1 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
3 * reg-stack.c: Include regs.h.
4 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
5 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6 FIRST_STACK_REG, not DFmode.
7 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
8 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
9 (change stack): Default register mode to the reg_raw_mode of
10 FIRST_STACK_REG, not DFmode.
11 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
12 (*swapxf): Rename from swapxf.
14 2018-09-14 Carl Love <cel@us.ibm.com>
16 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
17 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
19 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
22 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
25 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
28 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
30 2018-09-14 Jason Merrill <jason@redhat.com>
32 Fix --enable-gather-detailed-mem-stats.
33 * hash-table.c (hash_table_usage): Change from variable to function.
34 * hash-table.h: Adjust.
35 * Makefile.in: Add missing dependencies on hash-table.h.
37 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
39 PR tree-optimization/87259
41 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
42 execute_cse_reciprocals_1 has tried transforming.
44 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
46 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
47 VR_VARYING for PLUS/MINUS_EXPR.
49 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
51 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
54 2018-09-14 Richard Biener <rguenther@suse.de>
57 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
58 bits for the merged partition.
60 2018-09-13 Martin Sebor <msebor@redhat.com>
61 Bernd Edlinger <bernd.edlinger@hotmail.de>
63 * builtins.h (c_srlen): Add argument.
64 * builtins.c (warn_string_no_nul): New function.
65 (c_strlen): Add argument and use it. Update recursive calls.
66 Pass DECL argument to string_constant to get info on non
67 terminated strings. Update *NONSTR as needed.
68 (fold_builtin_strlen): Add argument to calls to c_strlen.
69 Warn for unterminated arrays.
70 (warn_string_no_null): Add prototype.
71 * expr.c (string_constant): Update arguments. Update recursive
72 calls appropriately. Detect missing NUL terminator and outermost
73 declaration its missing in.
74 Improve checks for arrays with nonzero lower bound or elements
75 that are not a single byte. Simplify offset computation.
76 Simplify checks for non-NUL terminated strings.
77 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
78 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
80 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
82 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
84 * fold-const.c (c_getstr): Fix function comment. Remove unused third
85 argument. Fix range checks.
86 * fold-const.h (c_getstr): Adjust protoype.
87 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
88 string is constant but contains no NUL byte.
90 * expr.c (string_constant): Adjust function comment.
91 Remove bogus check for zero termination.
93 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
95 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
96 (get_constant_size): Don't make STRING_CSTs larger than they are.
97 (check_string_literal): New check function for STRING_CSTs.
98 (output_constant): Use it.
100 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
103 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
105 2018-09-13 Richard Biener <rguenther@suse.de>
107 PR tree-optimization/87263
108 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
109 (struct unwind_state): Add max_rpo field.
110 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
111 Compute max_rpo, the max RPO number a block can be backwards reached
112 from. Re-write non-iterating mode to a RPO ordered worklist approach,
113 separating it from the iterating mode.
115 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
117 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
118 (rfs_decision): New scheduling decision.
120 2018-09-13 Richard Biener <rguenther@suse.de>
123 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
124 (vn_nary_op_insert_pieces_predicated): Do not write useless
125 valid_dominated_by_p entry outside of the allocated storage.
127 2018-09-13 Omar Sandoval <osandov@osandov.com>
128 Tom de Vries <tdevries@suse.de>
131 * dwarf2out.c (is_c): New function.
132 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
134 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
137 * config/aarch64/aarch64.md (*aarch64_bfxil):
139 * config/aarch64/constraints.md (Ulc): Define.
140 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
142 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
145 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
147 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
148 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
149 aarch64_layout_frame call.
150 (aarch64_expand_epilogue): Likewise.
151 (aarch64_initial_elimination_offset): Likewise.
152 (aarch64_get_separate_components): Likewise.
153 (aarch64_use_return_insn_p): Likewise.
154 (aarch64_layout_frame): Remove unneeded check.
156 2018-09-13 Jakub Jelinek <jakub@redhat.com>
158 * configure.ac: Only append
159 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
160 gcc_config_arguments if it was never reconfigured or last reconfigure
161 was with different arguments.
162 * configure: Regenerated.
164 2018-09-13 Jakub Jelinek <jakub@redhat.com>
165 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
168 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
169 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
171 2018-09-13 Jakub Jelinek <jakub@redhat.com>
173 PR tree-optimization/87287
174 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
175 X % C == 0 to X % (unsigned) C == 0 optimization to ...
176 * match.pd (X % C == 0): ... here. New optimization.
178 2018-09-12 Jakub Jelinek <jakub@redhat.com>
181 * expr.h (maybe_optimize_mod_cmp): Declare.
182 * expr.c (mod_inv): New function.
183 (maybe_optimize_mod_cmp): New function.
184 (do_store_flag): Use it.
185 * cfgexpand.c (expand_gimple_cond): Likewise.
187 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
188 Julian Brown <julian@codesourcery.com>
191 * gimplify.c (gimplify_scan_omp_clauses): Set
192 target_firstprivatize_array_bases in OpenACC parallel and kernels
193 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
194 OpenACC data regions.
196 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
198 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
199 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
201 2018-09-12 Richard Biener <rguenther@suse.de>
203 PR tree-optimization/87280
204 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
205 edge but unreachable target.
206 (do_rpo_vn): For conservatively handling a PHI only mark
207 the backedge executable but not the block reachable.
209 2018-09-12 Richard Biener <rguenther@suse.de>
211 PR tree-optimization/87266
212 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
215 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
217 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
219 ("trunc<BFP:mode><DFP_ALL:mode>2")
220 ("trunc<DFP_ALL:mode><BFP:mode>2")
221 ("extend<BFP:mode><DFP_ALL:mode>2")
222 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
223 according to the target operand type.
225 2018-09-12 Jakub Jelinek <jakub@redhat.com>
226 Andreas Krebbel <krebbel@linux.ibm.com>
228 PR tree-optimization/86844
229 * gimple-ssa-store-merging.c
230 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
231 there are any overlapping stores in between them, make sure they are
232 also coalesced or we give up completely.
234 2018-09-12 Jakub Jelinek <jakub@redhat.com>
237 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
238 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
240 2018-09-12 Tom de Vries <tdevries@suse.de>
242 * common.opt (gdescribe-dies): Add option.
243 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
244 attribute for artifical and nameless decls.
245 (dwarf2out_register_external_die): Add description attribute to
246 external reference die.
247 (add_desc_attribute): New functions.
248 (gen_subprogram_die): Add description attribute to
249 DW_TAG_call_site_parameter.
250 * tree-pretty-print.c (print_generic_expr_to_str): New function.
251 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
252 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
254 (@item -gdescribe-dies): Add.
256 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
258 * tree-vrp.c (vrp_shift_undefined_p): Remove.
259 (extract_range_from_binary_expr_1: Call
260 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
261 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
264 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
266 * gimple-ssa-warn-alloca.c
267 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
268 field for ALLOCA_BOUND_*_LARGE.
270 2018-09-11 Nathan Sidwell <nathan@acm.org>
272 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
274 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
276 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
277 for clobbers. Remove obsolete comment.
279 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
281 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
282 mpxchk, mpxld and mpxst types.
283 (define_attr length_immediate): Remove all processing of mpx types.
284 (define_attr prefix_0f): Ditto.
285 (define_attr memory): Ditto.
287 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
289 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
290 (log<mode>2): Change operand 1 predicate to general_operand.
291 Extend operand 1 to XFmode and generate logxf3 insn.
292 (log10<mode>2): Change operand 1 predicate to general_operand.
293 Extend operand 1 to XFmode and generate log10xf3 insn.
294 (log2<mode>2): Change operand 1 predicate to general_operand.
295 Extend operand 1 to XFmode and generate log2xf3 insn.
296 (fyl2xp1_extend<mode>xf3_i387): Remove.
297 (log1p<mode>2): Change operand 1 predicate to general_operand.
298 Extend operand 1 to XFmode and generate log1pxf3 insn.
299 (fxtract_extend<mode>xf3_i387): Remove.
300 (logb<mode>2): Change operand 1 predicate to general_operand.
301 Extend operand 1 to XFmode and generate logbxf3 insn.
302 (ilogb<mode>2): Change operand 1 predicate to general_operand.
303 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
304 (significand<mode>2): Change operand 1 predicate to general_operand.
305 Extend operand 1 to XFmode and generate significandxf3 insn.
307 2018-09-11 Nathan Sidwell <nathan@acm.org>
309 * gcc.c (perror_with_name, pfatal_with_name): Delete.
310 (load_specs): Use fatal_error.
311 (DELETE_IF_ORDINARY, process_command): Use error.
312 (execute, run_attempt): Use fatal_error.
314 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
316 * diagnostic-core.h (sorry_at): New prototype.
317 * diagnostic.c (sorry_at): New function.
319 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
321 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
322 by zero as VR_UNDEFINED.
324 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
326 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
327 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
328 (<sincos>mode2): New expander.
329 (sincos_extend<mode>xf3_i387): Remove insn pattern.
330 (sincos -> sin, cos splitters): Remove splitter patterns.
331 (sincos<mode>3): Change operand 2 predicate to general_operand.
332 Extend operand 2 to XFmode and generate sincosxf3 insn.
333 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
334 Change operand 3 predicate to const1_operand.
335 (fptan_extend<mode>xf4_i387): Remove insn pattern.
336 (tanxf2): Update operands in the call to fptanxf4_i387.
337 (tan<mode>2): Change operand 1 predicate to general_operand.
338 Extend operand 1 to XFmode and generate tanxf3 insn.
339 (atan2xf3): Rename from *fpatanxf3_i387.
340 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
341 (atan2xf3): Remove expander.
342 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
343 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
344 (atan<mode>2): Change operand 1 predicate to general_operand.
345 Extend operand 1 to XFmode and generate atanxf3 insn.
347 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
349 * config/i386/i386.md (x87/SSE constant load splitter): Use
350 memory_operand instead of nonimmediate_operand for input operand
353 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
355 * config/i386/i386.md (float partial SSE register stall splitter): Move
356 splitter near its instruction pattern.
357 (float_extend partial SSE register stall splitter): Ditto.
358 (float_truncate partial SSE register stall splitter): Ditto.
360 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
363 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
364 to speculation_safe_value_not_needed.
367 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
368 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
369 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
370 leaf_function_p, instead use has_hard_reg_initial_val.
372 2018-09-09 Nathan Sidwell <nathan@acm.org>
374 * gcc.h (pfatal_with_name): Don't declare here.
375 * gcc.c (pfatal_with_name): Make static.
377 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
379 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
382 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
384 PR rtl-optimization/85458
385 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
386 priority hook to reduce the priority of EXPR.
388 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
390 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
391 DImode for x87 on 32bit targets. Conditionally disable x87 modes
392 with X87_ENABLE_FLOAT. Remove preparation code.
393 (*float<SWI48:mode><MODEF:mode>2): Rename from
394 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
395 math using "enabled" attribute.
396 (*floatdi<MODEF:mode>2_i387): Rename from
397 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
398 enable for 32bit targets only.
399 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
401 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
402 as operand 1 predicate. Rewrite as define_insn_and_split.
403 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
405 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
407 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
408 to fallthru to FLOAT case.
410 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
413 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
414 around folding of vec_sl to handle out of range shift values.
416 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
418 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
419 Update callers to gen_fix_trunc<mode>_i387_fisttp
420 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
421 nonimmediate_operand.
422 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
423 and corresponding splitters.
424 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
425 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
426 (fix_truncdi_i387_with_temp): Remove insn pattern
427 and corresponding splitters.
428 (fix_trunc<mode>_i387): Change operand 0 predicate to
429 nonimmediate_operand.
430 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
431 and corresponding splitters.
432 (*fistdi2_1): Remove.
434 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
435 (lrintxfdi2): Remove expander. Reimplement as define_insn.
436 (*fist<mode>2_1): Remove.
437 (fist<mode>2): Ditto.
438 (fist<mode>2_with_temp): Remove insn pattern and corresponding
440 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
441 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
442 (fistdi2_<rounding>): Change operand 0 predicate to
443 nonimmediate_operand.
444 (fistdi2_<rounding>_with_temp): Remove insn pattern
445 and corresponding splitters.
446 (fist<mode>2_<rounding>): Change operand 0 predicate to
447 nonimmediate_operand.
448 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
449 and corresponding splitters.
451 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
453 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
455 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
458 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
460 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
461 early gimple folding of vec_splat().
462 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
463 * gimple-fold.h: Add an extern define for tree_vec_extract().
465 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
467 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
468 wrappers around TREE_TYPE comparisons.
470 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
473 * config/s390/predicates.md: Add nonsym_memory_operand.
474 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
475 contains a SYMBOL_REF, load it into an intermediate pseudo.
476 (s390_emit_compare_and_swap): Legitimize operand.
477 * config/s390/s390.md: Use the new nonsym_memory_operand
478 with UNSPECV_CAS patterns.
480 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
483 * config/s390/s390-passes.def: New file.
484 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
486 (make_pass_s390_early_mach): Add declaration.
487 * config/s390/s390.c (make_pass_s390_early_mach):
488 (s390_option_override): Remove dynamic registration.
489 * config/s390/t-s390: Add s390-passes.def.
491 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
493 * config/s390/s390.c (s390_decompose_constant_pool_ref):
494 Remove UNSPEC_LTREL_BASE check.
495 (annotate_constant_pool_refs): Likewise.
496 (find_constant_pool_ref): Likewise.
497 (find_ltrel_base): Removed.
498 (replace_ltrel_base): Removed.
499 (s390_mainpool_finish): Remove replace_ltrel_base call.
500 (s390_chunkify_start): Remove pending LTREL_BASE logic.
501 (s390_chunkify_finish): Remove replace_ltrel_base call.
502 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
504 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
507 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
508 to speculation_safe_value_not_needed.
510 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
511 Bernd Schmidt <bernds_cb1@t-online.de>
513 * config/nvptx/nvptx-opts.h: New file.
514 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
515 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
517 (TARGET_SM35): New macro.
518 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
520 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
523 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
525 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
527 * config/i386/i386.md (truncdfsf2): Remove expander.
528 (truncdfsf2_with_temp): Ditto.
529 (truncxf<mode>2): Ditto.
530 (*truncdfsf_fast_mixed): Remove insn pattern.
531 (*truncdfsf_fast_i387): Ditto.
532 (*truncdfsf_mixed): Ditto.
533 (*truncdfsf_i387): Ditto.
534 (*truncdfsf2_i387_1): Ditto.
535 (*truncxfsf2_mixed): Ditto.
536 (*truncxfdf2_mixed): Ditto.
537 (*truncxf<mode>2_i387_noop): Ditto. Update callers
538 to call gen_truncxf<mode>2 instead.
539 (*truncxf<mode>2_i387): Remove.
540 (reg->reg splitters): Remove splitter pattern.
541 (reg->mem splitters): Ditto.
543 (truncdfsf2): New insn pattern.
544 (truncxf<mode>2): Ditto.
546 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
548 * tree-ssa-math-opts.c (is_mult_by): New function.
549 (is_square_of): Use the above.
550 (optimize_recip_sqrt): New function.
551 (pass_cse_reciprocals::execute): Use the above.
553 2018-09-05 Richard Biener <rguenther@suse.de>
556 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
557 to zero-init the emplaced vec.
559 2018-09-05 Martin Liska <mliska@suse.cz>
561 PR tree-optimization/87205
562 * tree-switch-conversion.c (pass_lower_switch::execute):
563 Group cases for switch statements.
565 2018-09-05 Richard Biener <rguenther@suse.de>
567 PR tree-optimization/87217
568 * tree-ssa-sccvn.c (vuse_valueize): New.
569 (vn_reference_lookup_pieces): Use it.
570 (vn_reference_lookup): Likewise.
572 2018-09-05 Nathan Sidwell <nathan@acm.org>
575 * stor-layout.c (place_field): Scan forwards to check last
576 bitfield when ms_bitfield_placement is in effect.
578 2018-09-05 Richard Biener <rguenther@suse.de>
581 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
584 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
585 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
587 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
588 * config.gcc (extra_objs): Build it.
589 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
591 * config/aarch64/aarch64-passes.def
592 (pass_tag_collision_avoidance): New pass.
593 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
594 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
595 (aarch64_classify_address): Remove static qualifier.
596 (aarch64_address_info, aarch64_address_type): Move to...
597 * config/aarch64/aarch64-protos.h: ... here.
598 (make_pass_tag_collision_avoidance): New function.
599 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
602 2018-09-05 Martin Liska <mliska@suse.cz>
604 * doc/gcov.texi: Update documentation of humar
606 * gcov.c (format_count): Print one decimal place, it provides
607 more fine number of situations like '1G' vs. '1.4G'.
609 2018-09-05 Martin Liska <mliska@suse.cz>
612 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
613 * optc-gen.awk: Allow 'Var' for Deprecated options in order
614 to generate a MASK value.
616 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
619 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
620 if frame pointer isn't used.
621 (compute_frame_pointer_to_fb_displacement): Likewise.
623 2018-09-04 Jakub Jelinek <jakub@redhat.com>
626 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
627 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
628 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
629 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
630 and OPTION_MASK_ISA_XSAVEC_UNSET.
632 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
634 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
635 XOR operations in NAND case.
637 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
639 * wide-int-range.cc (wide_int_range_convert): New.
640 * wide-int-range.h (wide_int_range_convert): New.
641 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
642 code into wide_int_range_convert.
643 (extract_range_into_wide_ints): Do not munge anti range constants
644 into the entire domain. Just return the range back.
646 2018-09-04 Martin Liska <mliska@suse.cz>
648 * genmatch.c (output_line_directive): Add new argument
650 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
652 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
654 * doc/invoke.texi (Option Summary): Add whitespace.
656 * doc/invoke.texi (Option Summary): Add -Waligned-new.
658 2018-09-04 Richard Biener <rguenther@suse.de>
660 PR tree-optimization/87211
661 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
662 backedge value we're supposed to treat as VARYING also number
663 the PHI to VARYING in case it got a different value-number already.
665 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
667 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
668 (extract_range_from_binary_expr_1): Do not call
669 vrp_can_optimize_bit_op.
670 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
672 (wide_int_range_get_mask_and_bounds): New.
673 (wide_int_range_optimize_bit_op): New.
674 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
675 (wide_int_range_bit_and): Same.
676 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
677 (wide_int_range_optimize_bit_op): New.
678 (wide_int_range_get_mask_and_bounds): New.
680 2018-09-04 Richard Biener <rguenther@suse.de>
682 PR tree-optimization/87176
683 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
684 variable. When value-numbering a virtual PHI node make sure
685 to not value-number to the backedge value.
687 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
689 * doc/extend.texi (Long Long, Hex Floats): Document support for
690 long long and hex floats in more recent versions of ISO C++.
692 2018-09-03 Richard Biener <rguenther@suse.de>
694 PR tree-optimization/87177
695 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
698 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
700 * bb-reorder.c (edge_order): Convert to C-qsort-style
701 tri-state comparator.
702 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
704 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
706 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
707 tri-state comparator.
708 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
710 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
712 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
713 (mergesort): ... here as maximum count for using netsort.
714 (gcc_qsort): Set nlim to 3 if stable sort is requested.
715 (gcc_stablesort): New.
716 * system.h (gcc_stablesort): Declare.
718 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
720 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
721 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
722 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
724 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
726 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
727 lxsdx and stxsdx alternatives.
728 (*mov<mode>_hardfloat64): Ditto.
729 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
731 2018-09-03 Richard Biener <rguenther@suse.de>
733 PR tree-optimization/87200
734 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
737 2018-09-03 Martin Liska <mliska@suse.cz>
739 PR tree-optimization/87201
740 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
741 Fix parenthesis in an expression.
743 2018-09-03 Richard Biener <rguenther@suse.de>
745 PR tree-optimization/87197
746 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
747 visited. CSE the VN_INFO hashtable lookup.
749 PR tree-optimization/87169
750 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
751 iterating make sure there's no extra backedges from irreducible
752 regions feeding the header. Mark the destination block
755 2018-09-03 Martin Liska <mliska@suse.cz>
758 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
759 * common/common-targhooks.c (default_get_valid_option_values):
761 * common/common-targhooks.h (default_get_valid_option_values):
763 * common/config/i386/i386-common.c: Move processor_target_table
765 (ix86_get_valid_option_values): New function.
766 (TARGET_GET_VALID_OPTION_VALUES): New macro.
767 * config/i386/i386.c (struct ptt): Move to i386-common.c.
768 (PTA_*): Move all defined masks into i386-common.c.
769 (ix86_function_specific_restore): Use new processor_cost_table.
770 * config/i386/i386.h (struct ptt): Moved from i386.c.
771 (struct pta): Likewise.
772 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
773 * doc/tm.texi.in: Likewise.
774 * opt-suggestions.c (option_proposer::suggest_option):
775 Pass prefix to build_option_suggestions.
776 (option_proposer::get_completions): Likewise.
777 (option_proposer::build_option_suggestions): Use the new target
779 * opts.c (struct option_help_tuple): New struct.
780 (print_filtered_help): Use the new target hook.
782 2018-09-03 Martin Liska <mliska@suse.cz>
785 * predict.c (set_even_probabilities): Add likely_edges
786 argument and handle cases where we have precisely one
788 (combine_predictions_for_bb): Catch also likely_edges.
789 (tree_predict_by_opcode): Handle gswitch statements.
790 * tree-cfg.h (find_case_label_for_value): New declaration.
791 (find_taken_edge_switch_expr): Likewise.
792 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
793 Find pivot in decision tree based on probabily, not by number of
796 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
798 * doc/standards.texi (Standards): Update Objective-C reference.
800 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
802 * doc/install.texi (Prerequisites): Update link for MPC.
804 2018-09-01 Michael Matz <matz@suse.de>
806 PR tree-optimization/87074
807 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
808 PHIs for outer-loop uses.
810 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
812 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
813 * doc/invoke.texi (C Dialect Options): Ditto.
815 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
817 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
819 2018-08-31 Richard Biener <rguenther@suse.de>
821 PR tree-optimization/87168
822 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
823 (rpo_elim::eliminate_avail): When OP was not visited it must
826 2018-08-31 David Malcolm <dmalcolm@redhat.com>
828 * tree-vrp.c (copy_value_range): Convert param "from" from
829 "value_range *" to "const value_range *".
830 (range_is_null): Likewise for param "vr".
831 (range_int_cst_p): Likewise.
832 (range_int_cst_singleton_p): Likewise.
833 (symbolic_range_p): Likewise.
834 (value_ranges_intersect_p): Likewise for both params.
835 (value_range_nonnegative_p): Likewise for param "vr".
836 (value_range_constant_singleton): Likewise.
837 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
838 (extract_range_into_wide_ints): Likewise for param "vr".
839 (extract_range_from_multiplicative_op): Likewise for params "vr0"
841 (vrp_can_optimize_bit_op): Likewise.
842 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
844 (extract_range_from_unary_expr): Likewise.
845 (debug_value_range): Likewise for param "vr".
846 (value_range::dump): Add "const" qualifier.
847 (vrp_prop::check_array_ref): Convert local "vr" from
848 "value_range *" to "const value_range *".
849 (vrp_prop::check_mem_ref): Likewise.
850 (vrp_prop::visit_stmt): Likewise for local "old_vr".
851 (vrp_intersect_ranges_1): Likewise for param "vr_1".
852 (vrp_intersect_ranges): Likewise.
853 (simplify_stmt_for_jump_threading): Likewise for local "vr".
854 (vrp_prop::vrp_finalize): Likewise.
855 * tree-vrp.h (value_range::dump): Add "const" qualifier.
856 (vrp_intersect_ranges): Add "const" qualifier to params as above.
857 (extract_range_from_unary_expr): Likewise.
858 (value_range_constant_singleton): Likewise.
859 (symbolic_range_p): Likewise.
860 (copy_value_range): Likewise.
861 (extract_range_from_binary_expr_1): Likewise.
862 (range_int_cst_p): Likewise.
863 (vrp_set_zero_nonzero_bits): Likewise.
864 (range_int_cst_singleton_p): Likewise.
866 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
868 * config/aarch64/arm_neon.h (vabsd_s64): New.
869 (vnegd_s64): Likewise.
871 2018-08-31 Martin Jambor <mjambor@suse.cz>
873 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
875 2018-08-31 Martin Liska <mliska@suse.cz>
877 * ipa-icf.c (sem_item::add_type): Use
878 sem_item::m_type_hash_cache.
879 * ipa-icf.h: Move the cache from sem_item_optimizer
882 2018-08-31 Nathan Sidwell <nathan@acm.org>
884 * doc/extend.texi (Backwards Compatibility): Remove implicit
885 extern C leeway of () being (...).
887 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
889 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
891 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
895 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
897 2018-08-31 Jakub Jelinek <jakub@redhat.com>
900 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
901 gen_int_mode. Formatting fixes.
903 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
905 * target.def (custom_function_descriptors): Improve documentation.
906 * doc/tm.texi.in (Trampolines): Expand discussion of function
907 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
908 beginning of the section.
909 * doc/tm.texi: Regenerated.
911 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
913 * cfg.h (class auto_edge_flag): Spell out the template-id of the
914 base class in the initializer list. This is a workaround for
915 building with older GCC.
916 (class auto_bb_flag): Likewise.
918 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
920 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
921 (altivec_vcmpequ<VI_char>_p): Remove star.
922 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
924 (expand_strncmp_vec_sequence): New function.
925 (emit_final_str_compare_vec): New function.
926 (expand_strn_compare): Add support for vector strncmp.
927 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
928 length specification to bytes.
929 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
930 (vcmpnezb_p): New pattern.
931 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
932 for option -mstring-compare-inline-limit.
934 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
936 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
937 (PTA_SKYLAKE): Add PTA_AES.
938 (PTA_GOLDMONT): Likewise.
940 2018-08-29 Jan Hubicka <jh@suse.cz>
943 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
944 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
946 2018-08-29 Jan Hubicka <jh@suse.cz>
948 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
950 (hash_tree): Do not visit TYPE_STUB_DECL.
951 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
952 stream TYPE_STUB_DECL.
953 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
954 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
955 after free_lang_data.
956 (type_in_anonymous_namespace_p): Likewise.
958 2018-08-29 Jan Hubicka <jh@suse.cz>
960 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
961 comment that it has to be even number.
962 (class sreal): Change m_sig type to int32_t.
963 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
964 int64_t for temporary calculations.
965 (sreal_verify_basics): Drop one bit from minimum and maximum.
967 2018-08-30 Richard Biener <rguenther@suse.de>
969 PR tree-optimization/87147
970 * tree-ssa-sccvn.c (SSA_VISITED): New function.
971 (visit_phi): When the degenerate result is from the backedge and
972 we didn't visit its definition yet drop to VARYING.
973 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
975 2018-08-29 Jan Hubicka <jh@suse.cz>
977 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
979 (hash_tree): Likewise.
981 2018-08-29 Jan Hubicka <jh@suse.cz>
983 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
984 and TYPE_NEXT_REF_TO.
986 2018-08-29 Jan Hubicka <jh@suse.cz>
988 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
989 comment that it has to be even number.
990 (class sreal): Change m_sig type to int32_t.
991 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
992 int64_t for temporary calculations.
993 (sreal_verify_basics): Drop one bit from minimum and maximum.
995 2018-08-30 Tamar Christina <tamar.christina@arm.com>
997 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
999 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1002 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1003 if to_add is negative.
1005 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1008 * builtins.c (c_strlen): Improve range checks.
1010 2018-08-29 Martin Sebor <msebor@redhat.com>
1011 Jeff Law <law@redhat.com>
1013 PR tree-optimization/86714
1014 PR tree-optimization/86711
1015 * builtins.c (c_strlen): Add arguments to call to string_constant.
1016 * expr.c (string_constant): Add argument. Detect missing nul
1017 terminator and outermost declaration it's missing in.
1018 * expr.h (string_constant): Add argument.
1019 * fold-const.c (read_from_constant_string): Add arguments to call to
1021 (c_getstr): Likewise.
1022 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1024 * tree-ssa-strlen.c (get_stridx): Likewise.
1026 2018-08-29 Jan Hubicka <jh@suse.cz>
1028 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1029 Do not stream DECL_VINDEX.
1030 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1031 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1032 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1034 2018-08-29 Richard Biener <rguenther@suse.de>
1036 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1037 virtual operands that are not default defs to honor region
1039 (rpo_vn_valueize): Remove ineffective code here.
1041 2018-08-29 Richard Biener <rguenther@suse.de>
1043 PR tree-optimization/87132
1044 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1045 when skipping defs reachable over backedges.
1047 2018-08-29 Richard Biener <rguenther@suse.de>
1049 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1050 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1051 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1052 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1053 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1054 (vn_reference_lookup_3): Remove use of const_parms.
1055 (free_rpo_vn): Do not free const_parms.
1056 (do_rpo_vn): Do not call init_const_parms.
1057 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1058 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1059 (call_may_clobber_ref_p_1): Likewise.
1061 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1064 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1066 (-ftree-scev-cprop): Document.
1068 2018-08-29 Jan Hubicka <jh@suse.cz>
1070 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1072 (sreal constructor): Update.
1073 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1074 sreal:operator/): Update.
1076 2018-08-29 Martin Liska <mliska@suse.cz>
1078 * tree-switch-conversion.c (switch_conversion::expand):
1079 Strenghten assumption about gswitch statements.
1081 2018-08-29 Richard Biener <rguenther@suse.de>
1083 PR tree-optimization/87117
1084 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1085 re-value-number released SSA VDEFs.
1087 2018-08-29 Richard Biener <rguenther@suse.de>
1089 PR tree-optimization/87126
1090 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1092 2018-08-28 Jim Wilson <jimw@sifive.com>
1094 * config/riscv/pic.md: Rewrite.
1095 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1097 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1098 (SOFTF, default_load, softload, softstore): New.
1100 2018-08-28 Jeff Law <law@redhat.com>
1102 * fold-const.c (fold_binary_loc): Remove recently added assert.
1104 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1106 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1107 to OP parmeter of generated function.
1109 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1111 PR tree-optimization/87009
1112 * match.pd: Add boolean optimizations.
1114 2018-08-28 Martin Sebor <msebor@redhat.com>
1117 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1118 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1119 (pass_walloca::gate): Use it.
1120 (alloca_call_type): Same.
1121 (pass_walloca::execute): Same.
