordingly.
35093 (vector_builder::new_vector): Likewise.
35094 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
35095 (vector_builder::operator ==): Likewise.
35096 (vector_builder::finalize): Likewise.
35097 * int-vector-builder.h (int_vector_builder::int_vector_builder):
35098 Take the number of elements as a poly_uint64 rather than an
35100 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
35101 from unsigned int to poly_uint64.
35102 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
35103 (vec_perm_indices::new_vector): Likewise.
35104 (vec_perm_indices::length): Likewise.
35105 (vec_perm_indices::nelts_per_input): Likewise.
35106 (vec_perm_indices::input_nelts): Likewise.
35107 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
35108 number of elements per input as a poly_uint64 rather than an
35109 unsigned int. Use the original encoding for variable-length
35110 vectors, rather than clamping each individual element.
35111 For the second and subsequent elements in each pattern,
35112 clamp the step and base before clamping their sum.
35113 (vec_perm_indices::series_p): Handle polynomial element counts.
35114 (vec_perm_indices::all_in_range_p): Likewise.
35115 (vec_perm_indices_to_tree): Likewise.
35116 (vec_perm_indices_to_rtx): Likewise.
35117 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
35118 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
35119 (tree_vector_builder::new_binary_operation): Handle polynomial
35120 element counts. Return false if we need to know the number
35121 of elements at compile time.
35122 * fold-const.c (fold_vec_perm): Punt if the number of elements
35123 isn't known at compile time.
35125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35127 * vec-perm-indices.h (vec_perm_builder): Change element type
35128 from HOST_WIDE_INT to poly_int64.
35129 (vec_perm_indices::element_type): Update accordingly.
35130 (vec_perm_indices::clamp): Handle polynomial element_types.
35131 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
35132 (vec_perm_indices::all_in_range_p): Likewise.
35133 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
35135 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
35136 polynomial vec_perm_indices element types.
35137 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
35138 * fold-const.c (fold_vec_perm): Likewise.
35139 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
35140 * tree-vect-generic.c (lower_vec_perm): Likewise.
35141 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
35142 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
35143 element type to HOST_WIDE_INT.
35145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35146 Alan Hayward <alan.hayward@arm.com>
35147 David Sherwood <david.sherwood@arm.com>
35149 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
35150 rather than an int. Use plus_constant.
35151 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
35152 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
35154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35155 Alan Hayward <alan.hayward@arm.com>
35156 David Sherwood <david.sherwood@arm.com>
35158 * calls.c (emit_call_1, expand_call): Change struct_value_size from
35159 a HOST_WIDE_INT to a poly_int64.
35161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35162 Alan Hayward <alan.hayward@arm.com>
35163 David Sherwood <david.sherwood@arm.com>
35165 * calls.c (load_register_parameters): Cope with polynomial
35166 mode sizes. Require a constant size for BLKmode parameters
35167 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
35168 forces a parameter to be padded at the lsb end in order to
35169 fill a complete number of words, require the parameter size
35170 to be ordered wrt UNITS_PER_WORD.
35172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35173 Alan Hayward <alan.hayward@arm.com>
35174 David Sherwood <david.sherwood@arm.com>
35176 * reload1.c (spill_stack_slot_width): Change element type
35177 from unsigned int to poly_uint64_pod.
35178 (alter_reg): Treat mode sizes as polynomial.
35180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35181 Alan Hayward <alan.hayward@arm.com>
35182 David Sherwood <david.sherwood@arm.com>
35184 * reload.c (complex_word_subreg_p): New function.
35185 (reload_inner_reg_of_subreg, push_reload): Use it.
35187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35188 Alan Hayward <alan.hayward@arm.com>
35189 David Sherwood <david.sherwood@arm.com>
35191 * lra-constraints.c (process_alt_operands): Reject matched
35192 operands whose sizes aren't ordered.
35193 (match_reload): Refer to this check here.
35195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35196 Alan Hayward <alan.hayward@arm.com>
35197 David Sherwood <david.sherwood@arm.com>
35199 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
35200 that the mode size is in the set {1, 2, 4, 8, 16}.