1122 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1124 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1126 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1127 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1130 2018-08-28 Richard Biener <rguenther@suse.de>
1132 PR tree-optimization/87124
1133 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1134 constants before looking up avail.
1136 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1139 * calls.c (maybe_warn_nonstring_arg): Punt early if
1140 warn_stringop_overflow is zero. Don't call get_range_strlen
1141 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1142 Swap comparison operands to have constants on rhs. Only use
1143 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1144 increment lenrng[0].
1146 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1148 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1149 use of tree_to_shwi. Remove duplicated test for the size being
1150 a whole number of bytes.
1152 2018-08-28 Richard Biener <rguenther@suse.de>
1154 PR tree-optimization/87117
1155 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1156 Handle removed stmt without LHS (GIMPLE_NOP).
1158 2018-08-28 Richard Biener <rguenther@suse.de>
1160 PR tree-optimization/87117
1161 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1162 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1164 2018-08-28 Richard Biener <rguenther@suse.de>
1166 PR tree-optimization/87117
1167 * tree-ssa-pre.c (compute_avail): Do not make expressions
1168 with predicated values available.
1169 (get_expr_value_id): Assert we do not run into predicated value
1172 2018-08-28 Richard Biener <rguenther@suse.de>
1174 PR tree-optimization/87117
1175 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1176 get virtual operands.
1177 (get_expr_operands): Handle STRING_CST like other decls.
1179 2018-08-28 Martin Liska <mliska@suse.cz>
1181 * tree.h: Update documentation of fndecl_built_in_p
1185 2018-08-27 Jeff Law <law@redhat.com>
1186 PR tree-optimization/87110
1187 * tree-ssa-dse.c (compute_trims): Handle non-constant
1190 2018-08-27 Martin Sebor <msebor@redhat.com>
1192 PR tree-optimization/86914
1193 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1195 2018-08-27 Martin Sebor <msebor@redhat.com>
1197 PR tree-optimization/87112
1198 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1199 the type of the bound argument.
1201 2018-08-27 Jeff Law <law@redhat.com>
1203 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1204 type does not have a TYPE_SIZE_UNIT.
1206 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1208 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1209 with include of backend.h.
1211 2018-08-27 Richard Biener <rguenther@suse.de>
1213 PR tree-optimization/86927
1214 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1215 use const cond reduction code.
1217 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1219 PR tree-optimization/85758
1220 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1222 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1225 * diagnostic-show-locus.c (class layout_range): Update for
1226 conversion of show_caret_p to a tri-state.
1227 (layout_range::layout_range): Likewise.
1228 (make_range): Likewise.
1229 (layout::maybe_add_location_range): Likewise.
1230 (layout::should_print_annotation_line_p): Don't show annotation
1231 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1232 (layout::get_state_at_point): Update for conversion of
1233 show_caret_p to a tri-state. Bail out early for
1234 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1235 underlining or source colorization.
1236 (gcc_rich_location::add_location_if_nearby): Update for conversion
1237 of show_caret_p to a tri-state.
1238 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1239 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1241 (selftest::test_one_liner_labels): Likewise.
1242 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1243 conversion of show_caret_p to a tri-state.
1244 * pretty-print.c (text_info::set_location): Likewise.
1245 * pretty-print.h (text_info::set_location): Likewise.
1246 * substring-locations.c (format_warning_n_va): Likewise.
1247 * tree-diagnostic.c (default_tree_printer): Likewise.
1248 * tree-pretty-print.c (newline_and_indent): Likewise.
1250 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1253 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1254 line above for line-insertion fix-it hints.
1255 (selftest::test_fixit_insert_containing_newline): Update the
1256 expected results, and add a test with line-numbering enabled.
1258 2018-08-27 Martin Liska <mliska@suse.cz>
1261 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1262 params with DECL_HAS_VALUE_EXPR_P.
1264 2018-08-27 Martin Liska <mliska@suse.cz>
1266 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1267 selected expansion strategy.
1269 2018-08-27 Martin Liska <mliska@suse.cz>
1271 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1272 * builtins.c (is_builtin_fn): Likewise.
1273 * attribs.c (diag_attr_exclusions): Use new function
1274 fndecl_built_in_p and remove check for FUNCTION_DECL if
1276 (builtin_mathfn_code): Likewise.
1277 (fold_builtin_expect): Likewise.
1278 (fold_call_expr): Likewise.
1279 (fold_builtin_call_array): Likewise.
1280 (fold_call_stmt): Likewise.
1281 (set_builtin_user_assembler_name): Likewise.
1282 (is_simple_builtin): Likewise.
1283 * calls.c (gimple_alloca_call_p): Likewise.
1284 (maybe_warn_nonstring_arg): Likewise.
1285 * cfgexpand.c (expand_call_stmt): Likewise.
1286 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1287 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1288 (cgraph_node::verify_node): Likewise.
1289 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1290 (cgraph_node::create_clone): Likewise.
1291 * config/arm/arm.c (arm_insert_attributes): Likewise.
1292 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1293 * dse.c (scan_insn): Likewise.
1294 * expr.c (expand_expr_real_1): Likewise.
1295 * fold-const.c (operand_equal_p): Likewise.
1296 (fold_binary_loc): Likewise.
1297 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1298 * gimple-low.c (lower_stmt): Likewise.
1299 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1300 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1301 * gimple.c (gimple_build_call_from_tree): Likewise.
1302 (gimple_call_builtin_p): Likewise.
1303 (gimple_call_combined_fn): Likewise.
1304 * gimplify.c (gimplify_call_expr): Likewise.
1305 (gimple_boolify): Likewise.
1306 (gimplify_modify_expr): Likewise.
1307 (gimplify_addr_expr): Likewise.
1308 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1309 * ipa-cp.c (determine_versionability): Likewise.
1310 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1311 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1312 * ipa-split.c (visit_bb): Likewise.
1313 (split_function): Likewise.
1314 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1315 * lto-cgraph.c (input_node): Likewise.
1316 * lto-streamer-out.c (write_symbol): Likewise.
1317 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1318 (lower_omp_1): Likewise.
1319 * predict.c (strip_predict_hints): Likewise.
1320 * print-tree.c (print_node): Likewise.
1321 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1322 * trans-mem.c (is_tm_irrevocable): Likewise.
1323 (is_tm_load): Likewise.
1324 (is_tm_simple_load): Likewise.
1325 (is_tm_store): Likewise.
1326 (is_tm_simple_store): Likewise.
1327 (is_tm_abort): Likewise.
1328 (tm_region_init_1): Likewise.
1329 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1330 * tree-cfg.c (verify_gimple_call): Likewise.
1331 (move_stmt_r): Likewise.
1332 (stmt_can_terminate_bb_p): Likewise.
1333 * tree-eh.c (lower_eh_constructs_2): Likewise.
1334 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1335 * tree-inline.c (remap_gimple_stmt): Likewise.
1336 (copy_bb): Likewise.
1337 (estimate_num_insns): Likewise.
1338 (fold_marked_statements): Likewise.
1339 * tree-sra.c (scan_function): Likewise.
1340 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1341 (optimize_stack_restore): Likewise.
1342 (pass_fold_builtins::execute): Likewise.
1343 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1344 (mark_all_reaching_defs_necessary_1): Likewise.
1345 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1346 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1347 (pass_forwprop::execute): Likewise.
1348 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1349 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1350 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1351 * tree-ssa-strlen.c (get_string_length): Likewise.
1352 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1353 (find_func_aliases_for_call): Likewise.
1354 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1355 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1356 * tree-tailcall.c (find_tail_calls): Likewise.
1357 * tree.c (need_assembler_name_p): Likewise.
1358 (free_lang_data_in_decl): Likewise.
1359 (get_call_combined_fn): Likewise.
1360 * ubsan.c (is_ubsan_builtin_p): Likewise.
1361 * varasm.c (incorporeal_function_p): Likewise.
1362 * tree.h (DECL_BUILT_IN): Remove and replace with
1364 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1365 (fndecl_built_in_p): New.
1367 2018-08-27 Martin Liska <mliska@suse.cz>
1369 PR tree-optimization/86847
1370 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1371 Dump also subtree probability.
1372 (switch_decision_tree::do_jump_if_equal): New function.
1373 (switch_decision_tree::emit_case_nodes): Handle special
1374 situations in balanced tree that can be emitted much simpler.
1375 Fix calculation of probabilities that happen in tree expansion.
1376 * tree-switch-conversion.h (struct cluster): Add
1378 (struct simple_cluster): Likewise.
1379 (struct case_tree_node): Add new function has_child.
1380 (do_jump_if_equal): New.
1382 2018-08-27 Martin Liska <mliska@suse.cz>
1384 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1385 Add new argument to bit_test_cluster constructor.
1386 (bit_test_cluster::emit): Set bits really number of values
1388 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1389 probability argument.
1390 * tree-switch-conversion.h (struct bit_test_cluster):
1391 Add m_handles_entire_switch.
1393 2018-08-27 Martin Liska <mliska@suse.cz>
1395 PR tree-optimization/86702
1396 * tree-switch-conversion.c (jump_table_cluster::emit):
1397 Make probabilities even for values in jump table
1398 according to number of cases handled.
1399 (switch_decision_tree::compute_cases_per_edge): Pass
1400 argument to reset_out_edges_aux function.
1401 (switch_decision_tree::analyze_switch_statement): Likewise.
1402 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1405 2018-08-27 Martin Liska <mliska@suse.cz>
1407 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1408 cfun argument explicitly.
1409 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1410 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1411 function gimple_switch_default_bb.
1412 (convert_switch_statements):
1414 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1415 * stmt.c (label_to_block_fn): Use label_to_block and pass
1416 cfun argument explicitly and use gimple_switch_label_bb.
1417 (expand_case): Likewise.
1418 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1419 cfun argument explicitly. Likewise.
1420 (make_edges_bb): Likewise.
1421 (make_cond_expr_edges): Likewise.
1422 (get_cases_for_edge): Likewise.
1423 (make_gimple_switch_edges): Likewise.
1424 (label_to_block_fn): Likewise.
1425 (label_to_block): Likewise.
1426 (make_goto_expr_edges): Likewise.
1427 (make_gimple_asm_edges): Likewise.
1428 (main_block_label): Likewise.
1429 (group_case_labels_stmt): Likewise.
1430 (find_taken_edge_computed_goto): Likewise.
1431 (find_taken_edge_switch_expr): Likewise.
1432 (gimple_verify_flow_info): Likewise.
1433 (gimple_redirect_edge_and_branch): Likewise.
1434 (gimple_switch_label_bb): New function.
1435 (gimple_switch_default_bb): Likewise.
1436 (gimple_switch_edge): Likewise.
1437 (gimple_switch_default_edge): Likewise.
1438 * tree-cfg.h (label_to_block_fn): Remove and replace ...
1439 (label_to_block): ... with this.
1440 (gimple_switch_label_bb): New.
1441 (gimple_switch_default_bb): Likewise.
1442 (gimple_switch_edge): Likewise.
1443 (gimple_switch_default_edge): Likewise.
1444 * tree-cfgcleanup.c (convert_single_case_switch): Use
1445 new gimple functions and pass new argument to label_to_block.
1446 (cleanup_control_flow_bb):
1447 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
1448 cfun argument explicitly.
1449 (make_eh_edges): Likewise.
1450 (redirect_eh_dispatch_edge): Likewise.
1451 (lower_resx): Likewise.
1452 (lower_eh_dispatch): Likewise.
1453 (maybe_remove_unreachable_handlers): Likewise.
1454 (unsplit_eh): Likewise.
1455 (cleanup_empty_eh): Likewise.
1456 (verify_eh_edges): Likewise.
1457 (verify_eh_dispatch_edge): Likewise.
1458 * tree-ssa-dom.c (record_edge_info): Likewise.
1459 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
1460 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
1461 (thread_through_normal_block): Likewise.
1462 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
1463 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
1464 * tree-switch-conversion.c (switch_conversion::collect): Use new
1466 (switch_conversion::check_final_bb): Likewise.
1467 (switch_conversion::gather_default_values): Pass new argument
1469 (switch_conversion::build_constructors): Likewise.
1470 (switch_decision_tree::compute_cases_per_edge): Use new
1471 gimple_switch_edge function.
1472 (switch_decision_tree::analyze_switch_statement): Pass new argument
1474 (switch_decision_tree::try_switch_expansion): Use
1475 gimple_switch_default_edge.
1476 * tree-vrp.c (find_switch_asserts): Pass new argument
1478 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
1479 (vr_values::simplify_switch_using_ranges): Likewise.
1481 2018-08-27 Richard Biener <rguenther@suse.de>
1483 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
1484 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
1486 * tree-ssa-sccvn.h (struct vn_pval): New structure.
1487 (struct vn_nary_op_s): Add unwind_to member. Add
1488 predicated_values flag and put result into a union together
1489 with a linked list of vn_pval.
1490 (struct vn_ssa_aux): Add name member to make maintaining
1491 a map of SSA name to vn_ssa_aux possible. Remove no longer
1492 needed info, dfsnum, low, visited, on_sccstack, use_processed
1493 and range_info_anti_range_p members.
1494 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
1495 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
1497 (vn_valueize): New global.
1498 (vn_context_bb): Likewise.
1499 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
1500 VN_INFO_PTR_INFO): Remove.
1501 * tree-ssa-sccvn.c: ... (rewrite)
1502 (pass_fre::execute): For -O2+ initialize loops and run
1503 RPO VN in optimistic mode (iterating). For -O1 and -Og
1504 run RPO VN in non-optimistic mode.
1505 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
1506 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
1507 * doc/invoke.texi (sccvn-max-scc-size): Remove.
1508 (rpo-vn-max-loop-depth): Document.
1509 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
1510 when valuezing the VUSE signals we walked out of the region.
1511 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
1512 (phi_translate): Set VN context block to use for availability
1514 (compute_avail): Likewise.
1515 (pre_valueize): New function.
1516 (pass_pre::execute): Adjust to the RPO VN API.
1518 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
1519 (propagate_constants_for_unrolling): Remove.
1520 (tree_unroll_loops_completely): Perform value-numbering
1521 on the unrolled bodies loop parent.
1523 2018-08-27 Richard Biener <rguenther@suse.de>
1525 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
1526 for partial antic compute.
1528 2018-08-27 Jakub Jelinek <jakub@redhat.com>
1530 PR rtl-optimization/87065
1531 * combine.c (simplify_if_then_else): Formatting fix.
1532 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
1534 (known_cond): Don't return const_true_rtx for vector modes. Use
1535 CONST0_RTX instead of const0_rtx. Formatting fixes.
1537 2018-08-27 Martin Liska <mliska@suse.cz>
1539 PR gcov-profile/87069
1540 * gcov.c (process_file): Record files already processed
1541 and warn about a file being processed multiple times.
1543 2018-08-27 Martin Liska <mliska@suse.cz>
1546 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1547 Set default values for x_aarch64_*_string strings.
1548 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
1549 prefix. For -mabi do not print '=ABI' in help and use
1550 <option_value> format for -msve-vector-bits and -moverride
1553 2018-08-26 Jeff Law <law@redhat.com>
1555 * config/mips/frame-header-opt.c: Include "backend.h" rather than
1558 2018-08-26 Marek Polacek <polacek@redhat.com>
1560 PR c++/87029, Implement -Wredundant-move.
1561 * doc/invoke.texi: Document -Wredundant-move.
1563 2018-08-25 Martin Sebor <msebor@redhat.com>
1565 PR tree-optimization/87059
1566 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
1567 to the same type as the other.
1568 * fold-const.c (fold_binary_loc): Assert expectation.
1570 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
1572 * config/darwin.c (machopic_legitimize_pic_address): Clean up
1573 extraneous parentheses, dead code section and formatting.
1575 2018-08-24 David Malcolm <dmalcolm@redhat.com>
1578 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
1579 wide enough for jumps in the line-numbering to be visible.
1580 (layout::print_gap_in_line_numbering): New member function.
1581 (layout::calculate_line_spans): When using line numbering, merge
1582 line spans that are only 1 line apart.
1583 (diagnostic_show_locus): When printing line numbers, show gaps in
1584 line numbering directly, rather than printing headers.
1585 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
1586 line-numbering with multiple line spans.
1587 (selftest::test_fixit_insert_containing_newline_2): Add test of
1588 line-numbering, in which the spans are close enough to be merged.
1590 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1592 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
1593 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
1594 * tree-vrp.c (range_is_nonnull): Remove.
1595 (range_includes_zero_p): Accept value_range instead of min/max.
1596 (extract_range_from_binary_expr_1): Do not early bail on
1598 Use range_includes_zero_p instead of range_is_nonnull.
1599 (extract_range_from_unary_expr): Use range_includes_zero_p instead
1600 of range_is_nonnull.
1601 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
1602 special case VR_ANTI_RANGE.
1603 (vrp_finalize): Same.
1604 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1606 (range_is_nonnull): Remove.
1607 * vr-values.c (vrp_stmt_computes_nonzero): Use
1608 range_includes_zero_p instead of range_is_nonnull.
1609 (extract_range_basic): Pass value_range to range_includes_zero_p
1610 instead of range_is_nonnull.
1612 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
1614 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1615 * emit-rtl.h (rtl_data): Remove return_bnd.
1616 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1617 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1618 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1619 (POINTER_BOUNDS_MODE): Remove definition.
1620 (make_pointer_bounds_mode): Remove.
1621 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1622 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1623 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1624 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1625 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1626 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1627 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1629 * config/i386/i386-modes.def (BND32, BND64): Remove.
1630 * config/i386/i386.c (dbx_register_map): Remove bound registers.
1631 (dbx64_register_map): Ditto.
1632 (svr4_dbx_register_map): Ditto.
1633 (indirect_thunk_bnd_needed): Remove.
1634 (indirect_thunks_bnd_used): Ditto.
1635 (indirect_return_bnd_needed): Ditto.
1636 (indirect_return_via_cx_bnd): Ditto.
1637 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1638 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1639 (output_indirect_thunk): Ditto. Remove need_prefix argument.
1640 (output_indirect_thunk_function): Remove handling of
1641 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1642 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1643 (ix86_save_reg): Remove handling of crtl->return_bnd.
1644 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1645 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1646 and UNSPEC_BNDLX_ADDR.
1647 (ix86_output_indirect_branch_via_reg): Remove handling of
1648 indirect_thunk_prefix_bnd.
1649 (ix86_output_indirect_branch_via_push): Ditto.
1650 (ix86_output_function_return): Ditto.
1651 (ix86_output_indirect_function_return): Ditto.
1652 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1653 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1654 (CALL_USED_REGISTERS): Ditto.
1655 (REG_ALLOC_ORDER): Update for removal of bound registers.
1656 (HI_REGISTER_NAMES): Ditto.
1657 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1658 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1659 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1660 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1661 (FIRST_PSEUDO_REG): Update.
1662 (BND): Remove mode iterator.
1663 * config/i386/predicates.md (bnd_mem_operator): Remove.
1665 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1667 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1670 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1672 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1673 the case in which the permute needs only a single element and
1674 repeats for every vector of the result. Extend that case to
1675 handle variable-length vectors.
1676 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1678 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
1681 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1682 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1684 2018-08-24 Richard Biener <rguenther@suse.de>
1686 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1687 bb_flags_allocated members.
1688 (auto_flag): New RAII class for allocating flags.
1689 (auto_edge_flag): New RAII class for allocating edge flags.
1690 (auto_bb_flag): New RAII class for allocating bb flags.
1691 * cfgloop.c (verify_loop_structure): Allocate temporary edge
1693 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1694 in favor of temporarily allocated BB flag.
1695 * hsa-brig.c: Re-order includes.
1696 * hsa-dump.c: Likewise.
1697 * hsa-regalloc.c: Likewise.
1698 * print-rtl.c: Likewise.
1699 * profile-count.c: Likewise.
1701 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
1704 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1707 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1710 * wide-int-range.cc (wide_int_range_div): Do not ignore result
1711 from wide_int_range_multiplicative_op.
1713 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1715 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1718 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1720 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1723 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
1725 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1726 full GENERIC expression used for replacement.
1728 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
1730 * tree-vrp.c (abs_extent_range): Remove.
1731 (extract_range_into_wide_ints): Pass wide ints by reference.
1732 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
1733 Pass wide ints by reference in all calls to
1734 extract_range_into_wide_ints.
1735 * wide-int-range.cc (wide_int_range_div): New.
1736 * wide-int-range.h (wide_int_range_div): New.
1737 (wide_int_range_includes_zero_p): New.
1738 (wide_int_range_zero_p): New.
1740 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
1742 * config/aarch64/aarch64.md (arches): New enum.
1743 (arch): New enum attr.
1744 (arch_enabled): New attr.
1745 (enabled): Now uses arch_enabled only.
1746 (simd, sve, fp16): Removed attribute.
1747 (fp): Attr now defined in terms of 'arch'.
1748 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
1749 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
1750 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
1751 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
1752 attributes into 'arch'.
1753 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
1754 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
1755 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
1756 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
1757 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
1758 'simd' attribute into 'arch'.
1759 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
1760 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
1761 Convert use of 'fp' attribute to 'arch'.
1762 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
1763 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
1765 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
1766 (different modes) Merge 'fp' and 'simd' into 'arch'.
1767 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
1770 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
1772 PR rtl-optimization/87026
1773 * expmed.c (canonicalize_comparison): If we can no longer create
1774 pseudoregisters, don't.
1776 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
1779 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
1781 * config/arm/arm.c (speculation_barrier_libfunc): New static
1783 (arm_init_libfuncs): Initialize it.
1784 (arm_emit_speculation_barrier): New function.
1785 * config/arm/arm.md (speculation_barrier): Call
1786 arm_emit_speculation_barrier for architectures that do not have
1788 (speculation_barrier_insn): Only match on Armv7 or later.
1790 2018-08-23 Richard Biener <rguenther@suse.de>
1793 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
1796 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1798 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
1799 of single-vector TBLs.
1800 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
1803 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1806 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
1807 aarch64_evpc_tbl guard.
1809 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
1811 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
1814 2018-08-22 Martin Sebor <msebor@redhat.com>
1817 * tree-pretty-print.c (pretty_print_string): Add argument.
1818 (dump_generic_node): Call to pretty_print_string with string size.
1820 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
1822 PR rtl-optimization/86771
1823 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
1824 of two SETs into those two SETs, one to be placed at i2, if that SETs
1825 destination is modified between i2 and i3.
1827 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1829 PR tree-optimization/86725
1830 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
1832 (vect_analyze_scalar_cycles_1): Check it.
1834 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1836 PR tree-optimization/86725
1837 * tree-vect-loop.c (vect_is_simple_reduction): When treating
1838 an outer loop phi as a double reduction, make sure that the
1839 single user of the phi result is an inner loop phi.
1841 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1843 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
1844 grouped stores with gaps to a strided group.
1846 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1848 * tree-vect-stmts.c (get_group_load_store_type)
1849 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
1850 first statement in a group.
1852 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1854 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
1855 the sequence used in gcc/gcc.c.
1857 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1860 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
1863 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1865 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
1866 Darwin10-specific unwinder-shim.
1867 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
1868 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
1869 New to cater for Darwin10 Rosetta.
1871 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1873 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
1874 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
1876 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1881 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
1882 (dwarf2out_switch_text_section): Generate a local label for the
1883 second function sub-section and apply it as the second FDE start
1885 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
1886 second sub-section start.
1888 2018-08-22 Richard Biener <rguenther@suse.de>
1890 PR tree-optimization/86988
1891 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
1893 2018-08-22 Richard Biener <rguenther@suse.de>
1895 PR tree-optimization/86945
1896 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
1898 2018-08-22 Alexandre Oliva <oliva@adacore.com>
1900 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
1901 a comment about how uses of r2 for .sdata2 come about.
1903 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
1905 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
1907 2018-08-21 Marek Polacek <polacek@redhat.com>
1909 PR c++/86981, Implement -Wpessimizing-move.
1910 * doc/invoke.texi: Document -Wpessimizing-move.
1912 2018-08-21 Jan Hubicka <jh@suse.cz>
1914 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
1915 * tree.h (is_redundant_typedef): Remove.
1916 * dwarf2out.c (is_redundant_typedef): Turn into static function.
1918 2018-08-21 Jan Hubicka <jh@suse.cz>
1920 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
1923 2018-08-21 Tamar Christina <tamar.christina@arm.com>
1925 * expmed.c (extract_low_bits): Reject invalid subregs early.
1927 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1930 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
1933 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1935 * config/vxworks.h: Guard vxworks_asm_out_constructor and
1936 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
1937 * config/vxworks.c: Likewise.
1939 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1941 * config/vxworks.c: Set targetm.have_ctors_dtors
1942 if HAVE_INITFINI_ARRAY_SUPPORT.
1943 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
1944 if HAVE_INITFINI_ARRAY_SUPPORT.
1946 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1948 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
1949 default search path for VxWorks < 7.
1951 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1953 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
1954 (get_format_string): Refer to c_getstr.
1956 2018-08-21 Tom de Vries <tdevries@suse.de>
1958 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1959 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1960 (debuginfo_early_stop): Declare.
1961 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1962 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1963 (debuginfo_early_stop): New function.
1964 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
1965 and debuginfo_early_stop.
1966 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
1967 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
1968 (general_init): Call debuginfo_early_init.
1969 (finalize): Call debuginfo_fini.
1970 (do_compile): Call debuginfo_init.
1971 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
1973 (@item -fdump-debug, @item -fdump-earlydebug): Add.
1975 2018-08-21 Tom de Vries <tdevries@suse.de>
1977 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
1978 flag_dump_noaddr and flag_dump_unnumbered.
1980 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
1982 * wide-int-range.cc (wide_int_range_abs): New.
1983 (wide_int_range_order_set): Rename from wide_int_range_min_max.
1984 * wide-int-range.h (wide_int_range_abs): New.
1985 (wide_int_range_min_max): New.
1986 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
1987 case to call wide_int_range_abs.
1988 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
1989 (extract_range_from_abs_expr): Delete.
1991 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
1994 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
1995 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
1998 2018-08-20 Nathan Sidwell <nathan@acm.org>
1999 Jeff Law <law@redhat.com>
2001 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2002 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2004 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2007 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2008 (decl_attributes): Likewise.
2009 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2011 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2012 * diagnostic-core.h (class auto_diagnostic_group): New class.
2013 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2014 (diagnostic_report_diagnostic): Handle the first diagnostics within
2016 (emit_diagnostic): Add auto_diagnostic_group instance.
2018 (inform_n): Likewise.
2019 (warning): Likewise.
2020 (warning_at): Likewise.
2021 (warning_n): Likewise.
2022 (pedwarn): Likewise.
2023 (permerror): Likewise.
2025 (error_n): Likewise.
2026 (error_at): Likewise.
2028 (fatal_error): Likewise.
2029 (internal_error): Likewise.
2030 (internal_error_no_backtrace): Likewise.
2031 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2032 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2033 * diagnostic.h (struct diagnostic_context): Add fields
2034 "diagnostic_group_nesting_depth",
2035 "diagnostic_group_emission_count", "begin_group_cb",
2037 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2038 Add auto_diagnostic_group instance(s).
2039 (find_explicit_erroneous_behavior): Likewise.
2040 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2041 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2042 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2043 (gimplify_va_arg_expr): Likewise.
2044 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2045 (HSA_SORRY_AT): Likewise.
2046 * ipa-devirt.c (compare_virtual_tables): Likewise.
2047 (warn_odr): Likewise.
2048 * multiple_target.c (expand_target_clones): Likewise.
2049 * opts-common.c (cmdline_handle_error): Likewise.
2050 * reginfo.c (globalize_reg): Likewise.
2051 * substring-locations.c (format_warning_n_va): Likewise.
2052 * tree-inline.c (expand_call_inline): Likewise.
2053 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2054 * tree-ssa-loop-niter.c
2055 (do_warn_aggressive_loop_optimizations): Likewise.
2056 * tree-ssa-uninit.c (warn_uninit): Likewise.
2057 * tree.c (warn_deprecated_use): Likewise.
2059 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2062 * config/i386/i386.md (eh_return): Always update EH return
2063 address in word_mode.
2065 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2067 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2068 TARGET_SPLIT_COMPLEX_ARG is defined.
2070 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2072 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2074 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2077 * expr.c (expand_assignment): Assert that bitpos is positive.
2078 (store_field): Likewise
2079 (expand_expr_real_1): Make sure that bitpos is positive.
2080 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2083 2018-08-20 Nathan Sidwell <nathan@acm.org>
2085 * Makefile.in (CPP_ID_DATA_H): Delete.
2086 (CPP_INTERNAL_H): Don't add it.
2087 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2088 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2090 2018-08-20 Richard Biener <rguenther@suse.de>
2092 PR tree-optimization/78655
2093 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2094 pointer + offset nonnull if either operand is nonnull work.
2096 2018-08-20 Tom de Vries <tdevries@suse.de>
2098 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2099 unless the referenced die describes the added property using
2100 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2101 Otherwise, add a DW_AT_location to the referenced die.
2103 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2106 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2107 register_operand when calling ix86_set_reg_reg_cost.
2108 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2109 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2110 predicate and to 1 otherwise.
2112 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2114 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2115 emit a diagnostic that it is not supported and reset the option.
2116 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2117 supported and consume the option. (ASM_FINAL_SPEC): New.
2119 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2121 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2124 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2126 C-SKY port: Documentation
2128 * doc/extend.texi (C-SKY Function Attributes): New section.
2129 * doc/invoke.texi (Option Summary): Add C-SKY options.
2130 (C-SKY Options): New section.
2131 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2133 2018-08-17 Jojo <jijie_rong@c-sky.com>
2134 Huibin Wang <huibin_wang@c-sky.com>
2135 Sandra Loosemore <sandra@codesourcery.com>
2136 Chung-Lin Tang <cltang@codesourcery.com>
2138 C-SKY port: Backend implementation
2140 * config/csky/*: New.
2141 * common/config/csky/*: New.
2143 2018-08-17 Jojo <jijie_rong@c-sky.com>
2144 Huibin Wang <huibin_wang@c-sky.com>
2145 Sandra Loosemore <sandra@codesourcery.com>
2146 Chung-Lin Tang <cltang@codesourcery.com>
2147 Andrew Jenner <andrew@codesourcery.com>
2149 C-SKY port: Configury
2151 * config.gcc (csky-*-*): New.