35202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35203 Alan Hayward <alan.hayward@arm.com>
35204 David Sherwood <david.sherwood@arm.com>
35206 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
35207 Use plus_constant instead of gen_rtx_PLUS.
35209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35210 Alan Hayward <alan.hayward@arm.com>
35211 David Sherwood <david.sherwood@arm.com>
35213 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
35214 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
35215 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
35216 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
35217 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
35218 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
35219 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
35220 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
35221 * config/i386/i386.c (ix86_push_rounding): ...this new function.
35222 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
35224 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
35225 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
35226 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
35227 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
35228 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
35229 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
35230 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
35231 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
35232 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
35233 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
35235 * expr.c (emit_move_resolve_push): Treat the input and result
35236 of PUSH_ROUNDING as a poly_int64.
35237 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
35238 (emit_push_insn): Likewise.
35239 * lra-eliminations.c (mark_not_eliminable): Likewise.
35240 * recog.c (push_operand): Likewise.
35241 * reload1.c (elimination_effects): Likewise.
35242 * rtlanal.c (nonzero_bits1): Likewise.
35243 * calls.c (store_one_arg): Likewise. Require the padding to be
35244 known at compile time.
35246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35247 Alan Hayward <alan.hayward@arm.com>
35248 David Sherwood <david.sherwood@arm.com>
35250 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
35251 Use plus_constant instead of gen_rtx_PLUS.
35253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35254 Alan Hayward <alan.hayward@arm.com>
35255 David Sherwood <david.sherwood@arm.com>
35257 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
35258 rather than an int.
35260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35261 Alan Hayward <alan.hayward@arm.com>
35262 David Sherwood <david.sherwood@arm.com>
35264 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
35265 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
35266 via stack temporaries. Treat the mode size as polynomial too.
35268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35269 Alan Hayward <alan.hayward@arm.com>
35270 David Sherwood <david.sherwood@arm.com>
35272 * expr.c (expand_expr_real_2): When handling conversions involving
35273 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
35274 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
35275 as a poly_uint64 too.
35277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35278 Alan Hayward <alan.hayward@arm.com>
35279 David Sherwood <david.sherwood@arm.com>
35281 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
35283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35284 Alan Hayward <alan.hayward@arm.com>
35285 David Sherwood <david.sherwood@arm.com>
35287 * combine.c (can_change_dest_mode): Handle polynomial
35288 REGMODE_NATURAL_SIZE.
35289 * expmed.c (store_bit_field_1): Likewise.
35290 * expr.c (store_constructor): Likewise.
35291 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
35292 and polynomial REGMODE_NATURAL_SIZE.
35293 (gen_lowpart_common): Likewise.
35294 * reginfo.c (record_subregs_of_mode): Likewise.
35295 * rtlanal.c (read_modify_subreg_p): Likewise.
35297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35298 Alan Hayward <alan.hayward@arm.com>
35299 David Sherwood <david.sherwood@arm.com>
35301 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
35302 numbers of elements.
35304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35305 Alan Hayward <alan.hayward@arm.com>
35306 David Sherwood <david.sherwood@arm.com>
35308 * match.pd: Cope with polynomial numbers of vector elements.
35310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35311 Alan Hayward <alan.hayward@arm.com>
35312 David Sherwood <david.sherwood@arm.com>
35314 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
35315 in a POINTER_PLUS_EXPR.
35317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35318 Alan Hayward <alan.hayward@arm.com>
35319 David Sherwood <david.sherwood@arm.com>
35321 * omp-simd-clone.c (simd_clone_subparts): New function.
35322 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
35323 (ipa_simd_modify_function_body): Likewise.
35325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35326 Alan Hayward <alan.hayward@arm.com>
35327 David Sherwood <david.sherwood@arm.com>
35329 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
35330 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
35331 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
35332 (expand_vector_condition, vector_element): Likewise.
35333 (subparts_gt): New function.
35334 (get_compute_type): Use subparts_gt.
35335 (count_type_subparts): Delete.
35336 (expand_vector_operations_1): Use subparts_gt instead of
35337 count_type_subparts.
35339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35340 Alan Hayward <alan.hayward@arm.com>
35341 David Sherwood <david.sherwood@arm.com>
35343 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
35344 (vect_compile_time_alias): ...this new function. Do the calculation
35345 on poly_ints rather than trees.
35346 (vect_prune_runtime_alias_test_list): Update call accordingly.