2152 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2153 * configure: Regenerated.
2155 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2157 * dump-context.h: Include "dumpfile.h".
2158 (dump_context::dump_printf_va): Convert final param from va_list
2159 to va_list *. Convert from ATTRIBUTE_PRINTF to
2160 ATTRIBUTE_GCC_DUMP_PRINTF.
2161 (dump_context::dump_printf_loc_va): Likewise.
2162 * dumpfile.c: Include "stringpool.h".
2163 (make_item_for_dump_printf_va): Delete.
2164 (make_item_for_dump_printf): Delete.
2165 (class dump_pretty_printer): New class.
2166 (dump_pretty_printer::dump_pretty_printer): New ctor.
2167 (dump_pretty_printer::emit_items): New member function.
2168 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2170 (dump_pretty_printer::emit_item): New member function.
2171 (dump_pretty_printer::stash_item): New member function.
2172 (dump_pretty_printer::format_decoder_cb): New member function.
2173 (dump_pretty_printer::decode_format): New member function.
2174 (dump_context::dump_printf_va): Reimplement in terms of
2175 dump_pretty_printer.
2176 (dump_context::dump_printf_loc_va): Convert final param from va_list
2178 (dump_context::begin_scope): Reimplement call to
2179 make_item_for_dump_printf.
2180 (dump_printf): Update for change to dump_printf_va.
2181 (dump_printf_loc): Likewise.
2182 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2183 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2184 with %T, %E, and %G.
2185 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2186 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2187 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2188 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2189 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2190 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2191 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2192 within a dump_printf_loc call to "%wu".
2193 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2194 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2195 missing space after "=".
2196 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2197 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2198 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2199 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2200 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2201 duplicate "vectorized" from message.
2203 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2205 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2206 polyNxK_t element's TYPE_STRING_FLAG.
2208 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2210 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2211 (they were unnamed before). Fix comments.
2213 2018-08-17 Nathan Sidwell <nathan@acm.org>
2215 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2217 2018-08-17 Richard Biener <rguenther@suse.de>
2219 PR tree-optimization/86841
2220 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2222 2018-08-17 Martin Liska <mliska@suse.cz>
2224 * common.opt: Remove Warn, Init and Report for options with
2225 Ignore/Deprecated flag. Warning is done automatically for
2227 * config/i386/i386.opt: Likewise.
2228 * config/ia64/ia64.opt: Likewise.
2229 * config/rs6000/rs6000.opt: Likewise.
2230 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2231 Remove usage of flag_check_pointer_bounds.
2232 * lto-wrapper.c (merge_and_complain): Do not handle
2233 OPT_fcheck_pointer_bounds.
2234 (append_compiler_options): Likewise.
2235 * opt-functions.awk: Do not handle Deprecated.
2236 * optc-gen.awk: Check that Var, Report and Init are not
2237 used for an option with Ignore/Deprecated flag.
2238 * opts-common.c (decode_cmdline_option): Do not report
2240 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2242 * opts.h (struct cl_option): Remove cl_deprecated flag.
2243 (CL_ERR_DEPRECATED): Remove error enum value.
2245 2018-08-17 Richard Biener <rguenther@suse.de>
2248 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2249 across a va-arg-pack using call adjust its return value accordingly.
2251 2018-08-16 Martin Sebor <msebor@redhat.com>
2253 PR tree-optimization/86853
2254 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2255 (struct fmtresult): Add member and initialize it in ctors.
2256 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2257 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2258 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2259 (parse_directive): Handle %C same as %c.
2260 (sprintf_dom_walker::compute_format_length): Adjust.
2261 (is_call_safe): Adjust.
2263 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2265 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2266 for determining how to count the elements.
2267 * builtins.h (c_strlen): Adjust prototype.
2268 * expr.c (string_constant): Add new parameter mem_size.
2269 Set *mem_size appropriately.
2270 * expr.h (string_constant): Adjust protoype.
2271 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2272 * gimple-fold.h (get_range_strlen): Adjust prototype.
2273 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2274 (format_string): Call get_string_length with eltsize.
2276 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2278 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2279 to emit the span, rather than setting it as the prefix.
2281 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2283 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2284 "margin_char" parameter, defaulting to space. Use it in place
2285 of pp_space for the initial part of the margin.
2286 (layout::print_leading_fixits): Use '+' when filling the margin
2287 of line-insertion fix-it hints.
2289 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2291 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2294 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2296 * config/rs6000/altivec.md: Don't set length attribute to the default
2298 * config/rs6000/darwin.md: Ditto.
2299 * config/rs6000/dfp.md: Ditto.
2300 * config/rs6000/htm.md: Ditto.
2301 * config/rs6000/rs6000.md: Ditto.
2302 * config/rs6000/sync.md: Ditto.
2303 * config/rs6000/vsx.md: Ditto.
2305 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2307 * config/rs6000/altivec.md: Don't set length attribute to the default
2308 value, for branch instructions.
2309 * config/rs6000/darwin.md: Ditto.
2310 * config/rs6000/rs6000.md: Ditto.
2312 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2314 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2315 (unnamed conditional branch define_insn): Set length to 4 or 8
2316 depending on offset.
2317 (<bd>_<mode>): Similar, for alternative 0.
2318 (<bd>tf_<mode>): Ditto.
2320 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2322 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2324 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2326 * doc/rtl.texi: Replace old RTX class names with new names.
2329 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2331 * expmed.h (canonicalize_comparison): New declaration.
2332 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2333 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2334 * optabs.c (prepare_cmp_insn): Likewise.
2335 * rtl.h (unsigned_condition_p): New function which checks if a
2336 comparison operator is unsigned.
2338 2018-08-16 Nathan Sidwell <nathan@acm.org>
2340 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2341 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2343 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2346 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2347 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2348 (mov<mov>): ..this and enable unconditionally.
2350 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2352 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2354 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2356 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2357 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2358 "Common" with "Target".
2360 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2362 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2363 * doc/invoke.texi (mmitigate-rop): Remove.
2364 * config/i386/i386.c: Do not include "regrename.h".
2365 (ix86_rop_should_change_byte_p, reg_encoded_number)
2366 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2368 (ix86_reorg): Remove call to ix86_mitigate_rop.
2369 * config/i386/i386.md (attr "modrm_class"): Remove.
2370 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2371 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2372 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2374 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2376 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2377 allow folding of mergeh() and mergel() for the float and double types.
2378 (fold_mergehl_helper): Rework to handle building a permute tree
2381 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2383 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2386 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2388 * common.opt (fdiagnostics-show-labels): New option.
2389 * diagnostic-show-locus.c (class layout_range): Add field
2391 (class layout): Add field "m_show_labels_p".
2392 (layout_range::layout_range): Add param "label" and use it to
2394 (make_range): Pass in NULL for new "label" param of layout_range's
2396 (layout::layout): Initialize m_show_labels_p.
2397 (layout::maybe_add_location_range): Pass in loc_range->m_label
2398 when constructing layout_range instances.
2399 (struct line_label): New struct.
2400 (layout::print_any_labels): New member function.
2401 (layout::print_line): Call it if label-printing is enabled.
2402 (selftest::test_one_liner_labels): New test.
2403 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2404 * diagnostic.c (diagnostic_initialize): Initialize
2405 context->show_labels_p.
2406 * diagnostic.h (struct diagnostic_context): Add field
2408 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2409 -fno-diagnostics-show-labels.
2410 * dwarf2out.c (gen_producer_string): Add
2411 OPT_fdiagnostics_show_labels to the ignored options.
2412 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2414 (gcc_rich_location::maybe_add_expr): Likewise.
2415 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2416 label" param, defaulting to NULL.
2417 (gcc_rich_location::add_expr): Add "label" param.
2418 (gcc_rich_location::maybe_add_expr): Likewise.
2419 (class text_range_label): New class.
2420 (class range_label_for_type_mismatch): New class.
2421 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2422 of format_warning_va.
2423 (fmtwarn_n): Likewise for new params of format_warning_n_va.
2424 * lto-wrapper.c (merge_and_complain): Add
2425 OPT_fdiagnostics_show_labels to the "pick one setting" options.
2426 (append_compiler_options): Likewise to the dropped options.
2427 (append_diag_options): Likewise to the passed-on options.
2428 * opts.c (common_handle_option): Handle the new option.
2429 * selftest-diagnostic.c
2430 (test_diagnostic_context::test_diagnostic_context): Enable
2432 * substring-locations.c: Include "gcc-rich-location.h".
2433 (format_warning_n_va): Add "fmt_label" and "param_label" params
2434 and use them as appropriate.
2435 (format_warning_va): Add "fmt_label" and "param_label" params,
2436 passing them on to format_warning_n_va.
2437 (format_warning_at_substring): Likewise.
2438 (format_warning_at_substring_n): Likewise.
2439 * substring-locations.h (format_warning_va): Add "fmt_label" and
2440 "param_label" params.
2441 (format_warning_n_va): Likewise.
2442 (format_warning_at_substring): Likewise.
2443 (format_warning_at_substring_n): Likewise.
2444 * toplev.c (general_init): Initialize global_dc->show_labels_p.
2446 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
2449 * builtins.c (expand_builtin_memcmp): Do not expand the call
2450 when overflow is detected.
2452 2018-08-15 Martin Sebor <msebor@redhat.com>
2454 PR tree-optimization/71625
2455 * config/aarch64/aarch64-builtins.c
2456 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
2458 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
2460 * config/s390/s390.c (s390_reorg): Remove loop.
2462 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2465 (darwin_function_switched_text_sections): Delete.
2466 * gcc/config/darwin.h
2467 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2469 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2472 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2473 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
2474 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2476 2018-08-15 Martin Liska <mliska@suse.cz>
2478 PR tree-optimization/86925
2479 * predict.c (expr_expected_value_1): When taking
2480 later predictor, assign also probability.
2481 Use fold_build2_initializer_loc in order to fold
2482 the expression in -frounding-math.
2484 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
2486 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
2488 (expand_vec_perm_1): Try the new method.
2490 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
2493 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
2494 Check whether lra_live_max_point is 0 before dividing.
2496 2018-08-14 Martin Sebor <msebor@redhat.com>
2498 PR tree-optimization/86650
2499 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
2500 (vrp_prop::check_mem_ref): Same.
2502 2018-08-13 Liu Hao <lh_mouse@126.com>
2504 * pretty-print.c (eat_esc_sequence): Swap the foreground and
2505 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2506 and clear it thereafter, as it only works for DBCS.
2508 2018-08-13 Liu Hao <lh_mouse@126.com>
2510 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2511 handle returned by _get_osf_handle().
2513 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2515 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
2516 for folding vec_perm.
2518 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2520 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
2521 Add support for gimple-folding of vec_pack() and vec_unpack()
2524 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2526 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
2527 vec_xst variants to the list.
2528 (rs6000_gimple_fold_builtin): Add support for folding unaligned
2529 vector loads and stores.
2531 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
2533 * config.gcc (rs6000-ibm-aix4.x): Delete.
2534 (rs6000-ibm-aix5.1): Delete.
2535 (rs6000-ibm-aix5.2): Delete.
2536 (rs6000-ibm-aix5.3): Delete.
2537 * config/rs6000/aix43.h: Delete.
2538 * config/rs6000/aix51.h: Delete.
2539 * config/rs6000/aix52.h: Delete.
2540 * config/rs6000/t-aix43: Delete.
2542 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
2544 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2546 (s390_decompose_address): Factor out constant pool ref
2549 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2551 * config/nds32/nds32-predicates.c
2552 (nds32_can_use_bclr_p): Change return type as bool.
2553 (nds32_can_use_bset_p): Ditto.
2554 (nds32_can_use_btgl_p): Ditto.
2555 (nds32_can_use_bitci_p): Ditto.
2556 * config/nds32/nds32-protos.h
2557 (nds32_can_use_bclr_p): Change declaration.
2558 (nds32_can_use_bset_p): Ditto.
2559 (nds32_can_use_btgl_p): Ditto.
2560 (nds32_can_use_bitci_p): Ditto.
2562 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2564 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
2565 Support -msched-prolog-epilog option.
2566 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
2568 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2570 * common/config/nds32/nds32-common.c
2571 (nds32_option_optimization_table): Enalbe -malways-align.
2573 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2575 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
2577 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
2578 OPT_misr_secure_ case.
2579 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
2580 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
2581 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
2583 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
2584 (nds32_isr_info): New field security_level.
2585 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
2586 * config/nds32/nds32.md (return_internal): Consider critical attribute.
2587 * config/nds32/nds32.opt (misr-secure): New option.
2588 * config/nds32/nds32_init.inc: New file.
2589 * config/nds32/nds32_isr.h: New file.
2591 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2593 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2594 Update comment for atomic instructions.
2595 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2596 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2598 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2599 load for __ATOMIC_SEQ_CST model.
2600 (atomic_loaddi_1): Remove float register target.
2601 (atomic_storedi): Handle CONST_INT values.
2602 (atomic_storedi_1): Remove float register source. Add special case
2604 (memory_barrier): New expander and insn.
2606 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2608 PR tree-optimization/86835
2609 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2610 new_stmt after def_gsi, make sure to insert new_square_stmt after
2611 that stmt, not 2 stmts before it.
2613 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
2616 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2619 2018-08-10 Martin Liska <mliska@suse.cz>
2622 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2624 * builtins.c (expand_builtin_expect_with_probability):
2626 (expand_builtin_expect_with_probability): New function.
2627 (build_builtin_expect_predicate): Add new argumnet probability
2628 for BUILT_IN_EXPECT_WITH_PROBABILITY.
2629 (fold_builtin_expect):
2632 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2633 * builtins.h (fold_builtin_expect): Set new argument.
2634 * doc/extend.texi: Document __builtin_expect_with_probability.
2635 * doc/invoke.texi: Likewise.
2636 * gimple-fold.c (gimple_fold_call): Pass new argument.
2637 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2638 also BUILT_IN_EXPECT_WITH_PROBABILITY.
2639 * predict.c (get_predictor_value): New function.
2640 (expr_expected_value): Add new argument probability. Assume
2641 that predictor and probability are always non-null.
2642 (expr_expected_value_1): Likewise. For __builtin_expect and
2643 __builtin_expect_with_probability set probability. Handle
2644 combination in binary expressions.
2645 (tree_predict_by_opcode): Simplify code by simply calling
2646 get_predictor_value.
2647 (pass_strip_predict_hints::execute): Add handling of
2648 BUILT_IN_EXPECT_WITH_PROBABILITY.
2649 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2651 * tree.h (DECL_BUILT_IN_P): New function.
2653 2018-08-10 Martin Liska <mliska@suse.cz>
2655 PR tree-optimization/85799
2656 * passes.def: Add argument for pass_strip_predict_hints.
2657 * predict.c (class pass_strip_predict_hints): Add new argument
2659 (strip_predictor_early): New function.
2660 (pass_strip_predict_hints::execute): Call the function to
2662 (strip_predict_hints): New function.
2663 * predict.def: Fix comment.
2665 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2667 * Makefile.in: Clarify which tm.texi to copy over to assert the
2668 right to grant a GFDL license for all.
2670 2018-08-09 Jeff Law <law@redhat.com>
2672 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2675 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2677 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2680 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2682 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2683 reductions for variable-length vectors.
2685 2018-08-09 David Malcolm <dmalcolm@redhat.com>
2688 * common.opt (fdiagnostics-show-line-numbers): New option.
2689 * diagnostic-show-locus.c (class layout): Add fields
2690 "m_show_line_numbers_p" and "m_linenum_width";
2691 (num_digits): New function.
2692 (test_num_digits): New function.
2693 (layout::layout): Initialize new fields. Update m_x_offset
2694 logic to handle any left margin.
2695 (layout::print_source_line): Print line number when requested.
2696 (layout::start_annotation_line): New member function.
2697 (layout::print_annotation_line): Call it.
2698 (layout::print_leading_fixits): Likewise.
2699 (layout::print_trailing_fixits): Likewise. Update calls to
2700 move_to_column for new parameter.
2701 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2702 it to potentially call start_annotation_line.
2703 (layout::show_ruler): Call start_annotation_line.
2704 (selftest::test_line_numbers_multiline_range): New selftest.
2705 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2706 and selftest::test_line_numbers_multiline_range.
2707 * diagnostic.c (diagnostic_initialize): Initialize
2708 show_line_numbers_p.
2709 * diagnostic.h (struct diagnostic_context): Add field
2710 "show_line_numbers_p".
2711 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2712 -fno-diagnostics-show-line-numbers.
2713 * dwarf2out.c (gen_producer_string): Add
2714 OPT_fdiagnostics_show_line_numbers to the ignored options.
2715 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2716 one setting" options.
2717 (append_compiler_options): Likewise to the dropped options.
2718 (append_diag_options): Likewise to the passed-on options.
2719 * opts.c (common_handle_option): Handle the new option.
2720 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2722 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
2724 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2725 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
2726 third argument of type "const signed char" to descriptions of
2727 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2728 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2729 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2730 __builtin_bcdsub_ov functions.
2732 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2734 PR tree-optimization/86858
2735 * tree-vect-loop.c (vect_is_simple_reduction): Restore
2736 flow_bb_inside_loop_p calls.
2738 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2740 PR tree-optimization/86871
2741 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2742 instead of gimple_assign_lhs.
2744 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
2747 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
2748 register constraint to operand 0.
2749 (add<mode>3_carryinC): Likewise.
2750 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
2752 2018-08-09 Martin Liska <mliska@suse.cz>
2755 * common.opt: Remove extra line.
2757 2018-08-09 Martin Liska <mliska@suse.cz>
2759 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
2760 at the end of a line, make first letter capital and end up
2761 a sentence with a dot.
2762 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
2763 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
2764 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
2765 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
2766 (PARAM_MAX_ISL_OPERATIONS): Likewise.
2767 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
2768 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
2769 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
2770 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
2771 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
2772 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
2773 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
2774 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
2775 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
2776 (PARAM_TREE_REASSOC_WIDTH): Likewise.
2777 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
2778 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
2779 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
2781 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
2784 * config/s390/s390.c (s390_option_override_internal): Reduce the
2785 stack-clash-protection-probe-interval param if it would be too big
2788 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
2791 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
2792 * config/m68k/m68k.c (handle_move_double): Don't call
2793 m68k_final_prescan_insn.
2794 (m68k_adjust_decorated_operand): Renamed from
2795 m68k_final_prescan_insn, remove first and third operand and
2797 (print_operand): Call it.
2798 (print_operand_address): Call it.
2800 2018-08-08 Nathan Sidwell <nathan@acm.org>
2802 * diagnostic.c (diagnostic_report_current_module): Use
2803 linemap_included_from & linemap_included_from_linemap.
2805 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
2807 * config/aarch64/aarch64-cores.def: Add phecda core.
2808 * config/aarch64/aarch64-tune.md: Regenerate.
2809 * doc/invoke.texi: Add phecda core.
2811 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2814 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
2816 * config/s390/s390.md ("movti"): Add more alternatives for
2817 constant to GPR copies.
2819 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2821 * config/s390/s390.c: Fix whitespace damage throughout the file.
2822 * config/s390/s390.h: Likewise.
2823 * config/s390/tpf.h: Likewise.
2825 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2827 * config/s390/s390.c (s390_loadrelative_operand_p):
2828 Remove TARGET_CPU_ZARCH usages.
2829 (s390_rtx_costs): Likewise.
2830 (s390_legitimate_constant_p): Likewise.
2831 (s390_cannot_force_const_mem): Likewise.
2832 (legitimate_reload_constant_p): Likewise.
2833 (s390_preferred_reload_class): Likewise.
2834 (legitimize_pic_address): Likewise.
2835 (legitimize_tls_address): Likewise.
2836 (s390_split_branches): Removed.
2837 (s390_add_execute): Removed.
2838 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
2839 (s390_mainpool_start): Likewise.
2840 (s390_mainpool_finish): Likewise.
2841 (s390_mainpool_cancel): Removed.
2842 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
2843 (s390_chunkify_cancel): Likewise.
2844 (s390_return_addr_rtx): Likewise.
2845 (s390_register_info): Remove split_branches_pending_p uages.
2846 (s390_optimize_register_info): Likewise.
2847 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
2848 split_branches_pending_p usages.
2849 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
2850 (s390_load_got): Likewise.
2851 (s390_expand_split_stack_prologue): Likewise.
2852 (output_asm_nops): Likewise.
2853 (s390_function_profiler): Likewise.
2854 (s390_emit_call): Likewise.
2855 (s390_conditional_register_usage): Likewise.
2856 (s390_optimize_prologue): Likewise.
2857 (s390_reorg): Remove TARGET_CPU_ZARCH and
2858 split_branches_pending_p usages.
2859 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
2861 (s390_output_indirect_thunk_function): Likewise.
2862 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
2863 (TARGET_CPU_ZARCH_P): Removed.
2864 (struct machine_function): Remove split_branches_pending_p.
2865 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
2867 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2869 * common/config/s390/s390-common.c (processor_flags_table):
2871 * config.gcc: Remove with_arch/with_tune support.
2872 * config/s390/2064.md: Remove cpu attribute comparisons.
2873 * config/s390/driver-native.c (s390_host_detect_local_cpu):
2875 * config/s390/linux.h (ASM_SPEC):
2876 Remove -march support.
2877 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
2878 Use a table to get an arch level.
2879 * config/s390/s390-opts.h (enum processor_type):
2881 * config/s390/s390.c
2882 (processor_table): Remove entries, add arch_level values.
2883 (s390_issue_rate): Remove cases.
2884 (s390_option_override): Adjust
2885 s390_option_override_internal() call.
2886 (s390_option_override_internal): Remove deprecation warning.
2887 (s390_valid_target_attribute_tree): Adjust
2888 s390_option_override_internal() call.
2889 * config/s390/s390.h (struct s390_processor):
2890 Share with s390-c.c, add arch_level field.
2891 * config/s390/s390.md:
2892 Remove occurrences in cpu attribute.
2893 * config/s390/s390.opt: Remove -march/-mtune support.
2894 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
2895 * doc/invoke.texi: Remove deprecation warning.
2897 2018-08-08 Luis Machado <luis.machado@linaro.org>
2899 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
2901 (qdf24xx_tunings): Set vector cost structure to
2902 qdf24xx_vector_cost.
2904 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2905 <register_sextend>: Set to 3.
2907 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
2910 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
2911 * config/aarch64/aarch64-simd.md
2912 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
2913 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
2914 (aarch64_frecpx<mode>): ...this new pattern.
2915 * config/aarch64/aarch64-simd-builtins.def: Remove comment
2916 about aarch64_frecp<FRECP:frecp_suffix><mode>.
2918 2018-08-07 Martin Liska <mliska@suse.cz>
2921 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
2922 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
2923 * predict.def (PRED_MALLOC_NONNULL): New predictor.
2924 * doc/extend.texi: Document that malloc attribute adds
2927 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
2930 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2931 Define to speculation_safe_value_not_needed.
2933 2018-08-06 Jeff Law <law@redhat.com>
2935 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
2936 the vr_values instance to cprop_into_stmt.
2937 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
2938 (cprop_operand): Also query EVRP to determine if OP is a constant.
2940 2018-08-06 Nathan Sidwell <nathan@acm.org>
2942 * diagnostic.c (diagnostic_report_current_module): Reroll
2943 included-at loop. Translate text.
2945 2018-08-06 David Malcolm <dmalcolm@redhat.com>
2947 * function-tests.c (selftest::test_expansion_to_rtl): Call
2948 free_after_compilation.
2950 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2952 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
2954 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
2956 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
2957 loops with memory block operations from getting unrolled.
2959 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
2962 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2963 Define to speculation_safe_value_not_needed.
2965 2018-08-06 Jeff Law <law@redhat.com>
2967 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
2970 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2973 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
2974 with all enabled __intN types.
2976 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
2978 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2980 * alias.c (record_set): Check for clobber high.
2981 * cfgexpand.c (expand_gimple_stmt): Likewise.
2982 * combine-stack-adj.c (single_set_for_csa): Likewise.
2983 * combine.c (find_single_use_1): Likewise.
2984 (set_nonzero_bits_and_sign_copies): Likewise.
2985 (get_combine_src_dest): Likewise.
2986 (is_parallel_of_n_reg_sets): Likewise.
2987 (try_combine): Likewise.
2988 (record_dead_and_set_regs_1): Likewise.
2989 (reg_dead_at_p_1): Likewise.
2990 (reg_dead_at_p): Likewise.
2991 * dce.c (deletable_insn_p): Likewise.
2992 (mark_nonreg_stores_1): Likewise.
2993 (mark_nonreg_stores_2): Likewise.
2994 * df-scan.c (df_find_hard_reg_defs): Likewise.
2995 (df_uses_record): Likewise.
2996 (df_get_call_refs): Likewise.
2997 * dwarf2out.c (mem_loc_descriptor): Likewise.
2998 * haifa-sched.c (haifa_classify_rtx): Likewise.
2999 * ira-build.c (create_insn_allocnos): Likewise.
3000 * ira-costs.c (scan_one_insn): Likewise.
3001 * ira.c (equiv_init_movable_p): Likewise.
3002 (rtx_moveable_p): Likewise.
3003 (interesting_dest_for_shprep): Likewise.
3004 * jump.c (mark_jump_label_1): Likewise.
3005 * postreload-gcse.c (record_opr_changes): Likewise.
3006 * postreload.c (reload_cse_simplify): Likewise.
3007 (struct reg_use): Add source expr.
3008 (reload_combine): Check for clobber high.
3009 (reload_combine_note_use): Likewise.
3010 (reload_cse_move2add): Likewise.
3011 (move2add_note_store): Likewise.
3012 * print-rtl.c (print_pattern): Likewise.
3013 * recog.c (decode_asm_operands): Likewise.
3014 (store_data_bypass_p): Likewise.
3015 (if_test_bypass_p): Likewise.
3016 * regcprop.c (kill_clobbered_value): Likewise.
3017 (kill_set_value): Likewise.
3018 * reginfo.c (reg_scan_mark_refs): Likewise.
3019 * reload1.c (maybe_fix_stack_asms): Likewise.
3020 (eliminate_regs_1): Likewise.
3021 (elimination_effects): Likewise.
3022 (mark_not_eliminable): Likewise.
3023 (scan_paradoxical_subregs): Likewise.
3024 (forget_old_reloads_1): Likewise.
3025 * reorg.c (find_end_label): Likewise.
3026 (try_merge_delay_insns): Likewise.
3027 (redundant_insn): Likewise.
3028 (own_thread_p): Likewise.
3029 (fill_simple_delay_slots): Likewise.
3030 (fill_slots_from_thread): Likewise.
3031 (dbr_schedule): Likewise.
3032 * resource.c (update_live_status): Likewise.
3033 (mark_referenced_resources): Likewise.
3034 (mark_set_resources): Likewise.
3035 * rtl.c (copy_rtx): Likewise.
3036 * rtlanal.c (reg_referenced_p): Likewise.
3037 (single_set_2): Likewise.
3038 (noop_move_p): Likewise.
3039 (note_stores): Likewise.
3040 * sched-deps.c (sched_analyze_reg): Likewise.
3041 (sched_analyze_insn): Likewise.
3043 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3045 * cse.c (invalidate_reg): New function extracted from...
3046 (invalidate): ...here.
3047 (canonicalize_insn): Check for clobber high.
3048 (invalidate_from_clobbers): invalidate clobber highs.
3049 (invalidate_from_sets_and_clobbers): Likewise.
3050 (count_reg_usage): Check for clobber high.
3051 (insn_live_p): Likewise.
3052 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3053 (cselib_invalidate_regno): Check for clobber in setter.
3054 (cselib_invalidate_rtx): Pass through setter.
3055 (cselib_invalidate_rtx_note_stores):
3056 (cselib_process_insn): Check for clobber high.
3057 * cselib.h (cselib_invalidate_rtx): Add operand.
3059 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3061 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3062 (mark_not_eliminable): Likewise.
3063 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3064 * lra-lives.c (process_bb_lives): Check for clobber high.
3065 * lra.c (new_insn_reg): Remember clobber highs.
3066 (collect_non_operand_hard_regs): Check for clobber high.
3067 (lra_set_insn_recog_data): Likewise.
3068 (add_regs_to_insn_regno_info): Likewise.
3069 (lra_update_insn_regno_info): Likewise.
3071 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3073 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3074 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3076 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3078 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3079 (copy_insn_1): Likewise.
3080 (gen_hard_reg_clobber_high): New gen function.
3081 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3082 * genemit.c (gen_exp): Likewise.
3083 (gen_emit_seq): Pass through info.
3084 (gen_insn): Check for CLOBBER_HIGH.
3085 (gen_expand): Pass through info.
3086 (gen_split): Likewise.
3087 (output_add_clobbers): Likewise.
3088 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3089 (remove_clobbers): Likewise.
3090 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3092 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3094 * doc/rtl.texi (clobber_high): Add.
3095 (parallel): Add in clobber high
3096 * rtl.c (rtl_check_failed_code3): Add function.
3097 * rtl.def (CLOBBER_HIGH): Add expression.
3098 * rtl.h (RTL_CHECKC3): Add macro.
3099 (rtl_check_failed_code3): Add declaration.
3100 (XC3EXP): Add macro.
3102 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3105 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3106 cfun->machine->max_used_stack_alignment if needed.
3108 2018-08-04 Martin Sebor <msebor@redhat.com>
3110 PR tree-optimization/86571
3111 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3114 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3117 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3120 2018-08-03 Jeff Law <law@redhat.com>
3123 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3124 Define to speculation_safe_value_not_needed.
3126 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3128 * doc/gcov.texi (-x): Remove duplicate "to".
3129 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3130 (-Wif-not-aligned): Remove duplicate "is".
3131 (-flto): Remove duplicate "the".
3132 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3133 duplicate "v5.00.b".
3134 (MSP430 Options): Remove duplicate "and" from the description
3135 of "-mgprel-sec=regexp".
3136 (x86 Options): Remove duplicate copies of "vmldLog102" and
3137 vmlsLog104 from description of "-mveclibabi=type".
3139 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3141 * internal-fn.h (first_commutative_argument): Declare.
3142 * internal-fn.c (first_commutative_argument): New function.
3143 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3144 restrictions for pattern statements. Use first_commutative_argument
3145 to look for commutative operands in calls to internal functions.