35348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35349 Alan Hayward <alan.hayward@arm.com>
35350 David Sherwood <david.sherwood@arm.com>
35352 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
35354 (vect_schedule_slp_instance): Likewise.
35356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35357 Alan Hayward <alan.hayward@arm.com>
35358 David Sherwood <david.sherwood@arm.com>
35360 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
35361 constant and extern definitions for variable-length vectors.
35362 (vect_get_constant_vectors): Note that the number of units
35363 is known to be constant.
35365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35366 Alan Hayward <alan.hayward@arm.com>
35367 David Sherwood <david.sherwood@arm.com>
35369 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
35370 of units as polynomial. Choose between WIDE and NARROW based
35373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35374 Alan Hayward <alan.hayward@arm.com>
35375 David Sherwood <david.sherwood@arm.com>
35377 * tree-vect-stmts.c (simd_clone_subparts): New function.
35378 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
35380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35381 Alan Hayward <alan.hayward@arm.com>
35382 David Sherwood <david.sherwood@arm.com>
35384 * tree-vect-stmts.c (vectorizable_call): Treat the number of
35385 vectors as polynomial. Use build_index_vector for
35386 IFN_GOMP_SIMD_LANE.
35388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35389 Alan Hayward <alan.hayward@arm.com>
35390 David Sherwood <david.sherwood@arm.com>
35392 * tree-vect-stmts.c (get_load_store_type): Treat the number of
35393 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
35394 for variable-length vectors.
35395 (vectorizable_mask_load_store): Treat the number of units as
35396 polynomial, asserting that it is constant if the condition has
35397 already been enforced.
35398 (vectorizable_store, vectorizable_load): Likewise.
35400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35401 Alan Hayward <alan.hayward@arm.com>
35402 David Sherwood <david.sherwood@arm.com>
35404 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
35405 of units as polynomial. Punt if we can't tell at compile time
35406 which vector contains the final result.
35408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35409 Alan Hayward <alan.hayward@arm.com>
35410 David Sherwood <david.sherwood@arm.com>
35412 * tree-vect-loop.c (vectorizable_induction): Treat the number
35413 of units as polynomial. Punt on SLP inductions. Use an integer
35414 VEC_SERIES_EXPR for variable-length integer reductions. Use a
35415 cast of such a series for variable-length floating-point
35418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35419 Alan Hayward <alan.hayward@arm.com>
35420 David Sherwood <david.sherwood@arm.com>
35422 * tree.h (build_index_vector): Declare.
35423 * tree.c (build_index_vector): New function.
35424 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
35425 of units as polynomial, forcibly converting it to a constant if
35426 vectorizable_reduction has already enforced the condition.
35427 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
35428 to create a {1,2,3,...} vector.
35429 (vectorizable_reduction): Treat the number of units as polynomial.
35430 Choose vectype_in based on the largest scalar element size rather
35431 than the smallest number of units. Enforce the restrictions
35434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35435 Alan Hayward <alan.hayward@arm.com>
35436 David Sherwood <david.sherwood@arm.com>
35438 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
35439 number of units as polynomial.
35441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35442 Alan Hayward <alan.hayward@arm.com>
35443 David Sherwood <david.sherwood@arm.com>
35445 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
35446 * target.def (autovectorize_vector_sizes): Return the vector sizes
35447 by pointer, using vector_sizes rather than a bitmask.
35448 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
35449 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
35450 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
35452 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
35453 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
35454 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
35455 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
35456 * omp-general.c (omp_max_vf): Likewise.
35457 * omp-low.c (omp_clause_aligned_alignment): Likewise.
35458 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
35459 * tree-vect-loop.c (vect_analyze_loop): Likewise.
35460 * tree-vect-slp.c (vect_slp_bb): Likewise.
35461 * doc/tm.texi: Regenerate.
35462 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
35464 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
35465 the vector size as a poly_uint64 rather than an unsigned int.
35466 (current_vector_size): Change from an unsigned int to a poly_uint64.
35467 (get_vectype_for_scalar_type): Update accordingly.
35468 * tree.h (build_truth_vector_type): Take the size and number of
35469 units as a poly_uint64 rather than an unsigned int.
35470 (build_vector_type): Add a temporary overload that takes
35471 the number of units as a poly_uint64 rather than an unsigned int.