3147 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3149 * Makefile.in (wide-int-range.o): New.
3150 * tree-vrp.c: Move all the wide_int_* functions to...
3151 * wide-int-range.cc: ...here.
3152 * tree-vrp.h: Move all the wide_int_* prototypes to...
3153 * wide-int-range.h: ...here.
3155 2018-08-03 Tom de Vries <tdevries@suse.de>
3157 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3159 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3160 * except.c (output_function_exception_table): Do early exit if
3161 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3163 2018-08-03 Martin Liska <mliska@suse.cz>
3165 * predict.c (dump_prediction): Change to 2 digits
3168 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3170 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3172 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3174 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3176 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3177 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3178 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3179 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3180 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3182 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3184 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3185 pp_take_prefix when saving the existing prefix.
3186 * diagnostic.c (diagnostic_append_note): Likewise.
3187 * langhooks.c (lhd_print_error_function): Likewise.
3188 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3189 param's type. Free the existing prefix.
3190 (pp_take_prefix): New function.
3191 (pretty_printer::pretty_printer): Drop the prefix parameter.
3192 Rename the length parameter to match the comment.
3193 (pretty_printer::~pretty_printer): Free the prefix.
3194 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3196 (struct pretty_printer): Drop the "const" from "prefix" field's
3197 type and clarify memory management.
3198 (pp_set_prefix): Drop the "const" from the 2nd param.
3199 (pp_take_prefix): New decl.
3201 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3203 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3204 for word_mode_ok here instead of passing as argument.
3205 (expand_block_compare): Change select_block_compare_mode() call.
3206 (expand_strncmp_gpr_sequence): New function.
3207 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3209 2018-08-02 Jeff Law <law@redhat.com>
3212 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3213 Define to speculation_safe_value_not_needed.
3216 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3217 Define to speculation_safe_value_not_needed.
3219 2018-08-02 Tom de Vries <tdevries@suse.de>
3222 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3223 function. Return UI_TARGET unconditionally.
3224 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3225 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3227 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3229 * genemit.c (print_overload_test): Fix typo.
3231 2018-08-02 Richard Biener <rguenther@suse.de>
3233 PR tree-optimization/86816
3234 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3235 which checks for value availability before querying it.
3236 (gvn_uses_equal): Use it.
3237 (same_succ_hash): Likewise.
3238 (gimple_equal_p): Likewise.
3240 2018-08-02 Nick Clifton <nickc@redhat.com>
3243 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3244 Define to speculation_safe_value_not_needed.
3247 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3248 Define to speculation_safe_value_not_needed.
3251 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3252 Define to speculation_safe_value_not_needed.
3255 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3256 Define to speculation_safe_value_not_needed.
3259 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3260 Define to speculation_safe_value_not_needed.
3263 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3264 Define to speculation_safe_value_not_needed.
3267 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3268 Define to speculation_safe_value_not_needed.
3271 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3272 Define to speculation_safe_value_not_needed.
3275 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3276 speculation_safe_value_not_needed.
3279 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3280 to speculation_safe_value_not_needed.
3282 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3284 * doc/md.texi: Expand the documentation of instruction names
3285 to mention port-local uses. Document '@' in pattern names.
3286 * read-md.h (overloaded_instance, overloaded_name): New structs.
3288 (md_reader::handle_overloaded_name): New member function.
3289 (md_reader::get_overloads): Likewise.
3290 (md_reader::m_first_overload): New member variable.
3291 (md_reader::m_next_overload_ptr): Likewise.
3292 (md_reader::m_overloads_htab): Likewise.
3293 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3294 m_next_overload_ptr and m_overloads_htab.
3295 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3296 (get_mode_token, get_code_token, get_int_token): New functions.
3297 (map_attr_string): Add an optional argument that passes back
3298 the associated iterator.
3299 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3300 (md_reader::handle_overloaded_name, add_overload_instance): New
3302 (apply_iterators): Handle '@' names. Report an error if '@'
3303 is used without iterators.
3304 (initialize_iterators): Initialize the new iterator_group fields.
3305 * genopinit.c (handle_overloaded_code_for)
3306 (handle_overloaded_gen): New functions.
3307 (main): Use them to print declarations of maybe_code_for_* and
3308 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3309 * genemit.c (print_overload_arguments, print_overload_test)
3310 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3311 (main): Use it to print definitions of maybe_code_for_* and
3312 maybe_gen_* functions.
3313 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3314 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3315 instead of explicit mode checks.
3316 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3317 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3318 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3319 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3320 (aarch64_expand_compare_and_swap): Likewise
3321 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3322 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3323 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3324 (aarch64_constant_pool_reload_icode): Delete.
3325 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3326 instead of aarch64_constant_pool_reload_icode. Use
3327 code_for_aarch64_reload_mov instead of explicit mode checks.
3328 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3329 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3330 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3331 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3332 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3333 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3334 (aarch64_atomic_load_op_code): Delete.
3335 (aarch64_emit_atomic_load_op): Likewise.
3336 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3337 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3338 instead of aarch64_emit_atomic_load_op.
3339 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3340 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3341 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3342 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3343 character before the pattern name.
3344 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3345 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3346 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3347 (aarch64_frecps<mode>): Likewise.
3348 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3349 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3350 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3351 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3352 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3354 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3356 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3357 Allow HFmode constants if TARGET_FP_F16INST.
3359 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3362 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3363 No longer check last store for clobber of address register.
3365 2018-08-02 Martin Liska <mliska@suse.cz>
3367 PR gcov-profile/86817
3368 * gcov.c (process_all_functions): New function.
3370 (process_file): Move functions processing to
3371 process_all_functions.
3373 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3375 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3376 "const" to the "gimple *" and "rtx_insn *" parameters.
3377 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3379 (dump_location_t::dump_location_t): Likewise.
3381 2018-08-01 Martin Sebor <msebor@redhat.com>
3383 PR tree-optimization/86650
3384 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3385 rather than a "gcall *". Directly pass the data of interest
3386 to percent_K_format, rather than building a temporary CALL_EXPR
3388 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3389 (gimple_fold_builtin_strncat): Adjust.
3390 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3391 gcall* argument with gimple*.
3392 * gimple-ssa-warn-restrict.c (check_call): Same.
3393 (wrestrict_dom_walker::before_dom_children): Same.
3394 (builtin_access::builtin_access): Same.
3395 (check_bounds_or_overlap): Same
3396 (maybe_diag_overlap): Same.
3397 (maybe_diag_offset_bounds): Same.
3398 * tree-diagnostic.c (default_tree_printer): Move usage of
3399 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3401 * tree-pretty-print.c (percent_K_format): Add argument.
3402 * tree-pretty-print.h: Add argument.
3403 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3404 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3405 (maybe_diag_stxncpy_trunc): Same.
3406 (handle_builtin_stxncpy): Same.
3407 (handle_builtin_strcat): Same.
3409 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3411 * match.pd: Optimise pointer range checks.
3413 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3415 PR tree-optimization/86758
3416 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3417 to remove pattern statements.
3419 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3421 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3422 result of dfs_enumerate_from when constructing stmt_vec_infos,
3423 instead of additionally calling get_loop_body.
3425 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3427 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3429 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
3430 When creating an iv, assert that the step is not known to be zero.
3431 (vect_setup_realignment): Update call accordingly.
3432 * tree-vect-stmts.c (vectorizable_store): Likewise.
3433 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
3435 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3437 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
3438 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
3439 (vectorizable_reduction): Likewise.
3440 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3441 (vect_detect_hybrid_slp_stmts): Likewise.
3442 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3444 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
3446 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
3447 (wide_int_set_zero_nonzero_bits): ...this.
3448 (zero_nonzero_bits_from_vr): Rename to...
3449 (vrp_set_zero_nonzero_bits): ...this.
3450 (extract_range_from_multiplicative_op_1): Abstract wide int
3452 (wide_int_range_multiplicative_op): ...here.
3453 (extract_range_from_binary_expr_1): Extract wide int binary
3454 operations into their own functions.
3455 (wide_int_range_lshift): New.
3456 (wide_int_range_can_optimize_bit_op): New.
3457 (wide_int_range_shift_undefined_p): New.
3458 (wide_int_range_bit_xor): New.
3459 (wide_int_range_bit_ior): New.
3460 (wide_int_range_bit_and): New.
3461 (wide_int_range_trunc_mod): New.
3462 (extract_range_into_wide_ints): New.
3463 (vrp_shift_undefined_p): New.
3464 (extract_range_from_multiplicative_op): New.
3465 (vrp_can_optimize_bit_op): New.
3466 * tree-vrp.h (value_range::dump): New.
3467 (wide_int_range_multiplicative_op): New.
3468 (wide_int_range_lshift):New.
3469 (wide_int_range_shift_undefined_p): New.
3470 (wide_int_range_bit_xor): New.
3471 (wide_int_range_bit_ior): New.
3472 (wide_int_range_bit_and): New.
3473 (wide_int_range_trunc_mod): New.
3474 (zero_nonzero_bits_from_bounds): Rename to...
3475 (wide_int_set_zero_nonzero_bits): ...this.
3476 (zero_nonzero_bits_from_vr): Rename to...
3477 (vrp_set_zero_nonzero_bits): ...this.
3478 (range_easy_mask_min_max): Rename to...
3479 (wide_int_range_can_optimize_bit_op): this.
3480 * vr-values.c (simplify_bit_ops_using_ranges): Rename
3481 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
3483 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3485 * tree-vectorizer.h (vect_orig_stmt): New function.
3486 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
3487 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
3488 (vect_create_epilog_for_reduction): Likewise.
3489 (vectorizable_live_operation): Likewise.
3490 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3491 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
3492 * tree-vect-stmts.c (vectorizable_call): Likewise.
3493 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
3495 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3497 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
3499 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
3500 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
3501 (vect_transform_loop): Likewise.
3502 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3504 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3506 * tree-vectorizer.h (vect_schedule_slp): Return void.
3507 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3508 (vect_schedule_slp): Likewise.
3510 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3512 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
3514 (vect_transform_loop): Update calls accordingly. Schedule SLP
3515 instances before the main loop, if any exist.
3517 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3519 PR tree-optimization/86749
3520 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
3521 If the lhs is used in a COND_EXPR, check that it is being used
3522 as the "then" or "else" value.
3524 2018-08-01 Tom de Vries <tdevries@suse.de>
3527 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3528 speculation_safe_value_not_needed.
3530 2018-08-01 Richard Biener <rguenther@suse.de>
3532 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
3535 2018-08-01 Martin Liska <mliska@suse.cz>
3537 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
3538 format how successful transformation is dumped.
3539 (gimple_mod_pow2_value_transform): Likewise.
3540 (gimple_mod_subtract_transform): Likewise.
3541 (gimple_stringops_transform): Likewise.
3543 2018-08-01 Martin Liska <mliska@suse.cz>
3546 * value-prof.c (interesting_stringop_to_profile_p):
3547 Simplify the code and add BUILT_IN_MEMMOVE.
3548 (gimple_stringops_transform): Likewise.
3550 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
3552 * config/aarch64/aarch64-simd.md
3553 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
3554 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
3555 use GPI iterator instead of SI mode.
3557 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3559 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
3560 rs6000_speculation_barrier.
3561 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
3562 new barrier pattern name.
3564 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3566 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
3567 (speculation_barrier): New insn.
3569 2018-08-01 Richard Biener <rguenther@suse.de>
3572 * graphite.h: Include isl/id.h and isl/space.h to allow build
3575 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3578 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3579 mode for COFF targets.
3580 * defaults.h (TARGET_COFF): Define.
3581 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3582 TARGET_COFF): Define.
3583 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3584 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3585 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3587 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3589 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
3590 * gimple-streamer-out.c (output_bb): Save it.
3591 * lto-streamer-in.c (input_struct_function_base): Restore
3592 instance discriminator if available. Create map on demand.
3593 * lto-streamer-out.c (output_struct_function_base): Save it if
3595 * final.c (decl_to_instance_map): Document LTO strategy.
3597 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3598 Olivier Hainque <hainque@adacore.com>
3600 * debug.h (decl_to_instance_map_t): New type.
3601 (decl_to_instance_map): Declare.
3602 (maybe_create_decl_to_instance_map): New inline function.
3603 * final.c (bb_discriminator, last_bb_discriminator): New statics,
3604 to track basic block discriminators.
3605 (final_start_function_1): Initialize them.
3606 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3608 (decl_to_instance_map): New variable.
3609 (map_decl_to_instance, maybe_set_discriminator): New functions.
3610 (notice_source_line): Set discriminator.
3612 2018-07-31 Ian Lance Taylor <iant@golang.org>
3614 * targhooks.c (default_have_speculation_safe_value): Add
3617 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3619 * dump-context.h: Include "pretty-print.h".
3620 (dump_context::refresh_dumps_are_enabled): New decl.
3621 (dump_context::emit_item): New decl.
3622 (class dump_context): Add fields "m_test_pp" and
3624 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3625 (temp_dump_context::get_dumped_text): New decl.
3626 (class temp_dump_context): Add field "m_pp".
3627 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3628 (dump_context::refresh_dumps_are_enabled): ...and add a test for
3630 (set_dump_file): Update for above change.
3631 (set_alt_dump_file): Likewise.
3632 (dump_loc): New overload, taking a pretty_printer *.
3633 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
3634 to any test pretty-printer.
3635 (make_item_for_dump_gimple_stmt): New function, adapted from
3636 optinfo::add_gimple_stmt.
3637 (dump_context::dump_gimple_stmt): Call it, and use the result,
3638 eliminating the direct usage of dump_file and alt_dump_file in
3639 favor of indirectly using them via emit_item.
3640 (make_item_for_dump_gimple_expr): New function, adapted from
3641 optinfo::add_gimple_expr.
3642 (dump_context::dump_gimple_expr): Call it, and use the result,
3643 eliminating the direct usage of dump_file and alt_dump_file in
3644 favor of indirectly using them via emit_item.
3645 (make_item_for_dump_generic_expr): New function, adapted from
3647 (dump_context::dump_generic_expr): Call it, and use the result,
3648 eliminating the direct usage of dump_file and alt_dump_file in
3649 favor of indirectly using them via emit_item.
3650 (make_item_for_dump_printf_va): New function, adapted from
3651 optinfo::add_printf_va.
3652 (make_item_for_dump_printf): New function.
3653 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3654 and use the result, eliminating the direct usage of dump_file and
3655 alt_dump_file in favor of indirectly using them via emit_item.
3656 (make_item_for_dump_dec): New function.
3657 (dump_context::dump_dec): Call it, and use the result,
3658 eliminating the direct usage of dump_file and alt_dump_file in
3659 favor of indirectly using them via emit_item.
3660 (make_item_for_dump_symtab_node): New function, adapted from
3661 optinfo::add_symtab_node.
3662 (dump_context::dump_symtab_node): Call it, and use the result,
3663 eliminating the direct usage of dump_file and alt_dump_file in
3664 favor of indirectly using them via emit_item.
3665 (dump_context::begin_scope): Reimplement, avoiding direct usage
3666 of dump_file and alt_dump_file in favor of indirectly using them
3668 (dump_context::emit_item): New member function.
3669 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3670 Set up test pretty-printer on the underlying context. Call
3671 refresh_dumps_are_enabled.
3672 (temp_dump_context::~temp_dump_context): Call
3673 refresh_dumps_are_enabled.
3674 (temp_dump_context::get_dumped_text): New member function.
3675 (selftest::verify_dumped_text): New function.
3676 (ASSERT_DUMPED_TEXT_EQ): New macro.
3677 (selftest::test_capture_of_dump_calls): Run all tests twice, with
3678 and then without optinfo enabled. Add uses of
3679 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
3680 * dumpfile.h: Update comment for the dump_* API.
3681 * optinfo-emit-json.cc
3682 (selftest::test_building_json_from_dump_calls): Update for new
3683 param for temp_dump_context ctor.
3684 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3685 and "m_owned" field.
3686 (optinfo_item::~optinfo_item): Likewise.
3687 (optinfo::add_item): New member function.
3688 (optinfo::emit): Update comment.
3689 (optinfo::add_string): Delete.
3690 (optinfo::add_printf): Delete.
3691 (optinfo::add_printf_va): Delete.
3692 (optinfo::add_gimple_stmt): Delete.
3693 (optinfo::add_gimple_expr): Delete.
3694 (optinfo::add_tree): Delete.
3695 (optinfo::add_symtab_node): Delete.
3696 (optinfo::add_dec): Delete.
3697 * optinfo.h (class dump_context): New forward decl.
3698 (optinfo::add_item): New decl.
3699 (optinfo::add_string): Delete.
3700 (optinfo::add_printf): Delete.
3701 (optinfo::add_printf_va): Delete.
3702 (optinfo::add_gimple_stmt): Delete.
3703 (optinfo::add_gimple_expr): Delete.
3704 (optinfo::add_tree): Delete.
3705 (optinfo::add_symtab_node): Delete.
3706 (optinfo::add_dec): Delete.
3707 (optinfo::add_poly_int): Delete.
3708 (optinfo_item::optinfo_item): Remove "owned" param.
3709 (class optinfo_item): Remove field "m_owned".
3711 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3714 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3715 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3716 requested variable alignment.
3717 (expand_one_ssa_partition): Likewise.
3718 (expand_one_var): Likewise.
3720 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3722 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3723 to speculation_safe_value_not_needed.
3725 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3727 * targhooks.h (speculation_safe_value_not_needed): New prototype.
3728 * targhooks.c (speculation_safe_value_not_needed): New function.
3729 * target.def (have_speculation_safe_value): Update documentation.
3730 * doc/tm.texi: Regenerated.
3732 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3734 * config/aarch64/iterators.md (ALLI_TI): New iterator.
3735 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
3737 (despeculate_copy<ALLI:mode>_insn): New insn.
3738 (despeculate_copyti_insn): New insn.
3739 (despeculate_simple<ALLI:mode>): New insn
3740 (despeculate_simpleti): New insn.
3741 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
3743 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
3744 aarch64_speculation_safe_value.
3745 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
3747 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3749 * config/aarch64/aarch64-speculation.cc: New file.
3750 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
3751 before pass_reorder_blocks.
3752 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
3754 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
3755 X14 and X15 when tracking speculation.
3756 * config/aarch64/aarch64.md (register name constants): Add
3757 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
3758 (unspec): Add UNSPEC_SPECULATION_TRACKER.
3759 (speculation_barrier): New insn attribute.
3760 (cmp<mode>): Allow SP in comparisons.
3761 (speculation_tracker): New insn.
3762 (speculation_barrier): Add speculation_barrier attribute.
3763 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
3764 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
3765 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
3767 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3769 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
3770 aarch64_track_speculation is true.
3771 (tb<optab><mode>1): Likewise.
3772 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
3773 generate CB[N]Z when tracking speculation.
3774 (aarch64_split_compare_and_swap): Likewise.
3775 (aarch64_split_atomic_op): Likewise.
3777 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3779 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
3781 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3783 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
3784 (speculation_barrier): New insn.
3786 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3788 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
3789 * config/arm/arm.md (speculation_barrier): New expand.
3790 (speculation_barrier_insn): New pattern.
3792 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3794 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
3795 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
3796 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
3797 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
3799 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
3800 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
3801 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
3802 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
3803 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
3804 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
3805 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
3806 * builtins.c (expand_speculation_safe_value): New function.
3807 (expand_builtin): Call it.
3808 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
3809 * doc/extend.texi: Document __builtin_speculation_safe_value.
3810 * doc/md.texi: Document "speculation_barrier" pattern.
3811 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
3812 TARGET_HAVE_SPECULATION_SAFE_VALUE.
3813 * doc/tm.texi: Regenerated.
3814 * target.def (have_speculation_safe_value, speculation_safe_value): New
3816 * targhooks.c (default_have_speculation_safe_value): New function.
3817 (default_speculation_safe_value): New function.
3818 * targhooks.h (default_have_speculation_safe_value): Add prototype.
3819 (default_speculation_safe_value): Add prototype.
3821 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3823 * dump-context.h (dump_context::dump_loc): New decl.
3824 * dumpfile.c (dump_context::dump_loc): New member function.
3825 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
3826 and dump_gimple_stmt.
3827 (dump_context::dump_gimple_expr_loc): Likewise, using
3829 (dump_context::dump_generic_expr_loc): Likewise, using
3831 (dump_context::dump_printf_loc_va): Likewise, using
3833 (dump_context::begin_scope): Explicitly using the global function
3834 "dump_loc", rather than the member function.
3836 2018-07-31 Martin Sebor <msebor@redhat.com>
3838 PR tree-optimization/86741
3839 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
3841 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
3843 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
3844 depend on whether prefetch instructions will be emitted or not.
3845 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
3846 will be emitted or not.
3847 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
3848 (TARGET_SETMEM_PFD): New macros.
3850 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3852 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
3853 (NULL_STMT_VEC_INFO): Delete.
3854 (stmt_vec_info::operator*): Likewise.
3855 (stmt_vec_info::operator gimple *): Likewise.
3856 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
3857 of NULL_STMT_VEC_INFO.
3858 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3859 (vect_reassociating_reduction_p): Likewise.
3860 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3861 (vectorizable_store): Likewise.
3862 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
3863 (vec_info::free_stmt_vec_infos): Likewise.
3865 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3867 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
3868 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
3870 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3872 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
3873 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3874 (vec_info::free_stmt_vec_info): New private member functions.
3875 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
3876 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
3877 * tree-parloops.c (gather_scalar_reductions): Remove calls to
3878 set_stmt_vec_info_vec and free_stmt_vec_infos.
3879 * tree-vect-loop.c (_loop_vec_info): Remove call to
3880 set_stmt_vec_info_vec.
3881 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
3882 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
3883 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
3884 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3885 (vec_info::free_stmt_vec_info): ...these new functions. Remove
3886 assignments in {vec_info::,}new_stmt_vec_info that are redundant
3887 with the clearing in the xcalloc.
3888 (stmt_vec_info_vec): Delete.
3889 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
3890 (vectorize_loops): Likewise.
3891 (vec_info::~vec_info): Remove argument from call to
3892 free_stmt_vec_infos.
3893 (vec_info::add_stmt): Remove vinfo argument from call to
3896 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3898 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
3899 rather than a gimple stmt.
3900 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
3901 information for pattern statements when passed the original
3902 statement; instead wait to be passed the pattern statement itself.
3903 Don't call set_vinfo_for_stmt here.
3904 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
3905 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
3906 stmt_vec_infos here.
3907 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
3908 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
3909 stmt_vec_infos entry.
3911 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3913 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
3914 * tree-vectorizer.c (vec_info::replace_stmt): New function.
3915 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
3916 * tree-vect-stmts.c (vectorizable_call): Likewise.
3917 (vectorizable_simd_clone_call): Likewise.
3919 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3921 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
3922 * tree-vectorizer.c (vec_info::remove_stmt): New function.
3923 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
3924 * tree-vect-loop.c (vect_transform_loop): Likewise.
3925 * tree-vect-slp.c (vect_schedule_slp): Likewise.
3926 * tree-vect-stmts.c (vect_remove_stores): Likewise.
3928 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3930 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
3931 (vect_dr_stmt): Delete.
3932 * tree-vectorizer.c (vec_info::lookup_dr): New function.
3933 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
3935 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
3936 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
3937 (vect_verify_datarefs_alignment, vect_peeling_supportable)
3938 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3939 (vect_analyze_data_refs): Likewise.
3940 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
3942 (vect_find_same_alignment_drs): Likewise.
3943 (vect_slp_analyze_node_dependences): Update calls accordingly.
3944 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3945 instead of DR_VECT_AUX.
3946 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
3947 of a vector data references. Use vec_info::lookup_dr instead of
3949 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
3950 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3951 instead of DR_VECT_AUX.
3953 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3955 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
3957 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3959 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3960 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3961 (vect_gen_prolog_loop_niters): Likewise.
3963 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3965 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
3966 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
3967 (vect_known_alignment_in_bytes, vect_dr_behavior)
3968 (vect_get_scalar_dr_size): Take references as dr_vec_infos
3969 instead of data_references. Update calls to other routines for
3970 which the same change has been made.
3971 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
3972 dr_vec_infos instead of stmt_vec_infos.
3973 (vect_analyze_data_ref_dependence): Update call accordingly.
3974 (vect_slp_analyze_data_ref_dependence)
3975 (vect_record_base_alignments): Use DR_VECT_AUX.
3976 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
3977 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
3978 (vector_alignment_reachable_p, vect_get_data_access_cost)
3979 (vect_peeling_supportable, vect_analyze_group_access_1)
3980 (vect_analyze_group_access, vect_analyze_data_ref_access)
3981 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
3982 (vect_compile_time_alias, vect_small_gap_p)
3983 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
3984 (vect_supportable_dr_alignment): Take references as dr_vec_infos
3985 instead of data_references. Update calls to other routines for
3986 which the same change has been made.
3987 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
3988 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3989 (vect_slp_analyze_and_verify_node_alignment)
3990 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3991 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3992 (vect_setup_realignment): Use dr_vec_infos. Update calls after
3994 (_vect_peel_info::dr): Replace with...
3995 (_vect_peel_info::dr_info): ...this new field.
3996 (vect_peeling_hash_get_most_frequent)
3997 (vect_peeling_hash_choose_best_peeling): Update accordingly.
3998 (vect_peeling_hash_get_lowest_cost):
3999 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4000 routines for which the same change has been made.
4001 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4003 * tree-vect-loop-manip.c (get_misalign_in_elems)
4004 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4006 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4007 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4008 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4009 (get_group_load_store_type, get_negative_load_store_type)
4010 (vect_get_data_ptr_increment, vectorizable_store)
4011 (vectorizable_load): Likewise.
4012 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4013 Update calls to other routines for which the same change has been made.
4015 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4017 * tree-vectorizer.h (vec_info::move_dr): New member function.
4018 (dataref_aux): Rename to...
4019 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4020 (_stmt_vec_info::dr_aux): Update accordingly.
4021 (_stmt_vec_info::data_ref_info): Delete.
4022 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4023 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4024 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4025 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4027 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4028 (STMT_VINFO_DR_INFO): New macro.
4029 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4030 (set_dr_misalignment): Update after rename of dataref_aux.
4031 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4032 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4033 initialization of STMT_VINFO_DATA_REF.
4034 * tree-vectorizer.c (vec_info::move_dr): New function.
4035 * tree-vect-patterns.c (vect_recog_bool_pattern)
4036 (vect_recog_mask_conversion_pattern)
4037 (vect_recog_gather_scatter_pattern): Use it.
4038 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4039 the "dr" and "stmt" fields of dr_vec_info instead of
4040 STMT_VINFO_DATA_REF.
4042 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4044 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4045 (is_pattern_stmt_p): Use it.
4046 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4047 on pattern statements.
4049 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4051 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4052 original stmt as a stmt_vec_info rather than a gimple stmt.
4053 (vect_pattern_recog_1): Take the statement directly as a
4054 stmt_vec_info, rather than via a gimple_stmt_iterator.
4055 Update call to vect_mark_pattern_stmts.
4056 (vect_pattern_recog): Update calls accordingly.
4058 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4060 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4061 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4062 a vect_def_type for the first argument.
4063 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4064 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4065 operand if it isn't defined by a vectorized statement.
4066 (vect_build_gather_load_calls): Remove the mask_dt argument and
4067 update calls to vect_get_vec_def_for_stmt_copy.
4068 (vectorizable_bswap): Likewise the dt argument.
4069 (vectorizable_call): Update calls to vectorizable_bswap and
4070 vect_get_vec_def_for_stmt_copy.
4071 (vectorizable_simd_clone_call, vectorizable_assignment)
4072 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4073 (vectorizable_comparison): Update calls to
4074 vect_get_vec_def_for_stmt_copy.
4075 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4077 (vect_get_loop_based_defs): Remove dt argument and update call
4078 to vect_get_vec_def_for_stmt_copy.
4079 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4080 and vect_get_vec_def_for_stmt_copy.
4081 (vectorizable_load): Update calls to vect_build_gather_load_calls
4082 and vect_get_vec_def_for_stmt_copy.
4083 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4084 (vectorizable_reduction, vectorizable_live_operation): Update calls
4085 to vect_get_vec_def_for_stmt_copy.
4087 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4089 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4090 and gimple stmt arguments with a stmt_vec_info.
4091 (vect_record_base_alignments): Update calls accordingly.
4092 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4093 and gimple stmt arguments with a stmt_vec_info.
4094 (vect_build_slp_tree_1): Remove vinfo argument and update call
4095 to vect_record_max_nunits.
4096 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4097 and vect_record_max_nunits.
4099 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4101 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4102 file and take a stmt_vec_info instead of a gimple stmt.
4103 (supportable_widening_operation, vect_finish_replace_stmt)
4104 (vect_finish_stmt_generation, vect_get_store_rhs)
4105 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4106 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4107 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4108 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4109 (vect_create_data_ref_ptr, bump_vector_ptr)
4110 (vect_permute_store_chain, vect_setup_realignment)
4111 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4112 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4113 (vectorizable_reduction, vectorizable_induction)
4114 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4115 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4116 than gimple stmts as arguments.
4117 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4118 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4119 (can_group_stmts_p, vect_check_gather_scatter)
4120 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4121 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4122 (vect_permute_load_chain, vect_shift_permute_load_chain)
4123 (vect_transform_grouped_load)
4124 (vect_record_grouped_load_vectors): Likewise.
4125 * tree-vect-loop.c (vect_fixup_reduc_chain)
4126 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4127 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4128 (vectorizable_reduction, vectorizable_induction)
4129 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4130 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4131 (vect_get_load_store_mask): Likewise.
4132 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4133 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4134 * tree-vect-stmts.c (vect_mark_relevant)
4135 (is_simple_and_all_uses_invariant)
4136 (exist_non_indexing_operands_for_use_p, process_use)
4137 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4138 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4139 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4140 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4141 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4142 (get_negative_load_store_type, get_load_store_type)
4143 (vect_check_load_store_mask, vect_check_store_rhs)
4144 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4145 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4146 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4147 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4148 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4149 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4150 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4151 (supportable_widening_operation): Likewise.
4153 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4155 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4156 a stmt_vec_info instead of a gcall.
4157 (vect_check_gather_scatter): Update call accordingly.
4158 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4160 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4161 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4162 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4163 instead of a gimple stmt.
4164 (vect_transform_loop): Update calls accordingly.
4165 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4166 stmt_vec_infos instead of gimple stmts.
4167 (vect_analyze_slp_instance): Update use accordingly.
4168 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4169 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4170 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4171 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4172 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4173 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4174 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4175 down to subroutines.
4177 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4179 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4180 of the worklist from a vector of gimple stmts to a vector of
4182 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4183 (vect_mark_stmts_to_be_vectorized): Likewise
4185 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4187 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4188 statement before passing it to vect_analyze_stmt.
4189 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4190 the chain of phi vector definitions. Track the exit phi via its
4192 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4193 STMT_VINFO_REDUC_DEF.
4194 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4195 stmt_vec_infos to handle the statement chains.
4196 (vect_get_slp_defs): Record the first statement in the node
4197 using a stmt_vec_info.
4198 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4199 statements here and pass their stmt_vec_info down to subroutines.
4200 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4201 down to vect_finish_stmt_generation.
4202 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4203 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4204 stmt_vec_infos to subroutines.
4205 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4208 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4210 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4211 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4212 (vect_permute_store_chain, vect_setup_realignment)
4213 (vect_permute_load_chain, vect_shift_permute_load_chain)
4214 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4215 stmts internally, and when passing values to other vectorizer routines.
4216 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4217 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4218 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4219 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4220 (vectorizable_reduction, vectorizable_induction)
4221 (vectorizable_live_operation, vect_transform_loop_stmt)
4222 (vect_transform_loop): Likewise.
4223 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4224 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4225 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4226 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4227 (vect_slp_analyze_node_operations_1): Likewise.
4228 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4229 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4230 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4231 (vect_finish_stmt_generation_1, get_group_load_store_type)
4232 (get_load_store_type, vect_build_gather_load_calls)
4233 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4234 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4235 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4236 (vectorizable_store, vectorizable_load, vectorizable_condition)
4237 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4238 (supportable_widening_operation): Likewise.
4239 (vect_get_vector_types_for_stmt): Likewise.
4240 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4242 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4244 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4245 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4246 (vect_permute_store_chain, vect_permute_load_chain)
4247 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4248 repeated stmt_vec_info lookups.
4249 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4250 (vect_update_ivs_after_vectorizer): Likewise.
4251 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4252 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4253 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4254 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4255 (vect_bb_slp_scalar_cost): Likewise.
4256 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4258 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4260 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4261 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4262 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4263 to get gassigns and gcalls, rather than operating on generc gimple
4265 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4266 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4267 (vectorizable_load, vect_analyze_stmt): Likewise.
4268 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4270 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4272 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4273 return stmt_vec_infos rather than gimple stmts. Do not accept
4275 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4277 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4278 Update use of get_later_stmt.
4279 (vect_get_constant_vectors): Update call accordingly.
4280 (vect_schedule_slp_instance): Likewise
4281 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4282 (vect_slp_analyze_instance_dependence): Likewise.
4283 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4285 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4287 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4288 (stmt_info_for_cost::stmt_info): ...this new field.
4289 (add_stmt_costs): Update accordingly.
4290 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4291 (vect_get_known_peeling_cost): Likewise.
4292 (vect_estimate_min_profitable_iters): Likewise.
4293 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4295 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4297 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4298 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4299 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4301 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4303 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4305 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4306 a gimple stmt to a stmt_vec_info.
4307 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4309 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4311 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4312 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4313 (_loop_vec_info::reduction_chains): Likewise.
4314 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4316 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4318 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4320 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4321 a gimple stmt to a stmt_vec_info.
4322 (_stmt_vec_info::next_element): Likewise.
4323 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4324 (vect_slp_analyze_and_verify_node_alignment)
4325 (vect_analyze_group_access_1, vect_analyze_group_access)
4326 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4327 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4328 (vect_supportable_dr_alignment): Update accordingly.
4329 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4330 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4331 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4332 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4333 * tree-vect-slp.c (vect_build_slp_tree_1)
4334 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4335 (vect_split_slp_store_group, vect_analyze_slp_instance)
4336 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4337 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4338 (get_group_load_store_type, get_load_store_type)
4339 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4340 (vect_transform_stmt, vect_remove_stores): Likewise.
4342 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4344 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4346 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4347 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4348 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4349 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4350 (vector_alignment_reachable_p, vect_get_data_access_cost)
4351 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4352 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4353 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4354 (vect_analyze_group_access_1, vect_analyze_group_access)
4355 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4356 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4357 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4358 result of vect_dr_stmt and use the stmt_vec_info instead of
4359 the associated gimple stmt.
4360 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4361 (vect_gen_prolog_loop_niters): Likewise.
4362 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4364 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4366 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4367 to a vec<stmt_vec_info>.
4368 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4369 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4371 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4372 to a vec<stmt_vec_info>.
4373 (bst_traits::value_type, bst_traits::value_type): Likewise.
4374 (bst_traits::hash): Update accordingly.
4375 (vect_get_and_check_slp_defs): Change the stmts parameter from
4376 a vec<gimple *> to a vec<stmt_vec_info>.
4377 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4378 (vect_build_slp_tree): Likewise.
4379 (vect_build_slp_tree_2): Likewise. Update uses of
4380 SLP_TREE_SCALAR_STMTS.
4381 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4382 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4383 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4384 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4385 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4386 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4387 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4388 (vect_get_constant_vectors, vect_get_slp_defs)
4389 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4390 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4391 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4392 instead of gimple stmts.
4393 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4394 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4395 (vect_slp_analyze_instance_dependence): Update uses of
4396 SLP_TREE_SCALAR_STMTS.
4397 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4398 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4399 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4400 (get_initial_defs_for_reduction): Likewise.
4401 (vect_create_epilog_for_reduction): Likewise.
4402 (vectorize_fold_left_reduction): Likewise.
4403 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4404 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4405 (can_vectorize_live_stmts): Likewise.
4407 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4409 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4410 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4411 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4413 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4415 (gather_scalar_reductions): Update after above interface changes.
4416 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4417 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4419 (vect_force_simple_reduction): Likewise.
4420 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4421 LOOP_VINFO_REDUCTIONS.
4422 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4424 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4426 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4427 a gimple stmt to a stmt_vec_info.
4428 * tree-vect-loop.c (vect_active_double_reduction_p)
4429 (vect_force_simple_reduction, vectorizable_reduction): Update
4432 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4434 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
4435 vec<gimple *> to a vec<stmt_vec_info>.
4436 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
4437 the reduction_phis argument from a vec<gimple *> to a
4439 (vectorizable_reduction): Likewise the phis local variable that
4440 is passed to vect_create_epilog_for_reduction. Update for new type
4441 of SLP_TREE_VEC_STMTS.
4442 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
4443 (vectorizable_live_operation): Likewise.
4444 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
4445 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
4447 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4449 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
4450 a gimple stmt to a stmt_vec_info.
4451 (vectorizable_condition, vectorizable_live_operation)
4452 (vectorizable_reduction, vectorizable_induction): Pass back the
4453 vectorized statement as a stmt_vec_info.
4454 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
4455 use of STMT_VINFO_VEC_STMT.
4456 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
4457 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
4458 as stmt_vec_infos rather than gimple stmts.
4459 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
4461 (vectorizable_live_operation): Likewise.
4462 (vectorizable_reduction, vectorizable_induction): Likewise,
4463 updating use of STMT_VINFO_VEC_STMT.
4464 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
4465 of STMT_VINFO_VEC_STMT.
4466 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
4467 (vectorizable_simd_clone_call, vectorizable_conversion)
4468 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4469 (vectorizable_store, vectorizable_load, vectorizable_condition)
4470 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
4471 from a gimple stmt to a stmt_vec_info.
4472 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
4473 pointer to a stmt_vec_info to the vectorizable_* routines.
4475 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4477 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
4478 a gimple stmt to a stmt_vec_info.
4479 (is_pattern_stmt_p): Update accordingly.
4480 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
4481 (vect_record_grouped_load_vectors): Likewise.
4482 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4483 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
4484 (vect_model_reduction_cost): Likewise.
4485 (vect_create_epilog_for_reduction): Likewise.
4486 (vectorizable_reduction, vectorizable_induction): Likewise.
4487 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4488 Return the stmt_vec_info for the pattern statement.
4489 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
4490 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
4491 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4492 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
4493 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4494 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
4495 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
4496 (free_stmt_vec_info, vect_is_simple_use): Likewise.
4498 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4500 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
4501 (vect_finish_stmt_generation): Likewise.
4502 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4503 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
4504 (vect_build_gather_load_calls): Use the return value of the above
4505 functions instead of a separate call to vinfo_for_stmt. Use narrow
4506 scopes for the input gimple stmt and wider scopes for the associated
4507 stmt_vec_info. Use vec_info::lookup_def when setting these
4508 stmt_vec_infos from an SSA_NAME definition.
4509 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4510 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4511 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4512 (vectorizable_store, vectorizable_load, vectorizable_condition)
4513 (vectorizable_comparison): Likewise.
4514 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
4515 (vectorizable_reduction): Likewise.
4517 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4519 * tree-vectorizer.h (vect_is_simple_use): Add an optional
4520 stmt_vec_info * parameter before the optional gimple **.
4521 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4522 (process_use, vect_get_vec_def_for_operand_1): Update callers.
4523 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
4524 * tree-vect-loop.c (vectorizable_reduction): Likewise.
4525 (vectorizable_live_operation): Likewise.
4526 * tree-vect-patterns.c (type_conversion_p): Likewise.
4527 (vect_look_through_possible_promotion): Likewise.
4528 (vect_recog_rotate_pattern): Likewise.
4529 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4531 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4533 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
4534 a typedef to a wrapper class.
4535 (NULL_STMT_VEC_INFO): New macro.
4536 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
4537 (stmt_vec_info::operator*): New function.
4538 (stmt_vec_info::operator gimple *): Likewise.
4539 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
4540 (add_stmt_costs): Likewise.
4541 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
4542 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4543 (vect_get_known_peeling_cost): Likewise.
4544 (vect_estimate_min_profitable_iters): Likewise.
4545 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4546 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
4547 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4548 (vectorizable_store, free_stmt_vec_infos): Likewise.
4549 (new_stmt_vec_info): Change return type of xcalloc to
4552 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4554 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
4555 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
4556 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
4557 a single_imm_use-based sequence.
4558 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
4560 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4562 * tree-vectorizer.h (vec_info::lookup_def): Declare.
4563 * tree-vectorizer.c (vec_info::lookup_def): New function.
4564 * tree-vect-patterns.c (vect_get_internal_def): Use it.
4565 (vect_widened_op_tree): Likewise.
4566 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4567 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
4568 (vectorizable_reduction): Likewise.
4569 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
4571 (vect_is_slp_reduction): Update calls accordingly. Use
4572 vec_info::lookup_def.
4573 (vect_is_simple_reduction): Likewise
4574 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
4576 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4578 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
4579 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
4580 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
4582 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
4583 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
4584 (vect_update_vf_for_slp, vect_analyze_loop_operations)
4585 (vect_is_slp_reduction, vectorizable_induction)
4586 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
4587 * tree-vect-patterns.c (vect_init_pattern_stmt):
4588 (vect_determine_min_output_precision_1, vect_determine_precisions)
4589 (vect_pattern_recog): Likewise.
4590 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
4591 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
4592 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
4593 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4594 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
4595 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
4596 info field from a loop to a loop_vec_info.
4598 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4600 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4601 (vec_info::add_stmt): Declare.
4602 * tree-vectorizer.c (vec_info::add_stmt): New function.
4603 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4604 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4605 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4606 (vectorizable_induction): Likewise.
4607 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4608 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4609 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4610 (vectorizable_load): Likewise.
4611 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4612 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4613 (vect_recog_gather_scatter_pattern): Likewise.
4614 (append_pattern_def_seq): Likewise. Remove a check that is
4615 performed by add_stmt itself.
4617 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4619 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4620 which make_ssa_name was called with new_stmt before new_stmt
4623 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4625 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4627 (vect_is_slp_reduction): ...here...
4628 (vect_is_simple_reduction): ...and here. Remove repetition of tests
4629 that are already known to be false.
4631 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4633 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4634 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
4635 STMT_VINFO_NUM_SLP_USES when it's true.
4636 (vect_free_slp_instance): Add a final_p parameter and pass it to
4638 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4639 (vect_analyze_slp_instance): Likewise.
4640 (vect_slp_analyze_operations): Likewise.
4641 (vect_slp_analyze_bb_1): Likewise.
4642 * tree-vectorizer.c (vec_info): Likewise.
4643 * tree-vect-loop.c (vect_transform_loop): Likewise.
4645 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4647 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4648 function is not called for second and subsequent members of
4651 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4653 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4654 cases for nested loops from here to ...
4655 (vect_create_epilog_for_reduction): ...here. Only call
4656 vect_is_simple_use for inner-loop reductions.
4658 2018-07-31 Martin Liska <mliska@suse.cz>
4660 PR gcov-profile/85338
4661 PR gcov-profile/85350
4662 PR gcov-profile/85372
4663 * profile.c (struct location_triplet): New.
4664 (struct location_triplet_hash): Likewise.
4665 (output_location): Do not output a BB that
4666 is already recorded for a line.
4667 (branch_prob): Use streamed_locations.
4669 2018-07-31 Martin Liska <mliska@suse.cz>
4671 PR gcov-profile/85370
4672 * coverage.c (coverage_begin_function): Do not mark target
4673 clones as artificial functions.
4675 2018-07-31 Martin Liska <mliska@suse.cz>
4677 PR gcov-profile/83813
4678 PR gcov-profile/84758
4679 PR gcov-profile/85217
4680 PR gcov-profile/85332
4681 * profile.c (branch_prob): Do not record GOTO expressions
4682 for GIMPLE statements which locations are already streamed.
4684 2018-07-31 Olivier Hainque <hainque@adacore.com>
4686 * gcc.c (handle_spec_function): Accept a soft_matched_part
4687 argument, as do_spec_1. Pass it down to ...
4688 (eval_spec_function): Accept a soft_matched_part argument,
4689 and pass it down to ...
4690 (do_spec_2): Accept a soft_matched_part argument, and pass
4691 it down to do_spec_1.
4692 (do_spec_1): Pass soft_matched_part to handle_spec_function.
4693 (handle_braces): Update call to handle_spec_function.
4694 (driver::set_up_specs): Update calls to do_spec_2.
4695 (compare_debug_dump_opt_spec_function): Likewise.
4696 (compare_debug_self_opt_spec_function): Likewise.
4698 2018-07-31 Olivier Hainque <hainque@adacore.com>
4700 * common.opt (nolibc): New option.
4701 * doc/invoke.texi (Link Options): Document it.
4702 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4703 * config/alpha/linux.h: Likewise.
4704 * config/arc/elf.h: Likewise.
4705 * config/arm/uclinux-elf.h: Likewise.
4706 * config/arm/unknown-elf.h: Likewise.
4707 * config/avr/avrlibc.h: Likewise.
4708 * config/bfin/bfin.h: Likewise.
4709 * config/bfin/linux.h: Likewise.
4710 * config/bfin/uclinux.h: Likewise.
4711 * config/darwin.h: Likewise.
4712 * config/darwin10.h: Likewise.
4713 * config/darwin12.h: Likewise.
4714 * config/gnu-user.h: Likewise.
4715 * config/lm32/uclinux-elf.h: Likewise.
4716 * config/pa/pa-hpux11.h: Likewise.
4717 * config/pa/pa64-hpux.h: Likewise.
4718 * config/sparc/sparc.h: Likewise.
4720 2018-07-31 Olivier Hainque <hainque@adacore.com>
4722 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4723 undefined variables.
4725 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4728 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4731 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4733 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4734 terminated string literal.
4736 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4738 PR rtl-optimization/85160
4739 * combine.c (is_just_move): New function.
4740 (try_combine): Allow combining two instructions into two if neither of
4741 the original instructions was a move.
4743 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
4746 * doc/extend.texi (Global Register Variables): Discourage use of type
4748 (Local Register Variables): Likewise.
4750 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
4752 PR tree-optimization/86506
4753 * hwint.h (ceil_log2): Resync with hwint.c implementation.
4755 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
4758 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
4759 hard_regno, make sure no insn between `from` and `to` clobbers it.
4761 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
4762 Tom de Vries <tdevries@suse.de>
4764 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
4765 (PTX_DEFAULT_RUNTIME_DIM): ... this.
4766 (nvptx_goacc_validate_dims): Set default worker and gang dims to
4767 PTX_DEFAULT_RUNTIME_DIM.
4768 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
4770 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
4772 * config/pa/pa.c (pa_output_addr_vec): Align address table.
4773 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
4774 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
4776 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
4778 * config/rs6000/constraints.md (wG constraint): Delete, no longer
4780 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
4781 predicate to reflect toc fusion has been deleted.
4782 (toc_fusion_mem_raw): Delete, no longer used.
4783 (toc_fusion_mem_wrapped): Likewise.
4784 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
4786 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
4787 Delete, no longer used.
4788 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
4789 meant to be used for toc fusion.
4790 (rs6000_debug_print_mode): Delete toc fusion debugging.
4791 (rs6000_debug_reg_global): Likewise.
4792 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
4793 fusion and secondary reload support that were never used.
4794 (rs6000_option_override_internal): Delete TOC fusion, that was only
4795 partially defined, and it did not work unless you also used the
4797 (rs6000_legitimate_address_p): Delete TOC fusion support.
4798 (rs6000_opt_masks): Likewise.
4799 (fusion_wrap_memory_address): Delete function, no longer used.
4800 (fusion_split_address); Delete TOC fusion support.
4801 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
4802 longer used with toc fusion being deleted.
4803 (TARGET_TOC_FUSION_FP): Likewise.
4804 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
4806 (toc fusion spliter): Delete TOC fusion support.
4807 (toc_fusionload_<mode>): Likewise.
4808 (toc_fusionload_di): Likewise.
4809 (fusion_gpr_load_<mode>): Delete generator function, this insn no
4810 longer needs to be named. Rename predicate to delete TOC fusion.
4811 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4812 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4813 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4814 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4815 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
4817 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
4819 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
4820 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
4821 __int128 in built-in function prototypes.
4822 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
4823 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
4825 2018-07-27 Martin Sebor <msebor@redhat.com>
4827 PR tree-optimization/86696
4828 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
4829 types, including enums.
4830 (handle_char_store): Be prepared for the above function to fail.
4832 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
4834 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
4835 when optimization level is lower than 2 or optimize for size.
4837 2018-07-26 Martin Sebor <msebor@redhat.com>
4839 PR tree-optimization/86043
4840 PR tree-optimization/86042
4841 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
4842 (get_string_cst_length): Rename...
4843 (get_min_string_length): ...to this. Add argument.
4844 (handle_char_store): Extend to handle multi-character stores by
4846 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
4847 * tree.h (initializer_zerop): Add argument.
4849 2018-07-26 Jakub Jelinek <jakub@redhat.com>
4852 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
4853 declare target to variables if they have always,{to,from,tofrom} map
4856 2018-07-26 Martin Liska <mliska@suse.cz>
4859 * lto-wrapper.c: Add linker_output as prefix
4860 for ltrans_output_file.
4862 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
4864 PR rtl-optimization/85805
4865 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
4866 value for hard registers if that was written in the same mode.
4868 2018-07-26 Martin Liska <mliska@suse.cz>
4870 PR gcov-profile/86536
4871 * gcov.c (format_gcov): Use printf format %.*f directly
4872 and do not handle special values.
4874 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4876 * common/config/arc/arc-common.c (arc_option_optimization_table):
4877 Update default optimizations for size.
4879 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4881 * config/arc/arc.md (movsf_insn): Add short instruction selection.
4882 * config/arc/constraints.md (CfZ): New constraint.
4883 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
4884 (subsf3_fpu): Likewise.
4885 (cmpsf_fpu): Likewise.
4886 (cmpsf_fpu_uneq): Likewise.
4888 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4890 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
4892 (arc_is_uncached_mem_p): uncached applies to both the variable and
4895 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4897 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
4900 2018-07-25 David Malcolm <dmalcolm@redhat.com>
4902 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
4903 field "m_scopes" from vec to auto_vec.
4905 2018-07-25 Martin Liska <mliska@suse.cz>
4907 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
4910 2018-07-25 Richard Biener <rguenther@suse.de>
4913 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
4914 special wrt context_die late.
4915 (gen_subprogram_die): Re-use DIEs in local scope.
4917 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
4919 PR tree-optimization/86644
4920 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
4922 2018-07-25 Martin Liska <mliska@suse.cz>
4925 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
4926 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
4928 2018-07-25 Martin Liska <mliska@suse.cz>
4931 * params.def: Explain ASan abbreviation and provide
4932 a documentation link.
4934 2018-07-24 Martin Sebor <msebor@redhat.com>
4936 PR tree-optimization/86622
4937 PR tree-optimization/86532
4938 * builtins.h (string_length): Declare.
4939 * builtins.c (c_strlen): Correct handling of non-constant offsets.
4940 (check_access): Be prepared for non-constant length ranges.
4941 (string_length): Make extern.
4942 * expr.c (string_constant): Only handle the minor non-constant
4943 array index. Use string_constant to compute the length of
4944 a generic string constant.
4946 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
4948 PR tree-optimization/86618
4949 * tree-vect-stmts.c (vectorizable_call): Don't take the address
4950 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
4952 2018-07-24 David Malcolm <dmalcolm@redhat.com>
4954 PR tree-optimization/86636
4955 * json.cc (json::object::set): Fix comment. Add assertions.
4956 (json::array::append): Move here from json.h. Add comment and an
4958 (json::string::string): Likewise.
4959 * json.h (json::array::append): Move to json.cc.
4960 (json::string::string): Likewise.
4961 * optinfo-emit-json.cc
4962 (optrecord_json_writer::impl_location_to_json): Assert that we
4963 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
4964 wrapper around it. Expand the location once, rather than three
4966 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
4967 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
4969 (optrecord_json_writer::optinfo_to_json): Likewise, in four
4970 places. Fix some overlong lines.
4972 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
4974 * config/aarch64/aarch64-simd.md
4975 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
4976 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
4977 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
4978 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
4979 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
4980 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
4981 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
4982 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
4983 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
4985 2018-07-24 Jakub Jelinek <jakub@redhat.com>
4988 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
4989 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
4990 and abs_d == d, do the power of two handling if profitable.
4992 2018-07-24 Richard Biener <rguenther@suse.de>
4994 * match.pd: Add BIT_FIELD_REF canonicalizations.
4996 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4999 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5001 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5003 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5004 terminated STRING_CST object.
5006 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5008 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5010 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5012 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5013 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5014 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5015 the elements into a register.
5016 (rs6000_split_v4si_init_di_reg): Delete.
5017 (rs6000_split_v4si_init): Delete.
5018 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5019 (vsx_init_v4si): Rewrite as a define_expand.
5021 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5023 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5024 zero_extend argument from memory): New.
5026 2018-07-22 Martin Sebor <msebor@redhat.com>
5029 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5030 diagnosing calls with unknown arguments unless -Walloca-larger-than
5031 is restricted to less than PTRDIFF_MAX bytes.
5033 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5035 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5037 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5039 * pretty-print.c (text_info::set_location): Remove redundant
5040 "line_table" parameter from call to rich_location::set_range.
5042 2018-07-20 Martin Sebor <msebor@redhat.com>
5045 * builtins.c (expand_builtin_alloca): Adjust.
5046 * calls.c (alloc_max_size): Simplify.
5047 * cgraphunit.c (cgraph_node::expand): Adjust.
5048 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5050 (frame_larger_than_size): Same.
5051 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5052 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5054 * doc/invoke.texi (GCC Command Options): Document option arguments.
5055 Explain byte-size arguments and suffixes.
5056 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5057 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5058 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5059 * doc/options.texi (UInteger): Expand.
5060 (Host_Wide_Int, ByteSize): Document new properties.
5061 * final.c (final_start_function_1): Include sizes in an error message.
5062 * function.c (frame_offset_overflow): Same.
5063 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5064 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5065 Diagnose unbounded alloca calls only for limits of less than
5067 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5068 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5070 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5071 only for limits of less than PTRDIFF_MAX.
5072 * langhooks-def.h (lhd_handle_option): Change function argument
5074 * langhooks.c (lhd_handle_option): Same.
5075 * langhooks.h (handle_option): Same.
5076 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5078 (var_type, var_type_struct): Same.
5079 (var_set): Handle ByteSize flag.
5080 * optc-gen.awk: Add comments to output to ease debugging. Make
5081 use of HOST_WIDE_INT where appropriate.
5082 * opts-gen-save.awk: Use %lx to format unsigned long.
5083 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5084 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5085 arguments. Parse bytes-size suffixes.
5086 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5087 (enum_value_to_arg): Same.
5088 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5089 (handle_option): Adjust.
5090 (generate_option): Change function argument to HOST_WIDE_INT.
5091 (cmdline_handle_error): Adjust.
5092 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5093 (set_option): Change function argument to HOST_WIDE_INT.
5094 (option_enabled): Handle cl_host_wide_int.
5095 (get_option_state): Handle CLVC_SIZE.
5096 (control_warning_option): Same.
5097 * opts.c (common_handle_option): Change function argument to
5098 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5099 OPT_Wvla_larger_than_.
5100 * opts.h (enum cl_var_type): Add an enumerator.
5101 * stor-layout.c (layout_decl): Print a more meaningful warning.
5102 * toplev.c (output_stack_usage): Adjust.
5104 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5106 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5107 call to inline_expand_builtin_string_cmp.
5108 (expand_builtin_strcmp): Likewise.
5109 (expand_builtin_strncmp): Likewise.
5110 (inline_string_cmp): Delete the last parameter, change char_type_node
5111 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5113 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5114 the inlining expansion on target where the type of the call has same or
5115 narrower precision than unsigned char.
5117 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5119 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5120 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5121 * common.opt (fsave-optimization-record): New option.
5122 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5123 * doc/invoke.texi (-fsave-optimization-record): New option.
5124 * dumpfile.c: Include "optinfo-emit-json.h".
5125 (struct kv_pair): Move to coretypes.h.
5126 (optgroup_options): Make non-static.
5127 (dump_context::end_scope): Call
5128 optimization_records_maybe_pop_dump_scope.
5129 * dumpfile.h (optgroup_options): New decl.
5130 * json.cc: New file.
5132 * optinfo-emit-json.cc: New file.
5133 * optinfo-emit-json.h: New file.
5134 * optinfo.cc: Include "optinfo-emit-json.h".
5135 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5136 (optinfo_enabled_p): Check optimization_records_enabled_p.
5137 (optinfo_wants_inlining_info_p): Likewise.
5138 * optinfo.h: Update comment.
5139 * profile-count.c (profile_quality_as_string): New function.
5140 * profile-count.h (profile_quality_as_string): New decl.
5141 (profile_count::quality): New accessor.
5142 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5143 and optinfo_emit_json_cc_tests.
5144 * selftest.h (selftest::json_cc_tests): New decl.
5145 (selftest::optinfo_emit_json_cc_tests): New decl.
5146 * toplev.c: Include "optinfo-emit-json.h".
5147 (compile_file): Call optimization_records_finish.
5148 (do_compile): Call optimization_records_start.
5149 * tree-ssa-live.c: Include optinfo.h.
5150 (remove_unused_scope_block_p): Retain inlining information if
5151 optinfo_wants_inlining_info_p returns true.
5153 2018-07-20 Richard Biener <rguenther@suse.de>
5156 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5157 to cover -flto-partition=none.
5159 2018-07-20 Martin Liska <mliska@suse.cz>
5161 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5162 (get_decl_source_range): Remove unused function.
5164 2018-07-20 Richard Biener <rguenther@suse.de>
5166 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5167 (struct vn_phi_s): Likewise.
5168 (struct vn_reference_s): Likewise.
5169 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5170 for searching the slot of an entry known to be in the hash itself.
5171 (vn_phi_hasher::equal): Likewise.
5172 (vn_reference_hasher::equal): Likewise.
5173 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5175 (optimistic_info, current_info): Remove, keeping only valid_info.
5176 (vn_reference_lookup_1): Remove fallback lookup.
5177 (vn_reference_lookup_2): Likewise.
5178 (vn_nary_op_lookup_1): Likewise.
5179 (vn_phi_lookup): Likewise.
5180 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5182 (vn_reference_insert): Adjust, chain the inserted hash element
5183 at last_inserted_ref.
5184 (vn_reference_insert_pieces): Likewise.
5185 (visit_reference_op_call): Likewise.
5186 (vn_nary_op_insert_into): Chain the inserted hash element at
5188 (vn_nary_op_insert_pieces): Adjust.
5189 (vn_nary_op_insert): Likewise.
5190 (vn_nary_op_insert_stmt): Likewise.
5191 (vn_phi_insert): Adjust, chain the inserted hash element at
5193 (process_scc): Remove clearing and copying the optimistic
5194 table. Instead remove elements inserted during an optimistic
5195 iteration from the single table we maintain.
5196 (init_scc_vn): Adjust.
5197 (free_scc_vn): Likewise.
5198 (sccvn_dom_walker::record_cond): Likewise.
5199 (sccvn_dom_walker::after_dom_children): Likewise.
5201 2018-07-19 Martin Sebor <msebor@redhat.com>
5203 PR tree-optimization/84047
5204 PR tree-optimization/83776
5205 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5206 (check_array_bounds): Call it.
5208 2018-07-19 Martin Sebor <msebor@redhat.com>
5210 * align.h (align_flags): Use member initialization.
5212 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5214 * Makefile.in (OBJS): Add optinfo.o.
5215 * coretypes.h (class symtab_node): New forward decl.
5216 (struct cgraph_node): New forward decl.
5217 (class varpool_node): New forward decl.
5218 * dump-context.h: New file.
5219 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5221 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5222 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5223 (set_alt_dump_file): Likewise.
5224 (dump_context::~dump_context): New dtor.
5225 (dump_gimple_stmt): Move implementation to...
5226 (dump_context::dump_gimple_stmt): ...this new member function.
5227 Add the stmt to any pending optinfo, creating one if need be.
5228 (dump_gimple_stmt_loc): Move implementation to...
5229 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5230 Start a new optinfo and add the stmt to it.
5231 (dump_gimple_expr): Move implementation to...
5232 (dump_context::dump_gimple_expr): ...this new member function.