35472 * tree.c (make_vector_type): Likewise.
35473 (build_truth_vector_type): Take the number of units as a poly_uint64
35474 rather than an unsigned int.
35476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35477 Alan Hayward <alan.hayward@arm.com>
35478 David Sherwood <david.sherwood@arm.com>
35480 * target.def (get_mask_mode): Take the number of units and length
35481 as poly_uint64s rather than unsigned ints.
35482 * targhooks.h (default_get_mask_mode): Update accordingly.
35483 * targhooks.c (default_get_mask_mode): Likewise.
35484 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
35485 * doc/tm.texi: Regenerate.
35487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35488 Alan Hayward <alan.hayward@arm.com>
35489 David Sherwood <david.sherwood@arm.com>
35491 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
35492 * omp-general.c (omp_max_vf): Likewise.
35493 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
35494 (expand_omp_simd): Handle polynomial safelen.
35495 * omp-low.c (omplow_simd_context): Add a default constructor.
35496 (omplow_simd_context::max_vf): Change from int to poly_uint64.
35497 (lower_rec_simd_input_clauses): Update accordingly.
35498 (lower_rec_input_clauses): Likewise.
35500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35501 Alan Hayward <alan.hayward@arm.com>
35502 David Sherwood <david.sherwood@arm.com>
35504 * tree-vectorizer.h (vect_nunits_for_cost): New function.
35505 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
35506 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
35507 (vect_analyze_slp_cost): Likewise.
35508 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
35509 (vect_model_load_cost): Likewise.
35511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35512 Alan Hayward <alan.hayward@arm.com>
35513 David Sherwood <david.sherwood@arm.com>
35515 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
35516 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
35517 from an unsigned int * to a poly_uint64_pod *.
35518 (calculate_unrolling_factor): New function.
35519 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
35521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35522 Alan Hayward <alan.hayward@arm.com>
35523 David Sherwood <david.sherwood@arm.com>
35525 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
35526 from an unsigned int to a poly_uint64.
35527 (_loop_vec_info::slp_unrolling_factor): Likewise.
35528 (_loop_vec_info::vectorization_factor): Change from an int
35530 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
35531 (vect_get_num_vectors): New function.
35532 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
35533 (vect_get_num_copies): Use vect_get_num_vectors.
35534 (vect_analyze_data_ref_dependences): Change max_vf from an int *
35535 to an unsigned int *.
35536 (vect_analyze_data_refs): Change min_vf from an int * to a
35538 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
35539 than an unsigned HOST_WIDE_INT.
35540 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
35541 (vect_analyze_data_ref_dependence): Change max_vf from an int *
35542 to an unsigned int *.
35543 (vect_analyze_data_ref_dependences): Likewise.
35544 (vect_compute_data_ref_alignment): Handle polynomial vf.
35545 (vect_enhance_data_refs_alignment): Likewise.
35546 (vect_prune_runtime_alias_test_list): Likewise.
35547 (vect_shift_permute_load_chain): Likewise.
35548 (vect_supportable_dr_alignment): Likewise.
35549 (dependence_distance_ge_vf): Take the vectorization factor as a
35550 poly_uint64 rather than an unsigned HOST_WIDE_INT.
35551 (vect_analyze_data_refs): Change min_vf from an int * to a
35553 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
35554 vfm1 as a poly_uint64 rather than an int. Make the same change
35555 for the returned bound_scalar.
35556 (vect_gen_vector_loop_niters): Handle polynomial vf.
35557 (vect_do_peeling): Likewise. Update call to
35558 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
35559 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
35561 * tree-vect-loop.c (vect_determine_vectorization_factor)
35562 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
35563 (vect_get_known_peeling_cost): Likewise.
35564 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
35565 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
35566 (vect_transform_loop): Likewise. Use the lowest possible VF when
35567 updating the upper bounds of the loop.
35568 (vect_min_worthwhile_factor): Make static. Return an unsigned int
35569 rather than an int.
35570 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
35571 polynomial unroll factors.
35572 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
35573 (vect_make_slp_decision): Likewise.
35574 (vect_supported_load_permutation_p): Likewise, and polynomial
35576 (vect_analyze_slp_cost): Handle polynomial vf.
35577 (vect_slp_analyze_node_operations): Likewise.