5233 Add the stmt to any pending optinfo, creating one if need be.
5234 (dump_gimple_expr_loc): Move implementation to...
5235 (dump_context::dump_gimple_expr_loc): ...this new member function.
5236 Start a new optinfo and add the stmt to it.
5237 (dump_generic_expr): Move implementation to...
5238 (dump_context::dump_generic_expr): ...this new member function.
5239 Add the tree to any pending optinfo, creating one if need be.
5240 (dump_generic_expr_loc): Move implementation to...
5241 (dump_context::dump_generic_expr_loc): ...this new member
5242 function. Add the tree to any pending optinfo, creating one if
5244 (dump_printf): Move implementation to...
5245 (dump_context::dump_printf_va): ...this new member function. Add
5246 the text to any pending optinfo, creating one if need be.
5247 (dump_printf_loc): Move implementation to...
5248 (dump_context::dump_printf_loc_va): ...this new member function.
5249 Start a new optinfo and add the stmt to it.
5250 (dump_dec): Move implementation to...
5251 (dump_context::dump_dec): ...this new member function. Add the
5252 value to any pending optinfo, creating one if need be.
5253 (dump_context::dump_symtab_node): New member function.
5254 (dump_context::get_scope_depth): New member function.
5255 (dump_context::begin_scope): New member function.
5256 (dump_context::end_scope): New member function.
5257 (dump_context::ensure_pending_optinfo): New member function.
5258 (dump_context::begin_next_optinfo): New member function.
5259 (dump_context::end_any_optinfo): New member function.
5260 (dump_context::s_current): New global.
5261 (dump_context::s_default): New global.
5262 (dump_scope_depth): Delete global.
5263 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5264 (dump_symtab_node): New function.
5265 (get_dump_scope_depth): Reimplement in terms of dump_context.
5266 (dump_begin_scope): Likewise.
5267 (dump_end_scope): Likewise.
5268 (selftest::temp_dump_context::temp_dump_context): New ctor.
5269 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5270 (selftest::verify_item): New function.
5271 (ASSERT_IS_TEXT): New macro.
5272 (ASSERT_IS_TREE): New macro.
5273 (ASSERT_IS_GIMPLE): New macro.
5274 (selftest::test_capture_of_dump_calls): New test.
5275 (selftest::dumpfile_c_tests): Call it.
5276 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5277 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5278 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5279 descriptive comment.
5280 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5281 (dump_node, dump_bb): Move these unrelated decls.
5282 (class dump_manager): Add leading comment.
5283 * optinfo.cc: New file.
5284 * optinfo.h: New file.
5286 2018-07-19 Michael Collison <michael.collison@arm.com>
5287 Richard Henderson <rth@redhat.com>
5289 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5290 (subti): Handle op1 zero.
5291 (subvti4, usub4ti4): New.
5292 (*sub<GPI>3_compare1_imm): New.
5293 (sub<GPI>3_carryinCV): New.
5294 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5295 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5297 2018-07-19 Michael Collison <michael.collison@arm.com>
5298 Richard Henderson <rth@redhat.com>
5300 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5301 (addti3): Create simpler code if low part is already known to be 0.
5302 (addvti4, uaddvti4): New.
5303 (*add<GPI>3_compareC_cconly_imm): New.
5304 (*add<GPI>3_compareC_cconly): New.
5305 (*add<GPI>3_compareC_imm): New.
5306 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5307 handle constants within this pattern..
5308 (*add<GPI>3_compareV_cconly_imm): New.
5309 (*add<GPI>3_compareV_cconly): New.
5310 (*add<GPI>3_compareV_imm): New.
5311 (add<GPI>3_compareV): New.
5312 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5313 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5314 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5315 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5316 with 'comparison' operator.
5317 (*add<GPI>3_compareV_cconly_imm): Ditto.
5318 (*add<GPI>3_compareV_cconly): Ditto.
5319 (*add<GPI>3_compareV_imm): Ditto.
5320 (add<GPI>3_compareV): Ditto.
5321 (add<mode>3_carryinC): Ditto.
5322 (*add<mode>3_carryinC_zero): Ditto.
5323 (*add<mode>3_carryinC): Ditto.
5324 (add<mode>3_carryinV): Ditto.
5325 (*add<mode>3_carryinV_zero): Ditto.
5326 (*add<mode>3_carryinV): Ditto.
5328 2018-07-19 Michael Collison <michael.collison@arm.com>
5329 Richard Henderson <rth@redhat.com>
5331 * config/aarch64/aarch64-modes.def (CC_V): New.
5332 * config/aarch64/aarch64-protos.h
5333 (aarch64_addti_scratch_regs): Declare
5334 (aarch64_subvti_scratch_regs): Declare.
5335 (aarch64_expand_subvti): Declare.
5336 (aarch64_gen_unlikely_cbranch): Declare
5337 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5338 for signed overflow using CC_Vmode.
5339 (aarch64_get_condition_code_1): Handle CC_Vmode.
5340 (aarch64_gen_unlikely_cbranch): New function.
5341 (aarch64_addti_scratch_regs): New function.
5342 (aarch64_subvti_scratch_regs): New function.
5343 (aarch64_expand_subvti): New function.
5345 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5347 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5349 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5350 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5353 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5356 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5357 address check not strict.
5359 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5361 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5363 (store_pair_lanes<mode>): Likewise.
5364 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5365 enum value 'ADDR_QUERY_LDP_STP_N'.
5366 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5367 (aarch64_print_address_internal): Add declaration.
5368 (aarch64_print_ldpstp_address): Remove.
5369 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5370 (aarch64_print_operand): Change printing of 'y'.
5371 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5372 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5373 'true' rather than '1'.
5374 * config/aarch64/constraints.md (Uml): Likewise.
5375 (Uml): Rename to Umn.
5378 2018-07-19 Richard Biener <rguenther@suse.de>
5380 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5382 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5383 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5384 (vn_reference_hasher): Likewise.
5385 (struct vn_tables_s): Remove obstack and alloc-pool members.
5386 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5387 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5388 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5389 (vn_reference_insert_pieces): Likewise.
5390 (alloc_vn_nary_op_noinit): Adjust.
5391 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5392 (vn_phi_eq): Adjust.
5393 (shared_lookup_phiargs): Remove.
5394 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5395 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5396 (visit_reference_op_call): Likewise.
5397 (copy_nary, copy_phi, copy_reference): Remove.
5398 (process_scc): Rewind the obstack when iterating. Do not
5399 copy the elements to valid_info but just move them from one
5400 hashtable to the other.
5401 (allocate_vn_table): Adjust.
5402 (free_vn_table): Likewise.
5403 (init_scc_vn): Likewise.
5404 (free_scc_vn): Likewise.
5406 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5409 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5410 indirect_return as function type attribute.
5411 (ix86_attribute_table): Change indirect_return to function
5413 * doc/extend.texi: Update indirect_return attribute.
5415 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5417 * wide-int.h (widest2_int): New.
5418 * gimple-fold.c (arith_overflowed_p): Use it.
5419 * tree.h (widest2_int_cst): New.
5420 * tree-vrp.c (wide_int_binop_overflow): Rename from
5421 vrp_int_const_binop.
5422 Rewrite to work on trees.
5423 (extract_range_from_multiplicative_op_1): Abstract code to...
5424 (wide_int_range_min_max): ...here.
5425 (wide_int_range_cross_product): ...and here.
5426 (extract_range_from_binary_expr_1): Abstract overflow code to...
5427 (wide_int_range_mult_wrapping): ...here.
5428 * tree-vrp.h (wide_int_range_cross_product): New.
5429 (wide_int_range_mult_wrapping): New.
5431 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
5432 Julia Koval <julia.koval@intel.com>
5434 * config/i386/x86-tune-costs.h (skylake_memcpy,
5435 skylake_memset): Replace rep_prefix with unrolling for size 512.
5437 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
5440 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
5441 comparision with EQ_EXPR in last stmt.
5443 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
5445 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
5446 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
5447 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
5448 previously known as "PowerPC AltiVec Built-in Functions". Move
5449 some material to new subsubsections "PowerPC AltiVec Built-in
5450 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5452 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5453 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5454 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
5455 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
5457 2018-07-18 Richard Biener <rguenther@suse.de>
5459 PR tree-optimization/86557
5460 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
5463 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
5465 * config/s390/s390.c (s390_function_profiler): Generate CFI.
5467 2018-07-17 Jeff Law <law@redhat.com>
5469 * config/arm/arm.c (get_label_padding): Update for recent
5470 changes to label_to_alignment.
5472 PR tree-optimization/86010
5473 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
5475 * config/mips/mips.c (vr4130_align_insns): Update for recent
5476 changes to label_to_alignment.
5478 * config/frv/frv.c (frv_label_align): Update for recent changes
5479 to label_to_alignment.
5481 * config/nios2/nios2.c (nios2_label_align): Update for recent
5482 changes which dropped ALIGN_LABELS_LOG.
5484 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
5486 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
5487 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
5489 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
5491 * config/arc/arc.c (arc_label_align): Use align_labels instead of
5492 deprecated align_labels_log.
5494 2018-07-17 Richard Biener <rguenther@suse.de>
5497 * dwarf2out.c (init_sections_and_labels): Always generate
5498 a debug_line_str_section for early LTO debug.
5499 (dwarf2out_finish): Reset debug_line_str_hash output early.
5500 Bump counter for extra dwarf5 .debug_loc labels to not conflict
5502 (dwarf2out_early_finish): Output debug_line_str.
5504 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5506 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
5507 index register on z196 or later.
5509 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5511 * config/s390/s390.c (s390_default_align): Set default function
5513 (s390_override_options_after_change): Call s390_default align.
5514 (s390_option_override_internal): Call s390_default_align.
5515 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
5517 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5520 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
5521 to _looptemp_ clauses, other than the first two.
5523 2018-07-17 Martin Liska <mliska@suse.cz>
5525 * opts.c: Do not enable OPT_falign_* for -Os.
5527 2018-07-17 Martin Liska <mliska@suse.cz>
5529 * align.h (MAX_CODE_ALIGN): New.
5530 (MAX_CODE_ALIGN_VALUE): New.
5531 * common/config/i386/i386-common.c (ix86_handle_option):
5532 (MAX_CODE_ALIGN): Moved to align.h.
5533 * final.c (MAX_CODE_ALIGN): Likewise.
5534 * opts.c (parse_and_check_align_values):
5535 (MAX_CODE_ALIGN): Likewise.
5536 (MAX_CODE_ALIGN_VALUE): Likewise.
5538 2018-07-17 Martin Liska <mliska@suse.cz>
5540 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
5541 in order to fulfil coding style.
5542 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
5543 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5544 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5545 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
5546 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
5547 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
5548 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
5549 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
5550 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5552 2018-07-17 Martin Liska <mliska@suse.cz>
5554 * align.h: New file.
5555 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
5557 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
5558 align_flags of label_to_alignment.
5559 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
5561 * config/m68k/m68k.c: Do not use removed align_labels_value and
5563 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
5564 (LOOP_ALIGN): Likewise.
5565 (LABEL_ALIGN): Likewise.
5566 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5567 Remove not used macro.
5568 (rs6000_loop_align): Change return type to align_flags.
5569 (rs6000_loop_align_max_skip): Remove.
5570 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
5571 Change return type to align_flags.
5572 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5573 Remove not used macro.
5574 (rs6000_loop_align): Change return type to align_flags.
5575 (rs6000_loop_align_max_skip): Remove.
5576 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
5577 * config/rx/rx-protos.h (rx_align_for_label): Make it
5579 * config/rx/rx.c (rx_align_for_label): Change return type
5581 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5583 into align_flags class.
5584 (LABEL_ALIGN): Likewise.
5585 (LOOP_ALIGN): Likewise.
5586 * config/s390/s390.c (s390_label_align): Use align_flags
5588 (s390_asm_output_function_label): Likewise.
5589 * config/sh/sh.c (sh_override_options_after_change):
5590 Use align_flags class directly without macros.
5591 (find_barrier): Likewise.
5592 (barrier_align): Likewise.
5593 (sh_loop_align): Likewise.
5594 * config/spu/spu.c (spu_option_override):
5595 Use align_flags_tuple::get_value instead of removed macros.
5596 (spu_sched_init): Likewise.
5597 * config/spu/spu.h (GTY): Likewise.
5598 * config/visium/visium.c (visium_option_override):
5599 Set "8" as default secondary alignment.
5600 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5601 in order to guarantee secondary alignment of 8.
5602 * coretypes.h: Include align.h header file.
5603 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5604 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5605 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5606 * doc/tm.texi.in: Likewise.
5607 * final.c (struct label_alignment): Remove not used structure.
5608 (LABEL_ALIGN): Change type to align_flags.
5609 (LOOP_ALIGN): Likewise.
5610 (JUMP_ALIGN): Likewise.
5611 (default_loop_align_max_skip): Remove.
5612 (default_label_align_max_skip): Likewise.
5613 (default_jump_align_max_skip): Likewise.
5614 (default_label_align_after_barrier_max_skip):
5615 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5616 (LABEL_TO_MAX_SKIP): Remove.
5617 (label_to_alignment): Return align_flags type instead of integer.
5618 (label_to_max_skip): Remove.
5619 (align_fuzz): Use align_flags type.
5620 (compute_alignments): Use align_flags type and use align_flags::max
5621 to combine multiple alignments.
5622 (grow_label_align): Grow vec instead of C array.
5623 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5624 (shorten_branches): Use align_flags type and use align_flags::max
5625 to combine multiple alignments.
5626 (final_scan_insn_1): Remove usage of secondary alignment that comes
5627 from label alignment, but instead use proper secondary alignment
5628 which is computed in grow_label_align.
5629 * flags.h (struct align_flags_tuple): Move to align.h.
5630 (struct align_flags): Likewise.
5631 (state_align_loops): Rename to align_loops.
5632 (state_align_jumps): Rename to align_jumps.
5633 (state_align_labels): Rename to align_labels.
5634 (state_align_functions): Rename to align_functions.
5635 (align_loops_log): Remove.
5636 (align_jumps_log): Remove.
5637 (align_labels_log): Remove.
5638 (align_functions_log): Remove.
5639 (align_loops_max_skip): Remove.
5640 (align_jumps_max_skip): Remove.
5641 (align_labels_max_skip): Remove.
5642 (align_functions_max_skip): Remove.
5643 (align_loops_value): Remove.
5644 (align_jumps_value): Remove.
5645 (align_labels_value): Remove.
5646 (align_functions_value): Remove.
5647 * output.h (label_to_alignment): Change return type to align_flags.
5648 (label_to_max_skip): Remove.
5649 * target.def: Remove loop_align_max_skip, label_align_max_skip,
5650 jump_align_max_skip macros.
5651 * targhooks.h (default_loop_align_max_skip): Remove.
5652 (default_label_align_max_skip): Likewise.
5653 (default_jump_align_max_skip): Likewise.
5654 (default_label_align_after_barrier_max_skip): Remove.
5655 * toplev.c (read_log_maxskip): Use ::normalize function.
5656 (parse_N_M): Remove not used argument and also call ::normalize.
5657 (parse_alignment_opts): Do not pass unused arguments.
5658 * varasm.c (assemble_start_function): Use directly align_functions
5659 instead of removed macros.
5660 * system.h: Do not poison removed macros.
5662 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5665 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5666 and cond temporaries don't have reference type if iterator has
5667 pointer type. For init use &for_pre_body instead of pre_p if
5668 for_pre_body is non-empty.
5670 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5672 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5673 double-double modes to SFmode directly directly.
5674 (trunc<mode>sf2_fprs): Delete.
5676 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5678 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5679 for conversions between IFmode and the decimal floating point modes.
5680 (init_float128_ieee): Use the correct names for conversions between
5681 KFmode and the decimal floating point modes.
5683 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5685 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5686 for the conversions between TDmode and IFmode.
5687 (init_float128_ieee): Use more correct names for the conversions
5688 between TDmode and KFmode.
5690 2018-07-16 Jakub Jelinek <jakub@redhat.com>
5692 PR tree-optimization/86526
5693 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5694 (inline_expand_builtin_string_cmp): Likewise.
5695 (inline_string_cmp): Likewise. Use c_readstr instead of
5696 builtin_memcpy_read_str. Add unit_mode temporary.
5698 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5701 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5702 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5704 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5706 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5707 Alphabetize prototypes of built-in functions, separating out
5708 built-in functions that are listed in this section but should be
5709 described elsewhere.
5711 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
5714 * expmed.c (emit_store_flag): Do not emit setcc followed by a
5715 conditional move when trapping comparison was split to a
5716 non-trapping one (and vice versa).
5718 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5720 * config/s390/s390.c (s390_function_profiler): Generate nops
5721 instead of profiler call sequences.
5722 * config/s390/s390.opt: Add the new option.
5724 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5726 * config/s390/s390.c (s390_function_profiler): Generate
5727 __mcount_loc section.
5728 * config/s390/s390.opt: Add the new option.
5730 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5732 * common.opt: Add the new warning.
5733 * config/s390/s390.c (s390_function_profiler): Emit "brasl
5734 %r0,__fentry__" when -mfentry is specified.
5735 (s390_option_override_internal): Disallow -mfentry for 31-bit
5737 * config/s390/s390.opt: Add the new option.
5739 2018-07-16 Richard Biener <rguenther@suse.de>
5742 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
5743 for function-local FUNCTION_DECL and RESULT_DECL immediately.
5745 2018-07-16 Martin Liska <mliska@suse.cz>
5748 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
5751 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5753 * config/arc/arcHS.md: Update ARCHS scheduling rules.
5755 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5757 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5759 * config/arc/arc-cpus.def (hs4x): New cpu.
5761 * config/arc/arc-tables.opt: Regenerate.
5762 * config/arc/arc.c (arc_sched_issue_rate): New function.
5763 (TARGET_SCHED_ISSUE_RATE): Define.
5764 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
5765 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
5767 (attr tune): Add ARCHS4x tune values.
5768 (attr tune_dspmpy): Define.
5769 (*tst): Correct instruction type.
5770 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
5771 * config/arc/arcHS4x.md: New file.
5772 * config/arc/fpu.md: Update instruction type attributes.
5773 * config/arc/t-multilib: Regenerate.
5775 2018-07-16 Tom de Vries <tdevries@suse.de>
5778 * var-tracking.c (vt_initialize): Fix pre_dec handling.
5780 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5782 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
5785 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
5787 PR tree-optimization/86514
5788 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
5789 conversion to a boolean type from a type with greater precision.
5791 2018-07-16 Tom de Vries <tdevries@suse.de>
5793 * var-tracking.c (vt_initialize): Print adjusted insn slim if
5794 dump_flags request TDF_SLIM.
5796 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
5798 * fold-const.c (int_const_binop_1): Abstract...
5799 (wide_int_binop): ...wide int code here.
5800 (poly_int_binop): ...poly int code here.
5801 Abstract the rest of int_const_binop_1 into int_const_binop.
5802 * fold-const.h (wide_int_binop): New.
5803 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
5804 Remove useless PLUS/MINUS_EXPR case.
5805 (zero_nonzero_bits_from_vr): Move wide int code...
5806 (zero_nonzero_bits_from_bounds): ...here.
5807 (extract_range_from_binary_expr_1): Move mask optimization code...
5808 (range_easy_mask_min_max): ...here.
5809 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
5810 (range_easy_mask_min_max): New.
5812 2018-07-15 Jeff Law <law@redhat.com>
5815 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
5818 2018-07-14 Jim Wilson <jimw@sifive.com>
5820 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
5822 2018-07-14 Paul Koning <ni1d@arrl.net>
5824 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
5826 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
5828 * lto-streamer-out.c (copy_function_or_variable): Dump info about
5831 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
5832 Steve Munroe <munroesj52@gmail.com>
5834 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
5835 (_mm_andnot_si128): Likewise.
5836 (_mm_or_si128): Likewise.
5837 (_mm_xor_si128): Likewise.
5839 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
5842 * builtins.c (expand_builtin_memcmp): Inline the calls first
5843 when result_eq is false.
5844 (expand_builtin_strcmp): Inline the calls first.
5845 (expand_builtin_strncmp): Likewise.
5846 (inline_string_cmp): New routine. Expand a string compare
5847 call by using a sequence of char comparison.
5848 (inline_expand_builtin_string_cmp): New routine. Inline expansion
5849 a call to str(n)cmp/memcmp.
5850 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
5852 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
5854 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
5856 * config/arm/driver-arm.c: Include arm-native.h.
5857 (host_detect_local_cpu): Use auto-generated data tables.
5858 (vendors, arm_cpu_table): Delete. Move part information to ...
5859 * config/arm/arm-cpus.in: ... here.
5860 * config/arm/parsecpu.awk (gen_native): New function.
5861 (vendor, part): New CPU fields.
5862 (END): Add support for building the native CPU detection tables.
5863 * config/arm/t-arm (arm-native.h): Add build rule.
5864 (driver-arm.o): Add dependency on arm-native.h.
5866 2018-07-13 Richard Biener <rguenther@suse.de>
5869 * match.pd (addr1 - addr2): Allow either of the operand to
5872 2018-07-13 Tom de Vries <tdevries@suse.de>
5874 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
5877 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
5879 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
5880 arrays instead of numbered variables.
5882 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
5884 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
5885 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
5887 2018-07-13 Richard Biener <rguenther@suse.de>
5890 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
5891 instead of get_context_die.
5893 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
5894 Richard Biener <rguenther@suse.de>
5897 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
5898 that the loop latch destination where phi is defined.
5900 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
5902 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
5903 (riscv_expand_epilogue): Add assertion to check interrupt mode.
5904 (riscv_set_current_function): Extract getting interrupt type to new
5906 (riscv_get_interrupt_type): New function.
5907 (riscv_merge_decl_attributes): New function, checking interrupt type is
5909 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
5911 2018-07-12 Paul Koning <ni1d@arrl.net>
5913 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
5916 2018-07-12 Paul Koning <ni1d@arrl.net>
5918 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
5919 zero reference, add doloop_end instead.
5920 * doc/md.texi (decrement_and_branch_until_zero): Remove.
5921 (Looping patterns): Remove decrement_and_branch_until_zero. Add
5922 detail for doloop_end.
5924 2018-07-12 Martin Sebor <msebor@redhat.com>
5927 * attribs.c (decl_attributes): Reject conflicting attributes before
5928 calling attribute handlers.
5930 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5932 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
5934 (gcc::dump_manager::get_dump_file_name): likewise.
5935 (dump_begin): Likewise.
5936 * dumpfile.h (dump_begin): Update prototype.
5937 (gcc::dump_manager::get_dump_file_name,
5938 gcc::dump_manager::get_dump_file_name): Update prototype.
5940 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5942 * internal-fn.h (vectorizable_internal_fn_p): New function.
5943 * tree-vect-slp.c (compatible_calls_p): Likewise.
5944 (vect_build_slp_tree_1): Remove nops argument. Handle calls
5945 to internal functions.
5946 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
5948 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5950 * fold-const.h (inverse_conditions_p): Declare.
5951 * fold-const.c (inverse_conditions_p): New function.
5952 * match.pd: Use inverse_conditions_p. Add folds of view_converts
5953 that test the inverse condition of a conditional internal function.
5954 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
5955 * internal-fn.c (internal_fn_mask_index): Handle conditional
5957 (vectorized_internal_fn_supported_p): New function.
5958 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
5959 (any_pred_load_store): Replace with...
5960 (need_to_predicate): ...this new variable.
5961 (redundant_ssa_names): New variable.
5962 (ifcvt_can_use_mask_load_store): Move initial checks to...
5963 (ifcvt_can_predicate): ...this new function. Handle tree codes
5964 for which a conditional internal function exists.
5965 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
5966 instead of ifcvt_can_use_mask_load_store. Update after variable
5968 (predicate_load_or_store): New function, split out from
5969 predicate_mem_writes.
5970 (check_redundant_cond_expr): New function.
5971 (value_available_p): Likewise.
5972 (predicate_rhs_code): Likewise.
5973 (predicate_mem_writes): Rename to...
5974 (predicate_statements): ...this. Use predicate_load_or_store
5975 and predicate_rhs_code.
5976 (combine_blocks, tree_if_conversion): Update after above name changes.
5977 (ifcvt_local_dce): Handle redundant_ssa_names.
5978 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
5979 general conditional functions.
5980 * tree-vect-stmts.c (vectorizable_call): Likewise.
5982 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5983 Alan Hayward <alan.hayward@arm.com>
5984 David Sherwood <david.sherwood@arm.com>
5986 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
5987 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
5988 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
5989 plus and minus and convert them into IFN_COND_FMA-based sequences.
5990 (convert_mult_to_fma): Handle conditional plus and minus.
5992 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5994 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
5995 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
5996 (cond_fnms_optab): New optabs.
5997 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
5999 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6000 * internal-fn.h (get_conditional_internal_fn): Declare.
6001 (get_unconditional_internal_fn): Likewise.
6002 * internal-fn.c (cond_ternary_direct): New macro.
6003 (expand_cond_ternary_optab_fn): Likewise.
6004 (direct_cond_ternary_optab_supported_p): Likewise.
6005 (FOR_EACH_COND_FN_PAIR): Likewise.
6006 (get_conditional_internal_fn): New function.
6007 (get_unconditional_internal_fn): Likewise.
6008 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6009 (gimple_match_op::gimple_match_op): Add a new overload for 5
6011 (gimple_match_op::set_op): Likewise.
6012 (gimple_resimplify5): Declare.
6013 * genmatch.c (decision_tree::gen): Generate simplifications for
6015 * gimple-match-head.c (gimple_simplify): Define an overload for
6016 5 operands. Handle calls with 5 arguments in the top-level overload.
6017 (convert_conditional_op): Handle conversions from unconditional
6018 internal functions to conditional ones.
6019 (gimple_resimplify5): New function.
6020 (build_call_internal): Pass a fifth operand.
6021 (maybe_push_res_to_seq): Likewise.
6022 (try_conditional_simplification): Try converting conditional
6023 internal functions to unconditional internal functions.
6024 Handle 3-operand unconditional forms.
6025 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6026 Define ternary equivalents of the current rules for binary conditional
6028 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6030 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6031 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6032 (optab): Handle them.
6033 (SVE_COND_FP_TERNARY): New int iterator.
6034 (sve_fmla_op, sve_fmad_op): New int attributes.
6035 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6036 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6037 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6039 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6041 * target.def (preferred_else_value): New target hook.
6042 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6043 * doc/tm.texi: Regenerate.
6044 * targhooks.h (default_preferred_else_value): Declare.
6045 * targhooks.c (default_preferred_else_value): New function.
6046 * internal-fn.h (conditional_internal_fn_code): Declare.
6047 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6048 (get_conditional_internal_fn): Use it.
6049 (conditional_internal_fn_code): New function.
6050 * gimple-match.h (gimple_match_cond): New struct.
6051 (gimple_match_op): Add a cond member function.
6052 (gimple_match_op::gimple_match_op): Update all forms to take a
6054 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6055 for the suboperation, but don't specify a particular else_value.
6056 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6057 (visit_nary_op, visit_reference_op_load): Pass
6058 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6059 * gimple-match-head.c: Include tree-eh.h
6060 (convert_conditional_op): New function.
6061 (maybe_resimplify_conditional_op): Likewise.
6062 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6063 (gimple_resimplify2): Likewise.
6064 (gimple_resimplify3): Likewise.
6065 (gimple_resimplify4): Likewise.
6066 (maybe_push_res_to_seq): Return null for conditional operations.
6067 (try_conditional_simplification): New function.
6068 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6070 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6072 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6074 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6076 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6078 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6080 (hash_tree): Do not hash DECL_FCONTEXT
6081 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6082 Do not stream DECL_FCONTEXT.
6083 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6084 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6086 2018-07-12 Richard Biener <rguenther@suse.de>
6089 * dwarf2out.c (gen_block_die): Only output blocks when they have
6090 at least one !DECL_IGNORED_P variable.
6092 2018-07-12 Richard Biener <rguenther@suse.de>
6095 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6096 Remove -mieee-fp handling.
6098 2018-07-12 Richard Biener <rguenther@suse.de>
6100 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6101 left-over from last patch.
6103 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6105 PR tree-optimization/86492
6106 * gimple-ssa-store-merging.c
6107 (imm_store_chain_info::coalesce_immediate_stores): Call
6108 check_no_overlap even for the merge_overlapping case. Formatting fix.
6110 2018-07-12 Richard Biener <rguenther@suse.de>
6113 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6114 move possibly trapping operations into the conditional.
6116 2018-07-12 Richard Biener <rguenther@suse.de>
6118 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6119 (vn_lookup_simplify_result): Remove recursion limit applied
6121 (vn_nary_build_or_lookup_1): Adjust.
6122 (try_to_simplify): Likewise.
6123 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6125 (gimple_resimplify2): Likewise.
6126 (gimple_resimplify3): Likewise.
6127 (gimple_resimplify4): Likewise.
6129 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6131 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6132 Use __mmask64 type instead of __mmask8 for __M argument.
6133 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6134 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6136 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6137 __mmask16 for __M argument.
6138 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6139 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6140 to __mmask16 instead of __mmask8.
6141 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6142 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6143 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6144 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6145 instead of __mmask16 for __U argument.
6146 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6147 __mmask16 instead of __mmask8 for __U argument.
6148 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6150 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6152 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6154 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6156 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6158 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6160 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6162 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6164 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6166 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6168 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6169 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6170 return type as well as __M argument type and all casts from __mmask8
6172 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6173 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6174 return type as well as __M argument type and all casts from __mmask8
6176 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6177 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6178 return type as well as __M argument type and all casts from __mmask8
6180 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6181 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6182 return type as well as __M argument type and all casts from __mmask8
6184 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6185 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6188 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6190 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6191 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6194 2018-07-11 Paul Koning <ni1d@arrl.net>
6196 * doc/md.texi (define_subst): Document how multiple occurrences of
6197 the same argument in the replacement pattern are handled.
6199 2018-07-11 Paul Koning <ni1d@arrl.net>
6201 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6203 (Common Type Attributes): Add "mode" attribute.
6205 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6207 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6208 stream DECL_ORIGINAL_TYPE.
6209 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6210 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6211 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6212 Do not walk original type.
6213 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6215 (write_ts_decl_non_common_tree_pointers): Do not stream
6217 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6218 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6220 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6222 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6223 thread twice from the same starting edge.
6225 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6227 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6228 * gimple.c (gimple_call_nonnull_result_p): ...here...
6229 (gimple_call_nonnull_arg): ...and here.
6230 * gimple.h (gimple_call_nonnull_result_p): New.
6231 (gimple_call_nonnull_arg): New.
6233 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6235 * config/arm/arm-cpus.in: Move information from fpu field of each
6236 cpu definition to the isa field.
6237 * config/arm/parsecpu.awk (fpu): Delete match rule.
6238 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6240 2018-07-11 Richard Biener <rguenther@suse.de>
6243 * dwarf2out.c (init_sections_and_labels): Use
6244 output_asm_line_debug_info consistently.
6245 (dwarf2out_early_finish): Likewise.
6246 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6249 2018-07-11 Richard Biener <rguenther@suse.de>
6251 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6252 Rework father_bb setting in a way to avoid propagating constants
6253 multiple times on a loop body.
6255 2018-07-10 Mark Wielaard <mark@klomp.org>
6258 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6260 2018-07-10 Richard Biener <rguenther@suse.de>
6262 * hash-map.h (hash_map::iterator::operator*): Return
6263 references to key and value.
6265 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6268 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6269 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6270 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6271 (gimplify_omp_for): For composite loops, move outer
6272 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6273 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6274 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6275 TREE_LIST for both the original class iterator and the "last" helper
6276 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6277 loop, remember has_decl_expr from outer composite loops for the
6278 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6280 2018-07-09 Martin Sebor <msebor@redhat.com>
6284 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6285 accessing implicitly initialized array elements.
6286 * expr.c (string_constant): Handle string initializers of
6287 character arrays within aggregates.
6288 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6289 Store element offset. As a special case, handle zero size.
6290 (fold_nonarray_ctor_reference): Same.
6291 (fold_ctor_reference): Add argument. Store subobject offset.
6292 * gimple-fold.h (fold_ctor_reference): Add argument.
6294 2018-07-09 Paul Koning <ni1d@arrl.net>
6296 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6297 (pdp11_insn_cost): New function.
6298 (pdp11_md_asm_adjust): New function.
6299 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6300 (pdp11_rtx_costs): Update to match machine better.
6301 (output_addr_const_pdp11): Correct format mismatch warnings.
6302 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6303 * config/pdp11/pdp11.md: General change to add base_cost and/or
6304 length attributes for use by new pdp11_insn_cost function.
6305 (MIN_BRANCH): Correct definition.
6307 (doloop_end): Use standard pattern name for looping pattern.
6308 (doloop_end_nocc): New.
6309 (movsf): Add another constraint alternative.
6310 (zero_extendqihi2): Add constraint alternatives for not in place
6312 (zero_extendhisi2): Remove.
6313 (shift patterns): Add CC handling variants.
6317 (define_peephole2): New peephole to recognize mov that sets CC for
6320 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6322 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6323 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6326 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6329 * real.c (real_nextafter): Return true if result is denormal.
6331 2018-07-09 Martin Liska <mliska@suse.cz>
6333 * common.opt: Add back wrongly removed attribute.
6335 2018-07-09 Richard Biener <rguenther@suse.de>
6338 * dwarf2out.c (gen_block_die): For an early generated DIE
6339 always output high/low PC attributes.
6341 2018-07-09 Tom de Vries <tdevries@suse.de>
6343 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6344 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6345 onto VAR_DECL with abstract origin.
6347 2018-07-07 Jim Wilson <jimw@sifive.com>
6349 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6351 2018-07-07 Tom de Vries <tdevries@suse.de>
6353 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6355 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6357 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6359 (combine_bound): Use wide-int overflow calculation instead of
6361 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6363 * fold-const.c (int_const_binop_2): Same.
6364 (extract_muldiv_1): Same.
6365 (fold_div_compare): Same.
6366 (fold_abs_const): Same.
6368 * poly-int.h (add): Same.
6372 * predict.c (predict_iv_comparison): Same.
6373 * profile-count.c (slow_safe_scale_64bit): Same.
6374 * simplify-rtx.c (simplify_const_binary_operation): Same.
6375 * tree-chrec.c (tree_fold_binomial): Same.
6376 * tree-data-ref.c (split_constant_offset_1): Same.
6377 * tree-if-conv.c (idx_within_array_bound): Same.
6378 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6379 * tree-ssa-phiopt.c (minmax_replacement): Same.
6380 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6381 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6382 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6383 * wide-int.cc (wi::add_large): Same.
6384 (wi::mul_internal): Same.
6385 (wi::sub_large): Same.
6386 (wi::divmod_internal): Same.
6387 * wide-int.h: Change overflow type to overflow_type for neg, add,
6388 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6389 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6390 mul_internal, divmod_internal.
6391 (overflow_type): New enum.
6392 (accumulate_overflow): New.
6394 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6396 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6397 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6399 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6401 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6402 argument is checked for zero before entering loop, avoid checking again.
6404 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6406 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6407 * gimplify.h (generic_expr_could_trap_p): Likwise.
6408 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6410 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6412 PR tree-optimization/86401
6413 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6414 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6415 (fold_bit_and_mask): ... here. New helper function for match.pd.
6416 * fold-const.h (fold_bit_and_mask): Declare.
6417 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6419 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
6422 * target.def (translate_mode_attribute): New hook.
6423 * targhooks.h (default_translate_mode_attribute): Declare.
6424 * targhooks.c (default_translate_mode_attribute): New function.
6425 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6426 * doc/tm.texi: Regenerate.
6427 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6428 (rs6000_translate_mode_attribute): New function.
6430 2018-07-06 Paul Koning <ni1d@arrl.net>
6432 * doc/md.texi (define_split): Document DONE and FAIL.
6433 (define_peephole2): Ditto.
6435 2018-07-05 Jeff Law <law@redhat.com>
6437 PR tree-optimization/86010
6438 * tree-ssa-dse.c (compute_trims): More aggressively trim at
6439 both the head and tail of mem* and str* calls.
6441 2018-07-05 Jim Wilson <jimw@sifive.com>
6443 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
6445 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
6447 * config/aarch64/aarch64-simd.md: correct flags text for
6448 MIN_EXPR replacement.
6450 2018-07-05 James Clarke <jrtc27@jrtc27.com>
6452 * configure: Regenerated.
6454 2018-07-05 Carl Love <cel@us.ibm.com>
6456 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
6457 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
6458 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
6459 VSX_BUILTIN_DOUBLEL_V4SF.
6461 2018-07-05 Martin Sebor <msebor@redhat.com>
6464 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
6465 than its domain to compute its the upper bound of a char array.
6467 2018-07-05 Nathan Sidwell <nathan@acm.org>
6469 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
6470 * doc/cpp.texi: Update comment.
6471 * doc/tm.texi: Rebuilt.
6472 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
6473 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
6474 * doc/extend.texi (Backwards Compatibility): Clarify it is system
6475 headers affected by extern "C".
6476 * system.h: Poison NO_IMPLICIT_EXTERN_C.
6477 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
6478 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
6479 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
6480 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
6481 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
6482 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
6483 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
6484 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
6485 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
6486 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
6487 config/sparc/sp64-elf.h, config/spu/spu.h,
6488 config/stormy16/stormy16.h, config/v850/v850.h,
6489 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
6490 define NO_IMPLICIT_EXTERN_C.
6491 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
6493 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6496 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6497 instead of GET_MODE_SIZE when comparing Units.
6499 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6502 * rtlanal.c (set_noop_p): Constrain on mode change,
6503 include hard-reg-set.h
6505 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6507 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
6509 2018-07-05 Jakub Jelinek <jakub@redhat.com>
6512 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6515 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6517 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6519 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6522 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6524 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6526 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
6529 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
6531 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
6533 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
6534 neg_*_op* variables.
6536 2018-07-04 Martin Liska <mliska@suse.cz>
6538 * tree-switch-conversion.c: Define
6539 max_ratio_for_speed and max_ratio_for_size constants.
6541 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6542 Martin Liska <mliska@suse.cz>
6547 * common.opt: Rename align options with 'str_' prefix.
6548 * common/config/i386/i386-common.c (set_malign_value): New
6550 (ix86_handle_option): Use it to set -falign-* options/
6551 * config/aarch64/aarch64-protos.h (struct tune_params): Change
6552 type from int to string.
6553 * config/aarch64/aarch64.c: Update default values from int
6555 * config/alpha/alpha.c (alpha_override_options_after_change):
6557 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
6558 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6559 max skip conditionally.
6560 * config/i386/freebsd.h (SUBALIGN_LOG): New.
6561 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6562 max skip conditionally.
6563 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6564 max skip conditionally.
6565 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
6566 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6567 max skip conditionally.
6568 * config/i386/i386.c (struct ptt): Change type from int to
6570 (ix86_default_align): Set default values.
6571 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
6572 max skip conditionally.
6573 * config/i386/iamcu.h (SUBALIGN_LOG): New.
6574 (ASM_OUTPUT_MAX_SKIP_ALIGN):
6575 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
6576 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6577 max skip conditionally.
6578 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
6579 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
6580 * config/i386/x86-64.h (SUBALIGN_LOG): New.
6581 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6582 max skip conditionally.
6583 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6584 * config/ia64/ia64.c (ia64_option_override): Set default values
6585 for alignment options.
6586 * config/m68k/m68k.c: Handle new str_align_* options.
6587 * config/mips/mips.c (mips_set_compression_mode): Change
6589 (mips_option_override): Set default values for options.
6590 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6592 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6594 * config/rx/rx.c (rx_option_override): Likewise.
6595 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
6596 (LABEL_ALIGN): Use align_labels_log.
6597 (LOOP_ALIGN): Use align_loops_align.
6598 * config/s390/s390.c (s390_asm_output_function_label): Use new
6600 * config/sh/sh.c (sh_override_options_after_change):
6601 Change type of constants.
6602 * config/spu/spu.c (spu_sched_init): Likewise.
6603 * config/sparc/sparc.c (sparc_option_override): Set default
6605 * config/visium/visium.c (visium_option_override): Likewise.
6606 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6607 emit p2align format with last argument if it's not needed.
6608 * doc/invoke.texi: Document extended format of -falign-*.
6609 * final.c: Use align_labels alignment.
6610 * flags.h (struct target_flag_state): Change type to use
6612 (struct align_flags_tuple): New.
6613 (struct align_flags): Likewise.
6614 (align_loops_log): Redefine macro to use new types.
6615 (align_loops_max_skip): Redefine macro to use new types.
6616 (align_jumps_log): Redefine macro to use new types.
6617 (align_jumps_max_skip): Redefine macro to use new types.
6618 (align_labels_log): Redefine macro to use new types.
6619 (align_labels_max_skip): Redefine macro to use new types.
6620 (align_functions_log): Redefine macro to use new types.
6621 (align_loops): Redefine macro to use new types.
6622 (align_jumps): Redefine macro to use new types.
6623 (align_labels): Redefine macro to use new types.
6624 (align_functions): Redefine macro to use new types.
6625 (align_functions_max_skip): Redefine macro to use new types.
6626 (align_loops_value): New macro.
6627 (align_jumps_value): New macro.
6628 (align_labels_value): New macro.
6629 (align_functions_value): New macro.
6630 * function.c (invoke_set_current_function_hook): Propagate
6631 alignment values from flags to global variables default in
6633 * ipa-icf.c (sem_function::equals_wpa): Use
6634 cl_optimization_option_eq instead of memcmp.
6635 * lto-streamer.h (cl_optimization_stream_out): Support streaming
6637 (cl_optimization_stream_in): Likewise.
6638 * optc-save-gen.awk: Support strings in cl_optimization.
6639 * opth-gen.awk: Likewise.
6640 * opts.c (finish_options): Remove error checking of invalid
6642 (MAX_CODE_ALIGN): Remove.
6643 (MAX_CODE_ALIGN_VALUE): Likewise.
6644 (parse_and_check_align_values): New function.
6645 (check_alignment_argument): Likewise.
6646 (common_handle_option): Use check_alignment_argument.
6647 * opts.h (parse_and_check_align_values): Declare.
6648 * toplev.c (init_alignments): Remove.
6649 (read_log_maxskip): New.
6650 (parse_N_M): Likewise.
6651 (parse_alignment_opts): Likewise.
6652 (backend_init_target): Remove usage of init_alignments.
6653 * toplev.h (parse_alignment_opts): Declare.
6654 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6656 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6657 * tree.c (cl_option_hasher::equal): New.
6658 * varasm.c: Use new global macros.
6660 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6662 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6663 Use a simpler align directive also if MAXSKIP = ALIGN-1.
6664 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6665 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6666 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6667 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6668 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6669 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6670 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6671 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6672 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6673 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6674 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6676 2018-07-04 Martin Liska <mliska@suse.cz>
6677 Jonathan Wakely <jwakely@redhat.com>
6679 * coverage.c: Use correct type.
6680 * doc/invoke.texi: Language correction.
6682 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
6685 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6686 ENDBRANCH for non-tail call which may return via indirect branch.
6687 * doc/extend.texi: Document indirect_return attribute.
6689 2018-07-03 Martin Sebor <msebor@redhat.com>
6691 PR tree-optimization/86274
6692 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6694 (format_floating): Correct handling of infinities and NaNs.
6696 2018-07-03 Martin Sebor <msebor@redhat.com>
6698 * print-tree.c (print_real_cst): New function.
6699 (print_node_brief): Call it.
6700 (print_node): Ditto.
6702 2018-07-03 Jeff Law <law@redhat.com>
6704 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6705 into a single pattern.
6707 * config/h8300/h8300.md (ors code_iterator): New.
6708 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6709 a single pattern and single splitter.
6710 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6711 (iorqi3_1, xorqi3_1): Likewise.
6712 (iorqi3, xorqi3 expanders): Similarly.
6714 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6715 (movmd_internal) into a single pattern using the P mode iterator.
6716 (movmd splitters): Similarly.
6717 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6718 (movsd splitters): Similarly.
6720 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6721 ADDB, ADDW and ADDL into a single ADD attribute which selects the
6722 right table based on the size of the operand.
6723 * config/h8300/h8300.md (length_table): Corresponding changes. All
6724 references to "addb", "addw" and "addl" changed to "add".
6725 (btst patterns): Merge two variants into a single pattern.
6726 (tstqi, tsthi): Likewise.
6727 (addhi3_incdec, addsi3_incdec): Likewise.
6728 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6729 (mulhi3, mulsi3): Likewise.
6730 (udivhi3, udivsi3): Likewise.
6731 (divhi3, divsi3): Likewise.
6732 (andorqi3, andorhi3, andorsi3): Likewise.
6734 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
6737 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
6738 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
6740 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6742 PR tree-optimization/85694
6743 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
6744 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
6746 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
6747 (<u>avg<mode>3_ceil): New patterns.
6749 2018-07-03 David Malcolm <dmalcolm@redhat.com>
6751 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
6752 scan-tree-dump directive.
6753 * gcc.dg/vect/slp-perm-2.c: Likewise.
6754 * gcc.dg/vect/slp-perm-3.c: Likewise.
6755 * gcc.dg/vect/slp-perm-5.c: Likewise.
6756 * gcc.dg/vect/slp-perm-6.c: Likewise.
6757 * gcc.dg/vect/slp-perm-7.c: Likewise.
6758 * gcc.dg/vect/slp-perm-8.c: Likewise.
6760 2018-07-03 Marek Polacek <polacek@redhat.com>
6763 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
6765 2018-07-03 Richard Biener <rguenther@suse.de>
6768 * tree-ssa-structalias.c (find_func_clobbers): Properly
6769 handle indirect calls.
6771 2018-07-03 Jeff Law <law@redhat.com>
6773 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
6774 (shifts): New code iterator.
6775 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
6776 expander. Fix HImode handling on H8/SX.
6777 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
6778 (subqi3, subhi3, subsi3 expanders): Likewise.
6779 (andqi3, andhi3, andsi3 expanders): Likewise.
6780 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
6781 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
6782 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
6783 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
6784 (zero_extendqihi2, zero_extendqisi2): Likewise.
6785 (extendqihi2, extendqisi2): Likewise.
6786 (rotlqi3, rotlhi3, rotlsi3): Likewise.
6787 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
6788 (rotlqi3_1, rotlhi3_1): Likewise.
6789 (logicalhi3_sn, logicalsi3_sn): Likewise.
6790 (logicalhi3, logicalsi3): Likewise.
6792 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6794 * tree-vect-patterns.c (vect_recog_rotate_pattern)
6795 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6796 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
6797 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
6798 type to append_pattern_def_seq instead of creating a stmt_vec_info
6800 (build_mask_conversion): Likewise. Remove vinfo argument.
6801 (vect_add_conversion_to_patterm): Likewise, renaming to...
6802 (vect_add_conversion_to_pattern): ...this.
6803 (vect_recog_mask_conversion_pattern): Update call to
6804 build_mask_conversion. Pass the vector type to
6805 append_pattern_def_seq here too.
6806 (vect_recog_gather_scatter_pattern): Update call to
6807 vect_add_conversion_to_pattern.
6809 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6811 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
6812 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6813 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
6814 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
6815 STMT_VINFO_PATTERN_DEF_SEQ to null here.
6816 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
6817 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
6818 append_pattern_def_seq instead of new_pattern_def_seq.
6819 (vect_recog_divmod_pattern): Do both of the above.
6820 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
6823 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6825 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
6826 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
6827 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
6828 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
6829 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
6830 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6831 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
6832 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6833 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6834 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
6835 parameter with a single stmt_vec_info.
6836 (vect_recog_func_ptr): Likewise.
6837 (vect_recog_gather_scatter_pattern): Likewise, folding in...
6838 (vect_try_gather_scatter_pattern): ...this.
6839 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
6840 the stmt_vec_info of the statement to be matched. Don't clear
6841 STMT_VINFO_RELATED_STMT.
6842 (vect_pattern_recog): Update call accordingly.
6844 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6846 PR tree-optimization/85694
6847 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
6848 (uavgM3_ceil): Document new optabs.
6849 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
6850 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
6852 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
6853 (savg_ceil_optab): New optabs.
6854 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
6855 (vect_vect_recog_func_ptrs): Add it.
6856 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
6857 constant directly from the associated lhs.
6859 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6861 * tree-vect-patterns.c (vect_split_statement): New function.
6862 (vect_convert_input): Use it to try to split an existing cast.
6864 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6866 * poly-int.h (print_hex): New function.
6867 * dumpfile.h (dump_dec, dump_hex): Declare.
6868 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
6869 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
6870 min_input_precision, operation_precision and operation_sign.
6871 * tree-vect-patterns.c (vect_get_range_info): New function.
6872 (vect_same_loop_or_bb_p, vect_single_imm_use)
6873 (vect_operation_fits_smaller_type): Delete.
6874 (vect_look_through_possible_promotion): Add an optional
6875 single_use_p parameter.
6876 (vect_recog_over_widening_pattern): Rewrite to use new
6877 stmt_vec_info infomration. Handle one operation at a time.
6878 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
6879 (vect_truncatable_operation_p, vect_set_operation_type)
6880 (vect_set_min_input_precision): New functions.
6881 (vect_determine_min_output_precision_1): Likewise.
6882 (vect_determine_min_output_precision): Likewise.
6883 (vect_determine_precisions_from_range): Likewise.
6884 (vect_determine_precisions_from_users): Likewise.
6885 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
6886 (vect_vect_recog_func_ptrs): Put over_widening first.
6888 (vect_pattern_recog): Call vect_determine_precisions.
6890 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6892 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
6893 statements that have been replaced by further pattern statements.
6894 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
6896 2018-07-03 Richard Biener <rguenther@suse.de>
6898 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
6899 always set *dt. Dump vectype in vectype overload.
6900 * dumpfile.h (dump_gimple_expr): New function.
6901 (dump_gimple_expr_loc): Likewise.
6902 * dumpfile.c (dump_gimple_expr): New function.
6903 (dump_gimple_expr_loc): Likewise.
6905 2018-07-02 Jeff Law <law@redhat.com>
6907 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
6908 the H8/300, H8/300H and H8/S variants into a single pattern.
6909 (movhi_h8300, movqi_h8300hs): Similarly.
6910 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
6911 (QHI mode iterator): New.
6913 * config/h8300/h8300.md: Remove trailing whitespace.
6915 2018-07-02 Jim Wilson <jimw@sifive.com>
6917 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
6918 instead of emit_insn for interrupt returns.
6919 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
6920 (riscv_sret, riscv_uret): Likewise.
6922 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6924 * pretty-print.c (selftest::test_pp_format): Move save and restore
6925 of quotes to class auto_fix_quotes, and add an instance.
6926 * selftest.c: Include "intl.h".
6927 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
6928 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
6929 * selftest.h (selftest::auto_fix_quotes): New class.
6931 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
6933 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
6934 (aarch64_sve_prepare_conditional_op): Remove.
6935 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
6936 Allow aarch64_simd_reg_or_zero as select operand; remove
6937 the aarch64_sve_prepare_conditional_op call.
6938 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
6939 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6940 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
6941 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
6942 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
6943 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
6944 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
6945 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
6946 and a splitters to match all of the *_any patterns.
6947 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
6949 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
6950 (SVE_COND_FP_BINARY_REV): Remove.
6951 (sve_int_op_rev, sve_fp_op_rev): New.
6952 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
6953 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
6954 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
6955 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
6956 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
6957 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
6958 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
6959 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
6960 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
6962 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
6963 Remove match_dup 1 from the inner unspec.
6964 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6966 * config/aarch64/aarch64.md (movprfx): New attr.
6967 (length): Default movprfx to 8.
6968 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
6969 (*madd<SVE_I>, *msub<SVE_I): Likewise.
6970 (*<su>mul<SVE_I>3_highpart): Likewise.
6971 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
6972 (*v<ASHIFT><SVE_I>3): Likewise.
6973 (*<su><MAXMIN><SVE_I>3): Likewise.
6974 (*<su><MAXMIN><SVE_F>3): Likewise.
6975 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
6976 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
6977 (*div<SVE_F>4): Likewise.
6979 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
6981 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
6984 2018-07-02 Richard Biener <rguenther@suse.de>
6986 PR tree-optimization/86363
6987 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
6988 memset argument refers to a non-variable address.
6990 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6992 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
6993 {PLUS,MINUS}_EXPR code to...
6994 (adjust_symbolic_bound): ...here,
6995 (combine_bound): ...here,
6996 (set_value_range_with_overflow): ...and here.
6998 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7000 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7002 (extract_range_from_abs_expr): ...here.
7004 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7006 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7007 -fno-omit-frame-pointer when not optimizing.
7009 2018-07-02 Martin Liska <mliska@suse.cz>
7012 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7013 (propagate_nothrow): Likewise.
7015 2018-07-02 Martin Liska <mliska@suse.cz>
7018 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7020 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7022 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7023 function in r262149, changing "loc" param from source_location to
7024 const dump_location_t &.
7025 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7026 declaration, as above.
7028 2018-07-01 Paul Koning <ni1d@arrl.net>
7030 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7031 -munit-asm, -mgnu-asm, -mdec-asm.
7032 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7033 (pdp11_output_labelref): New.
7034 (pdp11_output_def): New.
7035 (pdp11_output_addr_vec_elt): New.
7036 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7037 %# and %@ format codes.
7038 (pdp11_option_override): New.
7039 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7040 (pdp11_output_ident): New.
7041 (pdp11_asm_named_section): New.
7042 (pdp11_asm_init_sections): New.
7043 (pdp11_file_start): New.
7044 (pdp11_file_end): New.
7045 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7046 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7047 %o, like %c but octal.
7048 (pdp11_option_override): New.
7049 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7051 (DATA_SECTION_ASM_OP): Ditto.
7052 (READONLY_DATA_SECTION_ASM_OP): New.
7053 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7054 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7055 (ASM_OUTPUT_LABELREF): Ditto.
7056 (ASM_OUTPUT_DEF): Ditto.
7057 (ASM_OUTPUT_EXTERNAL): New.
7058 (ASM_OUTPUT_SOURCE_FILENAME): New.
7059 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7060 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7061 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7062 %# and %@ format codes.
7063 * config/pdp11/pdp11.opt (mgnu-asm): New.
7064 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7065 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7066 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7068 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7070 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7071 dereferencing path[] beyond its length.
7073 (debug_all_paths): New.
7074 (rewire_first_differing_edge): New.
7075 (adjust_paths_after_duplication): New.
7076 (duplicate_thread_path): Call adjust_paths_after_duplication.
7078 (thread_through_all_blocks): Add new argument to
7079 duplicate_thread_path.
7081 2018-06-30 Jim Wilson <jimw@sifive.com>
7083 * config/riscv/predicates.md (p2m1_shift_operand): New.
7084 (high_mask_shift_operand): New.
7085 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7086 pattern using p2m1_shift_operand.
7087 (lshsi3_zero_extend_3+2): New combiner pattern using
7088 high_mask_shift_operand.
7090 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7092 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7094 (vect_recog_rotate_pattern): ...here.
7095 (vect_convert_input): Try to insert casts of invariants in the
7097 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7098 preheader to be empty.
7100 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7102 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7103 vector type. If given, install it in the new statement's
7105 (vect_element_precision): New function.
7106 (vect_unpromoted_value): New struct.
7107 (vect_unpromoted_value::vect_unpromoted_value): New function.
7108 (vect_unpromoted_value::set_op): Likewise.
7109 (vect_look_through_possible_promotion): Likewise.
7110 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7111 (vect_widened_op_tree, vect_convert_input): Likewise.
7112 (vect_convert_inputs, vect_convert_output): Likewise.
7113 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7114 to handle the optional cast of the multiplication result and
7115 vect_widened_op_tree to detect the widened multiplication itself.
7116 Do not require the input and output of promotion casts to have
7117 the same sign, but base the signedness of the operation on the
7118 input rather than the result. If the pattern includes two
7119 promotions, check that those promotions have the same sign.
7120 Do not restrict the MULT_EXPR handling to a double-width result;
7121 handle quadruple-width results and wider. Use vect_convert_inputs
7122 to convert the inputs to the common type.
7123 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7124 to handle the optional cast of the ABS result. Also allow a sign
7125 change or a sign extension between the ABS and MINUS.
7126 Use vect_widened_op_tree to detect the widened subtraction and use
7127 vect_convert_inputs to convert the inputs to the common type.
7128 (vect_handle_widen_op_by_const): Delete.
7129 (vect_recog_widen_op_pattern): New function.
7130 (vect_recog_widen_mult_pattern): Use it.
7131 (vect_recog_widen_shift_pattern): Likewise.
7132 (vect_recog_widen_sum_pattern): Use
7133 vect_look_through_possible_promotion to handle the promoted
7136 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7138 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7139 the containing gimple_seq *.
7140 * gimple-iterator.h (gsi_for_stmt): Declare it.
7141 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7142 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7143 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7144 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7145 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7147 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7149 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7150 statement being replaced is part of an existing pattern
7151 definition sequence, inserting the new pattern statements before
7153 (vect_pattern_recog_1): Don't return a bool. If the statement
7154 is already part of a pattern, instead apply pattern matching
7155 to the pattern definition statements. Don't clear the
7156 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7157 (vect_pattern_recog): Don't break after the first match;
7158 continue processing the pattern definition statements instead.
7159 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7161 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7163 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7164 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7165 (vect_recog_widen_sum_pattern): Use it.
7167 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7169 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7170 phi is not a pattern statement and has not been replaced by
7171 a pattern statement.
7172 * tree-vect-patterns.c (type_conversion_p): Don't check
7173 STMT_VINFO_IN_PATTERN_P.
7174 (vect_recog_vector_vector_shift_pattern): Likewise.
7175 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7176 the pattern statement rather than the original statement; check
7177 directly for a WIDEN_MULT_EXPR here.
7178 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7179 vect_is_simple_use to return the pattern statement rather
7180 than the original statement; use is_pattern_stmt_p to check
7181 for such a pattern statement.
7182 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7183 to return the pattern statement rather than the original statement;
7184 don't do the same transformation here.
7185 (vect_is_simple_use): If the defining statement has been replaced
7186 by a pattern statement, return the pattern statement instead.
7187 Remove the corresponding (local) transformation from the vectype
7190 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7192 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7193 end and default to null.
7194 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7195 (vectorizable_reduction): Update calls accordingly, dropping the
7196 gimple ** argument if the passed-back statement isn't needed.
7197 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7198 (vect_recog_rotate_pattern): Likewise.
7199 (vect_recog_mask_conversion_pattern): Likewise.
7200 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7201 (vect_mask_constant_operand_p): Likewise.
7202 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7203 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7204 (get_group_load_store_type, get_load_store_type): Likewise.
7205 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7206 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7207 (vectorizable_conversion, vectorizable_assignment): Likewise.
7208 (vectorizable_shift, vectorizable_operation): Likewise.
7209 (vectorizable_store, vect_is_simple_cond): Likewise.
7210 (vectorizable_condition, vectorizable_comparison): Likewise.
7211 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7212 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7213 and move it to the end. Cope with null def_stmt_outs.
7215 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7217 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7219 2018-06-29 Jeff Law <law@redhat.com>
7221 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7222 displacements for TARGET_V850E2V3 and newer.
7223 (TARGET_LRA_P): Remove. Defaults to LRA now.
7224 * config/v850/v850.md (sign23byte_load): Remove.
7225 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7226 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7228 2018-06-29 Martin Liska <mliska@suse.cz>
7231 * coverage.c (coverage_init): Mangle full path name.
7232 * doc/invoke.texi: Document the change.
7233 * gcov-io.c (mangle_path): New.
7234 * gcov-io.h (mangle_path): Likewise.
7235 * gcov.c (mangle_name): Use mangle_path for path mangling.
7237 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7239 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7240 if starting source register is not even.
7242 2018-06-29 Martin Liska <mliska@suse.cz>
7244 PR tree-optimization/86263
7245 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7246 Make edge redirection.
7248 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7250 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7251 (dump_scope_depth): New variable.
7252 (get_dump_scope_depth): New function.
7253 (dump_begin_scope): New function.
7254 (dump_end_scope): New function.
7255 * dumpfile.h (get_dump_scope_depth): New declaration.
7256 (dump_begin_scope): New declaration.
7257 (dump_end_scope): New declaration.
7258 (class auto_dump_scope): New class.
7259 (AUTO_DUMP_SCOPE): New macro.