35578 (vect_slp_analyze_bb_1): Likewise.
35579 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
35580 than an unsigned HOST_WIDE_INT.
35581 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
35582 (vectorizable_load): Handle polynomial vf.
35583 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
35585 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
35587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35588 Alan Hayward <alan.hayward@arm.com>
35589 David Sherwood <david.sherwood@arm.com>
35591 * match.pd: Handle bit operations involving three constants
35592 and try to fold one pair.
35594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35596 * tree-vect-loop-manip.c: Include gimple-fold.h.
35597 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
35598 niters_maybe_zero parameters. Handle other cases besides a step of 1.
35599 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
35600 Add a path that uses a step of VF instead of 1, but disable it
35602 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
35603 and niters_no_overflow parameters. Update calls to
35604 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
35605 Create a new SSA name if the latter choses to use a ste other
35606 than zero, and return it via niters_vector_mult_vf_var.
35607 * tree-vect-loop.c (vect_transform_loop): Update calls to
35608 vect_do_peeling, vect_gen_vector_loop_niters and
35609 slpeel_make_loop_iterate_ntimes.
35610 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
35611 (vect_gen_vector_loop_niters): Update declarations after above changes.
35613 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
35615 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
35616 128-bit round to integer instructions.
35617 (ceil<mode>2): Likewise.
35618 (btrunc<mode>2): Likewise.
35619 (round<mode>2): Likewise.
35621 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
35623 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
35624 unaligned VSX load/store on P8/P9.
35625 (expand_block_clear): Allow the use of unaligned VSX
35626 load/store on P8/P9.
35628 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35630 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
35632 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
35633 swap associated with both a load and a store.
35635 2018-01-02 Andrew Waterman <andrew@sifive.com>
35637 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
35638 * config/riscv/riscv.md (clear_cache): Use it.
35640 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
35642 * web.c: Remove out-of-date comment.
35644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35646 * expr.c (fixup_args_size_notes): Check that any existing
35647 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
35648 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
35649 (emit_single_push_insn): ...here.
35651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35653 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
35654 (const_vector_encoded_nelts): New function.
35655 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
35656 (const_vector_int_elt, const_vector_elt): Declare.
35657 * emit-rtl.c (const_vector_int_elt_1): New function.
35658 (const_vector_elt): Likewise.
35659 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
35660 of CONST_VECTOR_ELT.
35662 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35664 * expr.c: Include rtx-vector-builder.h.
35665 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
35666 directly on the tree encoding.
35667 (const_vector_from_tree): Likewise.
35668 * optabs.c: Include rtx-vector-builder.h.
35669 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
35670 sequence of "u" values.
35671 * vec-perm-indices.c: Include rtx-vector-builder.h.
35672 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
35673 directly on the vec_perm_indices encoding.
35675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35677 * doc/rtl.texi (const_vector): Describe new encoding scheme.
35678 * Makefile.in (OBJS): Add rtx-vector-builder.o.
35679 * rtx-vector-builder.h: New file.
35680 * rtx-vector-builder.c: Likewise.
35681 * rtl.h (rtx_def::u2): Add a const_vector field.
35682 (CONST_VECTOR_NPATTERNS): New macro.
35683 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
35684 (CONST_VECTOR_DUPLICATE_P): Likewise.
35685 (CONST_VECTOR_STEPPED_P): Likewise.
35686 (CONST_VECTOR_ENCODED_ELT): Likewise.
35687 (const_vec_duplicate_p): Check for a duplicated vector encoding.
35688 (unwrap_const_vec_duplicate): Likewise.
35689 (const_vec_series_p): Check for a non-duplicated vector encoding.
35690 Say that the function only returns true for integer vectors.
35691 * emit-rtl.c: Include rtx-vector-builder.h.
35692 (gen_const_vec_duplicate_1): Delete.
35693 (gen_const_vector): Call gen_const_vec_duplicate instead of
35694 gen_const_vec_duplicate_1.
35695 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
35696 (gen_const_vec_duplicate): Use rtx_vector_builder.
35697 (gen_const_vec_series): Likewise.
35698 (gen_rtx_CONST_VECTOR): Likewise.
35699 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
35700 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35701 Build a new vector rather than modifying a CONST_VECTOR in-place.
35702 (handle_special_swappables): Update call accordingly.
35703 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
35704 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35705 Build a new vector rather than modifying a CONST_VECTOR in-place.
35706 (handle_special_swappables): Update call accordingly.
35708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35710 * simplify-rtx.c (simplify_const_binary_operation): Use
35711 CONST_VECTOR_ELT instead of XVECEXP.
35713 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35715 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
35716 the selector elements to be different from the data elements
35717 if the selector is a VECTOR_CST.
35718 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
35719 ssizetype for the selector.
35721 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35723 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
35724 before testing each element individually.
35725 * tree-vect-generic.c (lower_vec_perm): Likewise.
35727 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35729 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
35730 * selftest-run-tests.c (selftest::run_tests): Call it.
35731 * vector-builder.h (vector_builder::operator ==): New function.
35732 (vector_builder::operator !=): Likewise.
35733 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
35734 (vec_perm_indices::all_from_input_p): New function.
35735 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
35736 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
35737 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
35738 instead of reading the VECTOR_CST directly. Detect whether both
35739 vector inputs are the same before constructing the vec_perm_indices,
35740 and update the number of inputs argument accordingly. Use the
35741 utility functions added above. Only construct sel2 if we need to.
35743 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35745 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
35746 the broadcast of the low byte.
35747 (expand_mult_highpart): Use an explicit encoding for the permutes.
35748 * optabs-query.c (can_mult_highpart_p): Likewise.
35749 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
35750 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35751 (vectorizable_bswap): Likewise.
35752 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
35753 explicit encoding for the power-of-2 permutes.
35754 (vect_permute_store_chain): Likewise.
35755 (vect_grouped_load_supported): Likewise.
35756 (vect_permute_load_chain): Likewise.
35758 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35760 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
35761 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
35762 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
35763 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
35764 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
35765 (vect_gen_perm_mask_any): Likewise.
35767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35769 * int-vector-builder.h: New file.
35770 * vec-perm-indices.h: Include int-vector-builder.h.
35771 (vec_perm_indices): Redefine as an int_vector_builder.
35772 (auto_vec_perm_indices): Delete.
35773 (vec_perm_builder): Redefine as a stand-alone class.
35774 (vec_perm_indices::vec_perm_indices): New function.
35775 (vec_perm_indices::clamp): Likewise.
35776 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
35777 (vec_perm_indices::new_vector): New function.
35778 (vec_perm_indices::new_expanded_vector): Update for new
35779 vec_perm_indices class.
35780 (vec_perm_indices::rotate_inputs): New function.
35781 (vec_perm_indices::all_in_range_p): Operate directly on the
35782 encoded form, without computing elided elements.
35783 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
35784 encoding. Update for new vec_perm_indices class.
35785 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
35786 the given vec_perm_builder.
35787 (expand_vec_perm_var): Update vec_perm_builder constructor.
35788 (expand_mult_highpart): Use vec_perm_builder instead of
35789 auto_vec_perm_indices.
35790 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
35791 vec_perm_indices instead of auto_vec_perm_indices. Use a single
35792 or double series encoding as appropriate.
35793 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
35794 vec_perm_indices instead of auto_vec_perm_indices.
35795 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
35796 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
35797 (vect_permute_store_chain): Likewise.
35798 (vect_grouped_load_supported): Likewise.
35799 (vect_permute_load_chain): Likewise.
35800 (vect_shift_permute_load_chain): Likewise.
35801 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
35802 (vect_transform_slp_perm_load): Likewise.
35803 (vect_schedule_slp_instance): Likewise.
35804 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35805 (vectorizable_mask_load_store): Likewise.
35806 (vectorizable_bswap): Likewise.
35807 (vectorizable_store): Likewise.
35808 (vectorizable_load): Likewise.
35809 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
35810 vec_perm_indices instead of auto_vec_perm_indices. Use
35811 tree_to_vec_perm_builder to read the vector from a tree.
35812 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
35813 vec_perm_builder instead of a vec_perm_indices.
35814 (have_whole_vector_shift): Use vec_perm_builder and
35815 vec_perm_indices instead of auto_vec_perm_indices. Leave the
35816 truncation to calc_vec_perm_mask_for_shift.
35817 (vect_create_epilog_for_reduction): Likewise.
35818 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
35819 from auto_vec_perm_indices to vec_perm_indices.
35820 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
35821 instead of changing individual elements.
35822 (aarch64_vectorize_vec_perm_const): Use new_vector to install
35823 the vector in d.perm.
35824 * config/arm/arm.c (expand_vec_perm_d::perm): Change
35825 from auto_vec_perm_indices to vec_perm_indices.
35826 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
35827 instead of changing individual elements.
35828 (arm_vectorize_vec_perm_const): Use new_vector to install
35829 the vector in d.perm.
35830 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
35831 Update vec_perm_builder constructor.
35832 (rs6000_expand_interleave): Likewise.
35833 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
35834 (rs6000_expand_interleave): Likewise.
35836 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35838 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
35839 to qimode could truncate the indices.
35840 * optabs.c (expand_vec_perm_var): Likewise.
35842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35844 * Makefile.in (OBJS): Add vec-perm-indices.o.
35845 * vec-perm-indices.h: New file.
35846 * vec-perm-indices.c: Likewise.
35847 * target.h (vec_perm_indices): Replace with a forward class
35849 (auto_vec_perm_indices): Move to vec-perm-indices.h.
35850 * optabs.h: Include vec-perm-indices.h.
35851 (expand_vec_perm): Delete.
35852 (selector_fits_mode_p, expand_vec_perm_var): Declare.
35853 (expand_vec_perm_const): Declare.
35854 * target.def (vec_perm_const_ok): Replace with...
35855 (vec_perm_const): ...this new hook.
35856 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
35857 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
35858 * doc/tm.texi: Regenerate.
35859 * optabs.def (vec_perm_const): Delete.
35860 * doc/md.texi (vec_perm_const): Likewise.
35861 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
35862 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
35863 expand_vec_perm for constant permutation vectors. Assert that
35864 the mode of variable permutation vectors is the integer equivalent
35865 of the mode that is being permuted.
35866 * optabs-query.h (selector_fits_mode_p): Declare.
35867 * optabs-query.c: Include vec-perm-indices.h.
35868 (selector_fits_mode_p): New function.
35869 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
35870 is defined, instead of checking whether the vec_perm_const_optab
35871 exists. Use targetm.vectorize.vec_perm_const instead of
35872 targetm.vectorize.vec_perm_const_ok. Check whether the indices
35873 fit in the vector mode before using a variable permute.
35874 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
35875 vec_perm_indices instead of an rtx.
35876 (expand_vec_perm): Replace with...
35877 (expand_vec_perm_const): ...this new function. Take the selector
35878 as a vec_perm_indices rather than an rtx. Also take the mode of
35879 the selector. Update call to shift_amt_for_vec_perm_mask.
35880 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
35881 Use vec_perm_indices::new_expanded_vector to expand the original
35882 selector into bytes. Check whether the indices fit in the vector
35883 mode before using a variable permute.
35884 (expand_vec_perm_var): Make global.
35885 (expand_mult_highpart): Use expand_vec_perm_const.
35886 * fold-const.c: Includes vec-perm-indices.h.
35887 * tree-ssa-forwprop.c: Likewise.
35888 * tree-vect-data-refs.c: Likewise.
35889 * tree-vect-generic.c: Likewise.
35890 * tree-vect-loop.c: Likewise.
35891 * tree-vect-slp.c: Likewise.
35892 * tree-vect-stmts.c: Likewise.
35893 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
35895 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
35896 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
35897 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
35898 (aarch64_vectorize_vec_perm_const): ...this new function.
35899 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35900 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35901 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
35902 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
35903 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35904 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35905 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
35907 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
35908 check for NEON modes.
35909 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
35910 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
35911 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
35912 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
35914 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
35915 the old VEC_PERM_CONST conditions.
35916 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
35917 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
35918 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
35919 (ia64_vectorize_vec_perm_const_ok): Merge into...
35920 (ia64_vectorize_vec_perm_const): ...this new function.
35921 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
35922 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
35923 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
35924 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
35925 * config/mips/mips.c (mips_expand_vec_perm_const)
35926 (mips_vectorize_vec_perm_const_ok): Merge into...
35927 (mips_vectorize_vec_perm_const): ...this new function.
35928 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
35929 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
35930 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
35931 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
35932 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
35933 (rs6000_expand_vec_perm_const): Delete.
35934 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
35936 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35937 (altivec_expand_vec_perm_const_le): Take each operand individually.
35938 Operate on constant selectors rather than rtxes.
35939 (altivec_expand_vec_perm_const): Likewise. Update call to
35940 altivec_expand_vec_perm_const_le.
35941 (rs6000_expand_vec_perm_const): Delete.
35942 (rs6000_vectorize_vec_perm_const_ok): Delete.
35943 (rs6000_vectorize_vec_perm_const): New function.
35944 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
35945 an element count and rtx array.
35946 (rs6000_expand_extract_even): Update call accordingly.
35947 (rs6000_expand_interleave): Likewise.
35948 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
35949 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
35950 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
35951 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
35952 (rs6000_expand_vec_perm_const): Delete.
35953 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35954 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35955 (altivec_expand_vec_perm_const_le): Take each operand individually.
35956 Operate on constant selectors rather than rtxes.
35957 (altivec_expand_vec_perm_const): Likewise. Update call to
35958 altivec_expand_vec_perm_const_le.
35959 (rs6000_expand_vec_perm_const): Delete.
35960 (rs6000_vectorize_vec_perm_const_ok): Delete.
35961 (rs6000_vectorize_vec_perm_const): New function. Remove stray
35962 reference to the SPE evmerge intructions.
35963 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
35964 an element count and rtx array.
35965 (rs6000_expand_extract_even): Update call accordingly.
35966 (rs6000_expand_interleave): Likewise.
35967 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
35968 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
35970 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35974 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
35975 vector mode and that that mode matches the mode of the data
35977 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
35978 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
35979 directly using expand_vec_perm_1 when forcing selectors into
35981 (expand_vec_perm_var): New function, split out from expand_vec_perm.
35983 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35985 * optabs-query.h (can_vec_perm_p): Delete.
35986 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
35987 * optabs-query.c (can_vec_perm_p): Split into...
35988 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
35989 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
35990 particular selector is valid.
35991 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
35992 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
35993 (vect_grouped_load_supported): Likewise.
35994 (vect_shift_permute_load_chain): Likewise.
35995 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
35996 (vect_transform_slp_perm_load): Likewise.
35997 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35998 (vectorizable_bswap): Likewise.
35999 (vect_gen_perm_mask_checked): Likewise.
36000 * fold-const.c (fold_ternary_loc): Likewise. Don't take
36001 implementations of variable permutation vectors into account
36002 when deciding which selector to use.
36003 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
36004 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
36005 with a false third argument.
36006 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
36007 to test whether the constant selector is valid and can_vec_perm_var_p
36008 to test whether a variable selector is valid.
36010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
36012 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
36013 * optabs-query.c (can_vec_perm_p): Likewise.
36014 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
36015 instead of vec_perm_indices.
36016 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
36017 (vect_gen_perm_mask_checked): Likewise,
36018 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
36019 (vect_gen_perm_mask_checked): Likewise,
36021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
36023 * optabs-query.h (qimode_for_vec_perm): Declare.
36024 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
36025 (qimode_for_vec_perm): ...this new function.
36026 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
36028 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
36030 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
36031 does not have a conditional at the top.
36033 2018-01-02 Richard Biener <rguenther@suse.de>
36035 * ipa-inline.c (big_speedup_p): Fix expression.
36037 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
36040 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
36043 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
36047 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
36048 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
36049 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
36050 cond_taken_branch_cost 3->4.
36052 2018-01-01 Jakub Jelinek <jakub@redhat.com>
36054 PR tree-optimization/83581
36055 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
36056 TODO_cleanup_cfg if any changes have been made.
36058 PR middle-end/83608
36059 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
36060 convert_modes if target mode has the right side, but different mode
36063 PR middle-end/83609
36064 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
36065 last argument when extracting from CONCAT. If either from_real or
36066 from_imag is NULL, use expansion through memory. If result is not
36067 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
36068 the parts directly to inner mode, if even that fails, use expansion
36071 PR middle-end/83623
36072 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
36073 check for bswap in mode rather than HImode and use that in expand_unop
36076 Copyright (C) 2018 Free Software Foundation, Inc.
36078 Copying and distribution of this file, with or without modification,
36079 are permitted in any medium without royalty provided the copyright
36080 notice and this notice are preserved.