7260 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7263 2018-06-29 Richard Biener <rguenther@suse.de>
7265 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7266 compute_all_dependences succeeds.
7267 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7268 exceed --param loop-max-datarefs-for-datadeps.
7270 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7272 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7274 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7277 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7278 alternative 0 in preferred_for_speed attribute.
7280 2018-06-28 Paul Koning <ni1d@arrl.net>
7282 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7283 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7284 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7285 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7286 for shift insn patterns.
7288 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7290 * cgraph.c (cgraph_node::get_body): Replace assignments to
7291 "dump_file" with calls to set_dump_file.
7292 * dumpfile.c (alt_dump_file): Make static, and group with...
7293 (alt_flags): ...this definition.
7294 (dumps_are_enabled): New variable.
7295 (refresh_dumps_are_enabled): New function.
7296 (set_dump_file): New function.
7297 (set_alt_dump_file): New function.
7298 (gcc::dump_manager::dump_start): Replace assignments to
7299 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7301 (gcc::dump_manager::dump_finish): Likewise.
7302 * dumpfile.h (alt_dump_file): Delete decl.
7303 (dumps_are_enabled): New variable decl.
7304 (set_dump_file): New function decl.
7305 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7307 * tree-nested.c (lower_nested_functions): Replace assignments to
7308 "dump_file" with calls to set_dump_file.
7310 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7312 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7313 goto_locus of each outgoing edge of each basic block.
7315 2018-06-28 Richard Biener <rguenther@suse.de>
7317 * dwarf2out.c (decl_scope_table): Remove.
7318 (push_decl_scope): Likewise.
7319 (pop_decl_scope): Likewise.
7320 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7321 (gen_struct_or_union_type_die): Likewise.
7322 (gen_tagged_type_die): Likewise.
7323 (dwarf2out_init): Do not initialize decl_scope_table.
7324 (dwarf2out_c_finalize): Do not free it.
7326 2018-06-28 Richard Biener <rguenther@suse.de>
7328 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7329 deciding whether to not re-use a DIE.
7331 2018-06-28 Richard Biener <rguenther@suse.de>
7333 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7334 DW_AT_abstract_origin attribute.
7336 2018-06-28 Martin Liska <mliska@suse.cz>
7338 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7339 Use newly introduced constants.
7340 * tree-switch-conversion.h (struct jump_table_cluster):
7341 Define max_ratio_for_size and max_ratio_for_speed.
7343 2018-06-28 Martin Liska <mliska@suse.cz>
7345 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7346 Add new checking assert to catch invalid state.
7347 (jump_table_cluster::can_be_handled): Handle single case
7349 (jump_table_cluster::is_beneficial): Bail out for such case.
7350 (bit_test_cluster::find_bit_tests):
7351 Add new checking assert to catch invalid state.
7352 (bit_test_cluster::can_be_handled): Handle single case
7354 (bit_test_cluster::is_beneficial): Bail out for such case.
7355 (switch_decision_tree::analyze_switch_statement):
7358 2018-06-28 Martin Liska <mliska@suse.cz>
7360 * common.opt: Introduce -completion option.
7361 * gcc.c (driver_handle_option): Handle it.
7362 (driver::main): Print completions if completion
7364 * opt-suggestions.c (option_proposer::get_completions):
7366 (option_proposer::suggest_completion): Likewise.
7367 (option_proposer::find_param_completions): Likewise.
7368 (verify_autocompletions): Likewise.
7369 (test_completion_valid_options): Likewise.
7370 (test_completion_valid_params): Likewise.
7371 (in_completion_p): Likewise.
7372 (empty_completion_p): Likewise.
7373 (test_completion_partial_match): Likewise.
7374 (test_completion_garbage): Likewise.
7375 (opt_proposer_c_tests): Likewise.
7376 * opt-suggestions.h: Declare new functions.
7377 * opts.c (common_handle_option): Handle OPT__completion_.
7378 * selftest-run-tests.c (selftest::run_tests): Add
7379 opt_proposer_c_tests.
7380 * selftest.c (assert_str_startswith): New.
7381 * selftest.h (assert_str_startswith): Likewise.
7382 (opt_proposer_c_tests): New.
7383 (ASSERT_STR_STARTSWITH): Likewise.
7385 2018-06-28 Martin Liska <mliska@suse.cz>
7387 * Makefile.in: Add opt-suggestions.o.
7388 * gcc-main.c: Include opt-suggestions.h.
7389 * gcc.c (driver::driver): Likewise.
7390 (driver::~driver): Remove m_option_suggestions.
7391 (driver::build_option_suggestions): Moved to option_proposer.
7392 (driver::suggest_option): Likewise.
7393 (driver::handle_unrecognized_options): Use option_proposer.
7394 * gcc.h (class driver): Add new memver m_option_proposer.
7395 * opt-suggestions.c: New file.
7396 * opt-suggestions.h: New file.
7398 2018-06-28 Martin Liska <mliska@suse.cz>
7400 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7401 (auto_string_vec::~auto_string_vec): Likewise.
7403 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7405 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7406 prevent_decl_creation_for_types fields up and add reset_location field.
7407 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7408 statement if id->reset_location is true.
7409 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7410 id->reset_location is true.
7411 (copy_phis_for_bb): Force input_location on the arguments if
7412 id->reset_location is true.
7413 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7414 is set on the function to be inlined.
7416 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7418 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7420 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
7422 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7423 registers for Pmode.
7424 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7425 hard registers for the clobbered pseudo.
7427 2018-06-27 Paul Koning <ni1d@arrl.net>
7429 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7430 mutually exclusive options.
7431 * config/pdp11/constraints.md (h): New constraint.
7432 (O): Update definition to match shift code generation.
7433 (D): New constraint.
7434 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
7436 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
7438 (output_jump): Change arguments.
7439 (pdp11_fixed_cc_regs): New function.
7440 (pdp11_cc_mode): Ditto.
7441 (pdp11_expand_shift): Ditto.
7442 (pdp11_assemble_shift): Ditto.
7443 (pdp11_small_shift): Ditto.
7444 (pdp11_branch_cost): Remove.
7445 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
7447 (pdp11_register_move_cost): Update for CC registers.
7448 (pdp11_rtx_costs): Add case for LSHIFTRT.
7449 (pdp11_output_jump): Add CCNZ mode conditional branches.
7450 (notice_update_cc_on_set): Remove.
7451 (pdp11_cc_mode): New function.
7452 (simple_memory_operand): Correct pre/post decrement case.
7453 (no_side_effect_operand): New function.
7454 (pdp11_regno_reg_class): Add CC_REGS class.
7455 (pdp11_fixed_cc_regs): New function.
7456 (pdp11_small_shift): New function.
7457 (pdp11_expand_shift): New function to expand shift insns.
7458 (pdp11_assemble_shift): New function to output shifts.
7459 (pdp11_branch_cost): Remove.
7460 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
7461 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
7462 (WCHAR_TYPE): Ditto.
7463 (PTRDIFF_TYPE): Ditto.
7464 (ADJUST_INSN_LENGTH): New macro.
7465 (FIXED_REGISTERS): Add CC registers.
7466 (CALL_USED_REGISTERS): Ditto.
7468 (REG_CLASS_NAMES): Ditto.
7469 (REG_CLASS_CONTENTS): Ditto.
7470 (SELECT_CC_MODE): Use new function.
7471 (TARGET_FLAGS_REGNUM): New macro.
7472 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
7473 (cc0_reg_rtx): Remove.
7474 (CC_STATUS_MDEP): Remove.
7475 (CC_STATUS_MDEFP_INIT): Remove.
7476 (CC_IN_FPU): Remove.
7477 (NOTICE_UPDATE_CC): Remove.
7478 (REGISTER_NAMES): Add CC registers.
7479 (BRANCH_COST): Change to constant 1.
7480 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
7482 * config/pdp11/pdp11.opt (mbcopy): Remove.
7483 (mbcopy-builtin): Remove.
7484 (mbranch-cheap): Remove.
7485 (mbranch-expensive): Remove.
7486 * config/pdp11/predicates.md (expand_shift_operand): Update to
7487 match shift code generation.
7488 (ccnz_operator): New predicate.
7489 * doc/invoke.texi (PDP-11 Options): Remove deleted options
7490 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
7491 Remove non-existent option -mabshi, -mno-abshi. Document mutually
7493 * doc/md.texi (PDP-11): Document new D and h constraints. Update
7494 description of O constraint.
7496 2018-06-27 Jeff Law <law@redhat.com>
7497 Austin Law <austinklaw@gmail.com>
7499 * config/v850/v850.md (addsi3_set_flags): New pattern.
7500 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
7501 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
7502 (zero_extendhisi2_v850_set_flags): Likewise.
7503 (zero_extendqisi2_v850_set_flags): Likewise.
7504 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
7505 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
7506 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
7508 * config/v850/v850-protos.h (notice_update_cc): Remove.
7509 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
7510 (v850_print_operand): Handle 'D' and "d".
7511 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
7512 Add handling of arithmetic/logical operations compared against zero.
7513 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
7514 Do not look at v850_compare_op, instead get mode from last argument.
7515 (v850_gen_compare): Remove
7516 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
7517 after reload for prologue insns.
7518 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
7520 (construct_save_jarl): Likewise.
7521 (TARGET_FLAGS_REGNUM): Define.
7522 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
7523 (NOTICE_UPDATE_CC): Remove.
7524 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
7525 than cc0. Conditionalize on reload_completed.
7526 (cmpsi_insn, setfcc_insn): Likewise.
7527 (tst1 splitter): Turn into define_and_split which sets the flags
7529 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
7530 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
7531 (cstoresf4, cstoredf4): Clobber the flags.
7532 (cmpsi, cmpsf, cmpdf): Remove expanders.
7533 (setf_insn): Remove pattern.
7534 (addsi3): Turn into define_and_split which clobbers the flags after
7535 reload and a suitable pattern (addsi3_clobber_flags) for use after
7537 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
7538 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
7539 (ashrsi3, ashrsi3_v850e2): Likewise.
7540 (bins): Clobber the flags.
7541 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
7542 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
7543 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
7544 (call_value_internal_short, call_value_internal_long): Likewise.
7545 (callt_save_interrupt, callt_return_interrupt): Likewise.
7546 (save_interrupt, return_interrupt): Likewise.
7547 (callt_save_all_interrupt, save_all_interrupt): Likewise.
7548 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
7549 (restore_all_interrupt, _restore_all_interrupt): Likewise.
7550 (All FP comparisons): Only allow after reload has completed.
7552 (divh, divhu): Tweak output template.
7553 (branch_z_normal, branch_z_invert): Remove
7554 (branch_nz_normal, branch_nz_invert): Likewise.
7555 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
7557 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
7558 * config/v850/v850.c (notice_update_cc): Remove.
7559 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
7560 (CC_NO_CARRY): Likewise.
7561 (NOTICE_UPDATE_CC): Define to nothing.
7562 * config/v850/v850.md: Remove block comment on cc0 handling
7563 Remove "cc" attribute from all patterns. Remove cc_status handling
7564 from all patterns. Minor formatting fixes.
7566 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7568 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
7569 (cortex-a76.cortex-a55): Likewise.
7570 * config/aarch64/aarch64-tune.md: Regenerate.
7571 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
7572 cortex-a76.cortex-a55.
7574 2018-06-27 Jeff Law <law@redhat.com>
7576 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
7577 (MULTILIB_DIRNAMES): Similarly.
7579 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
7581 * gimple.h (gimple_return_retbnd): Delete.
7582 (gimple_return_set_retbnd): Likewise.
7583 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
7584 gimple_return_set_retbnd.
7585 * gimple-pretty-print.c (dump_gimple_return): Remove call to
7586 gimple_return_retbnd and adjust.
7587 * tree-inline.h (struct copy_body_data): Remove retbnd field.
7588 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
7589 Explicitly return NULL in a couple more cases. Move assertion
7590 on debug statements and remove unreachable code.
7591 (reset_debug_binding): Do not test id->retbnd.
7592 (expand_call_inline): Do not set it.
7594 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7596 * configure.ac: Add --disable-gcov option.
7597 * configure: Regenerate.
7598 * Makefile.in: Honour @enable_gcov@.
7599 * doc/install.texi: Document --disable-gcov.
7601 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7603 * config/arm/arm-cpus.in (cortex-a76): New entry.
7604 (cortex-a76.cortex-a55): Likewise.
7605 * config/arm/arm-tables.opt: Regenerate.
7606 * config/arm/arm-tune.md: Likewise.
7607 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7608 * doc/invoke.texi (ARM Options): Document cortex-a76 and
7609 cortex-a76.cortex-a55.
7611 2018-06-27 Tamar Christina <tamar.christina@arm.com>
7614 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7616 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
7618 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7620 (EPILOGUE_USES): Likewise.
7622 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
7624 * tree-inline.c (remap_location): New function extracted from...
7625 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
7626 (copy_phis_for_bb): ...here. Call remap_location.
7627 (copy_cfg_body): Adjust call to copy_edges_for_bb.
7629 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
7631 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7632 unaligned vsx for 16B memset.
7634 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
7637 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7638 ieee128_float_type_node to long_double_type_node unless
7639 TARGET_LONG_DOUBLE_128 is set.
7641 2018-06-26 David Malcolm <dmalcolm@redhat.com>
7643 * cfgloop.c (get_loop_location): Convert return type from
7644 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7645 by implicit construction from rtx_insn *, and using
7646 dump_user_location_t::from_function_decl for the fallback case.
7647 * cfgloop.h (get_loop_location): Convert return type from
7648 location_t to dump_user_location_t.
7649 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7650 dump_printf_loc to pass in a dump_location_t rather than a
7651 location_t, via the gimple stmt.
7652 * coverage.c (get_coverage_counts): Update calls to
7653 dump_printf_loc to pass in dump_location_t rather than a
7655 * doc/optinfo.texi (Dump types): Convert example of
7656 dump_printf_loc from taking "locus" to taking "insn". Update
7657 description of the "_loc" calls to cover dump_location_t.
7658 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7660 (dump_user_location_t::dump_user_location_t): New constructors,
7661 from gimple *stmt and rtx_insn *.
7662 (dump_user_location_t::from_function_decl): New function.
7663 (dump_loc): Make static.
7664 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7665 const dump_location_t &.
7666 (dump_generic_expr_loc): Delete.
7667 (dump_printf_loc): Convert param "loc" from location_t to
7668 const dump_location_t &.
7669 (selftest::test_impl_location): New function.
7670 (selftest::dumpfile_c_tests): New function.
7671 * dumpfile.h: Include "profile-count.h".
7672 (class dump_user_location_t): New class.
7673 (struct dump_impl_location_t): New struct.
7674 (class dump_location_t): New class.
7675 (dump_printf_loc): Convert 2nd param from source_location to
7676 const dump_location_t &.
7677 (dump_generic_expr_loc): Delete.
7678 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7679 const dump_location_t &.
7680 * gimple-fold.c (fold_gimple_assign): Update call to
7681 dump_printf_loc to pass in a dump_location_t rather than a
7682 location_t, via the gimple stmt.
7683 (gimple_fold_call): Likewise.
7684 * gimple-loop-interchange.cc
7685 (loop_cand::analyze_iloop_reduction_var): Update for change to
7686 check_reduction_path.
7687 (tree_loop_interchange::interchange): Update for change to
7689 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7690 change in return-type of find_loop_location.
7691 (graphite_regenerate_ast_isl): Likewise.
7692 * graphite-optimize-isl.c (optimize_isl): Likewise.
7693 * graphite.c (graphite_transform_loops): Likewise.
7694 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7695 pass in a dump_location_t rather than a location_t, via the
7697 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7698 * ipa.c (walk_polymorphic_call_targets): Likewise.
7699 * loop-unroll.c (report_unroll): Convert "locus" param from
7700 location_t to dump_location_t.
7701 (decide_unrolling): Update for change to get_loop_location's
7703 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7704 location_t to dump_user_location_t.
7705 (grid_find_single_omp_among_assignments_1): Updates calls to
7706 dump_printf_loc to pass in a dump_location_t rather than a
7707 location_t, via the gimple stmt.
7708 (grid_parallel_clauses_gridifiable): Convert "tloc" from
7709 location_t to dump_location_t. Updates calls to dump_printf_loc
7710 to pass in a dump_location_t rather than a location_t, via the
7712 (grid_inner_loop_gridifiable_p): Likewise.
7713 (grid_dist_follows_simple_pattern): Likewise.
7714 (grid_gfor_follows_tiling_pattern): Likewise.
7715 (grid_target_follows_gridifiable_pattern): Likewise.
7716 (grid_attempt_target_gridification): Convert initialization
7717 of local "grid" from memset to zero-initialization; FIXME: does
7718 this require C++11? Update call to dump_printf_loc to pass in a
7719 optinfo_location rather than a location_t, via the gimple stmt.
7720 * profile.c (read_profile_edge_counts): Updates call to
7721 dump_printf_loc to pass in a dump_location_t rather than a
7723 (compute_branch_probabilities): Likewise.
7724 * selftest-run-tests.c (selftest::run_tests): Call
7726 * selftest.h (dumpfile_c_tests): New decl.
7727 * tree-loop-distribution.c (pass_loop_distribution::execute):
7728 Update for change in return type of find_loop_location.
7729 * tree-parloops.c (parallelize_loops): Likewise.
7730 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7731 "locus" from location_t to dump_user_location_t.
7732 (canonicalize_loop_induction_variables): Likewise.
7733 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
7734 for change in return type of find_loop_location.
7735 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
7736 to dump_printf_loc to pass in a dump_location_t rather than a
7737 location_t, via the stmt.
7738 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
7740 * tree-vect-loop-manip.c (find_loop_location): Convert return
7741 type from source_location to dump_user_location_t.
7742 (vect_do_peeling): Update for above change.
7743 (vect_loop_versioning): Update for change in type of
7745 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
7746 from location_t to dump_user_location_t.
7747 (vect_estimate_min_profitable_iters): Update for change in type
7749 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
7750 location_t to dump_location_t.
7751 (vect_slp_bb): Update for change in type of vect_location.
7752 * tree-vectorizer.c (vect_location): Convert from source_location
7753 to dump_user_location_t.
7754 (try_vectorize_loop_1): Update for change in vect_location's type.
7755 (vectorize_loops): Likewise.
7756 (increase_alignment): Likewise.
7757 * tree-vectorizer.h (vect_location): Convert from source_location
7758 to dump_user_location_t.
7759 (find_loop_location): Convert return type from source_location to
7760 dump_user_location_t.
7761 (check_reduction_path): Convert 1st param from location_t to
7762 dump_user_location_t.
7763 * value-prof.c (check_counter): Update call to dump_printf_loc to
7764 pass in a dump_user_location_t rather than a location_t; update
7765 call to error_at for change in type of "locus".
7766 (check_ic_target): Update call to dump_printf_loc to
7767 pass in a dump_user_location_t rather than a location_t, via the
7770 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
7772 * config/s390/s390.h (enum processor_flags): Do not use
7773 default tune parameter when -march was specified.
7775 2018-06-26 Jakub Jelinek <jakub@redhat.com>
7778 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
7779 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
7782 2018-06-26 Richard Biener <rguenther@suse.de>
7784 PR tree-optimization/86287
7786 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
7787 (vect_analyze_loop): Initialize n_stmts.
7789 2018-06-26 Richard Biener <rguenther@suse.de>
7792 * fold-const.c (fold_convertible_p): Pointer extension
7795 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
7798 * dwarf2out.c (loc_list_has_views): Adjust comments.
7799 (dw_loc_list): Split single cross-partition range with
7802 2018-06-25 Jeff Law <law@redhat.com>
7804 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
7805 on -mbig-switch by default.
7807 * config/v850/predicates.md (const_float_1_operand): Fix match_code
7809 (const_float_0_operand): Remove unused predicate.
7810 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
7811 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
7812 (recipsf2): New expander. Original pattern now called
7814 (recipdf2, recipdf2_insn): Similarly.
7815 (rsqrtsf2, rsqrtsf2_insn): Similarly
7816 (rsqrtdf2, rsqrtdf2_insn): Similarly
7818 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
7820 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
7821 Simplify logic for FreeBSD (twice).
7823 2018-06-25 Martin Sebor <msebor@redhat.com>
7825 PR tree-optimization/86204
7826 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
7827 a strnlen result if it's less than the length of the string.
7829 2018-06-25 Martin Sebor <msebor@redhat.com>
7831 PR tree-optimization/85700
7832 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
7833 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
7834 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
7836 2018-06-25 Martin Sebor <msebor@redhat.com>
7838 * doc/extend.texi (Zero-length arrays): Update and clarify.
7840 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
7842 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
7843 added IEEE/IBM long double multilib support on PowerPC little
7844 endian Linux systems.
7845 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
7846 (MULTILIB_DEFAULTS): Likewise.
7847 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7849 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
7850 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
7851 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
7853 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
7856 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
7857 (REORDER_45): Likewise.
7859 2018-06-25 Jeff Law <law@redhat.com>
7861 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
7862 dividend to 32 bits. Adjust length.
7863 (udivmodhi4): Cleanup output template. Fix length.
7865 2018-06-25 Carl Love <cel@us.ibm.com>
7867 * config/rs6000/vsx.md: Change word selector to prefered location.
7869 2018-06-25 Richard Biener <rguenther@suse.de>
7871 PR tree-optimization/86304
7872 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
7873 epilogue-if-converted loops as well.
7875 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
7877 * lto-section-out.c (lto_begin_section): Do not print section
7878 name for noaddr and unnumbered dumps.
7880 2018-06-25 Richard Biener <rguenther@suse.de>
7882 * tree-vectorizer.h (struct vec_info_shared): New structure
7883 with parts split out from struct vec_info and loop_nest from
7884 struct _loop_vec_info.
7885 (struct vec_info): Adjust accordingly.
7886 (struct _loop_vec_info): Likewise.
7887 (LOOP_VINFO_LOOP_NEST): Adjust.
7888 (LOOP_VINFO_DATAREFS): Likewise.
7889 (LOOP_VINFO_DDRS): Likewise.
7890 (struct _bb_vec_info): Likewise.
7891 (BB_VINFO_DATAREFS): Likewise.
7892 (BB_VINFO_DDRS): Likewise.
7893 (struct _stmt_vec_info): Add dr_aux member.
7894 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
7895 (DR_MISALIGNMENT_UNINITIALIZED): New.
7896 (set_dr_misalignment): Adjust.
7897 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
7898 (vect_analyze_loop): Adjust prototype.
7899 (vect_analyze_loop_form): Likewise.
7900 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7901 Compute dependences lazily.
7902 (vect_record_base_alignments): Use shared datarefs/ddrs.
7903 (vect_verify_datarefs_alignment): Likewise.
7904 (vect_analyze_data_refs_alignment): Likewise.
7905 (vect_analyze_data_ref_accesses): Likewise.
7906 (vect_analyze_data_refs): Likewise.
7907 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
7908 constructor parameter for shared part.
7909 (vect_analyze_loop_form): Pass in shared part and adjust.
7910 (vect_analyze_loop_2): Pass in storage for the number of
7911 stmts. Move loop nest finding to the caller. Compute
7913 (vect_analyze_loop): Pass in shared part.
7914 (vect_transform_loop): Verify shared datarefs are unchanged.
7915 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
7916 constructor parameter for shared part.
7917 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
7918 (vect_slp_bb): Verify shared datarefs are unchanged before
7920 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
7922 (new_stmt_vec_info): Initialize DR_AUX misalignment to
7923 DR_MISALIGNMENT_UNINITIALIZED.
7924 * tree-vectorizer.c (vec_info::vec_info): Add constructor
7925 parameter for shared part.
7926 (vec_info::~vec_info): Adjust.
7927 (vec_info_shared::vec_info_shared): New.
7928 (vec_info_shared::~vec_info_shared): Likewise.
7929 (vec_info_shared::save_datarefs): Likewise.
7930 (vec_info_shared::check_datarefs): Likewise.
7931 (try_vectorize_loop_1): Construct shared part live for analyses
7932 of a single loop for multiple vector sizes.
7933 * tree-parloops.c (gather_scalar_reductions): Adjust.
7935 2018-06-25 Richard Biener <rguenther@suse.de>
7937 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
7938 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
7939 (vect_analyze_data_refs): Remove similar code from here and
7940 simplify accordingly.
7942 2018-06-25 Richard Biener <rguenther@suse.de>
7944 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
7945 for reverse storage order accesses rather than asserting
7946 they cannot happen here.
7948 2018-06-25 Tom de Vries <tdevries@suse.de>
7951 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
7952 Use data16 instead of .byte for insn prefix.
7954 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
7957 * parser.c (make_char_string_pack): Pass this literal chars
7958 through cpp_interpret_string.
7959 (cp_parser_userdef_numeric_literal): Check the result of
7960 make_char_string_pack.
7962 2018-06-24 Maya Rashish <coypu@sdf.org>
7964 * ginclude/stddef.h: Simplify conditions around avoiding
7965 re-definition of __size_t.
7967 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7969 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
7970 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
7972 2018-06-22 Maya Rashish <coypu@sdf.org>
7974 * doc/invoke.texi (mno-fancy-math-387): Update for changes
7975 made to OpenBSD and NetBSD through the years.
7977 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7979 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7980 behavior of vec_pack (vector double, vector double) to match
7981 behavior of vec_float2 (vector double, vector double).
7983 2018-06-22 Olivier Hainque <hainque@adacore.com>
7985 * gimplify.c (gimplify_function_tree): Prevent creation
7986 of a trampoline for the address of the current function
7987 passed to entry/exit instrumentation hooks.
7989 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
7992 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
7995 2018-06-22 Martin Liska <mliska@suse.cz>
7997 PR tree-optimization/86263
7998 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7999 Bail out if is_enabled is false.
8000 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8002 (jump_table_cluster::is_enabled): New function.
8004 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8006 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8007 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8008 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8009 (lto_input_ts_binfo_tree_pointers): Likewise.
8010 * tree-streamer-out.c (streamer_write_tree_bitfields,
8011 write_ts_binfo_tree_pointers): Likewise.
8012 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8014 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8016 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8018 2018-06-22 Martin Liska <mliska@suse.cz>
8020 * symbol-summary.h (get): Make it pure and inline move
8021 functionality from ::get function.
8022 (get): Remove and inline into ::get and ::get_create.
8023 (get_create): Move code from ::get function.
8025 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8028 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8029 -x assembler-with-cpp.
8031 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8033 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8034 _FILE_OFFSET_BITS=64 for C++.
8036 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8038 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8039 conversion insn that shows up when pr85657-3.c is compiled using
8040 IEEE 128-bit long double.
8041 (neg<mode>2_internal): Use the correct mode to check whether the
8042 mode is IBM extended.
8043 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8044 multiply and divide external functions from being created more
8047 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8049 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8051 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8052 the edge can be forwarded.
8053 (cfg_layout_merge_blocks): Likewise.
8055 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8057 * except.c (finish_eh_generation): Commit edge insertions only after
8058 the EH edges have been redirected from post-landing to landing pads.
8060 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8062 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8063 create_tmp_var_for to create the FRAME decl.
8064 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8066 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8068 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8069 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8070 debug statement when resetting its value.
8071 (expand_call_inline): Copy the locus of the call onto the assignment
8072 of the return value, if any. Use local variable in more cases.
8074 2018-06-21 Martin Liska <mliska@suse.cz>
8076 * ipa-pure-const.c (propagate_nothrow): Use
8077 funct_state_summaries->get.
8078 (dump_malloc_lattice): Likewise.
8079 (propagate_malloc): Likewise.
8081 2018-06-21 Richard Biener <rguenther@suse.de>
8083 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8084 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8085 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8087 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8088 BLOCK_ABSTRACT_ORIGIN unconditionally.
8090 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8092 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8094 * ipa-reference.c (ipa_reference_c_finalize): Delete
8095 ipa_ref_opt_sum_summaries and set it to NULL.
8097 2018-06-21 Tom de Vries <tdevries@suse.de>
8099 PR tree-optimization/85859
8100 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8101 test with comment from bb_no_side_effects_p.
8103 2018-06-21 Richard Biener <rguenther@suse.de>
8105 PR tree-optimization/86232
8106 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8107 max for constant niter.
8109 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8111 * config/aarch64/aarch64-simd.md
8112 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8114 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8116 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8117 Make opernads of the unspec commutative.
8119 2018-06-21 Richard Biener <rguenther@suse.de>
8121 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8122 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8123 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8124 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8126 (vect_record_base_alignments): Do not record base alignment
8127 for gathers or scatters.
8128 (vect_compute_data_ref_alignment): Drop return value that is always
8129 true. Bail out early for gathers or scatters.
8130 (vect_enhance_data_refs_alignment): Bail out early for gathers
8132 (vect_find_same_alignment_drs): Likewise.
8133 (vect_analyze_data_refs_alignment): Remove dead code.
8134 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8135 (vect_analyze_data_refs): For possible gathers or scatters do
8136 not create an alternate DR, just check their possible validity
8137 and mark them. Adjust DECL_NONALIASED handling to not rely
8139 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8140 update inits of gathers or scatters.
8141 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8142 Also copy gather/scatter flag to pattern vinfo.
8144 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8146 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8147 behavior of vec_packsu (vector unsigned long long, vector unsigned
8148 long long) to match behavior of vec_packs with same signature.
8150 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8151 Thomas Schwinge <thomas@codesourcery.com>
8152 Cesar Philippidis <cesar@codesourcery.com>
8154 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8155 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8156 (gimplify_adjust_omp_clauses): Likewise.
8157 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8158 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8159 (gimplify_omp_target_update): Update handling of acc update and
8161 * omp-low.c (install_var_field): Remove unused parameter
8162 base_pointers_restrict.
8163 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8164 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8166 (omp_target_base_pointers_restrict_p): Delete.
8167 (scan_omp_target): Update call to scan_sharing_clauses.
8168 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8170 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8171 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8172 (convert_local_omp_clauses): Likewise.
8173 * tree-pretty-print.c (dump_omp_clause): Likewise.
8174 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8176 (omp_clause_code_name): Likewise.
8178 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8181 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8184 PR tree-optimization/86231
8185 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8186 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8188 2018-06-20 Tom de Vries <tdevries@suse.de>
8190 PR tree-optimization/86097
8191 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8192 iv type if signedness of iv type is not the same as that of *nit.
8194 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8196 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8197 EDGE_EH edges, verify they are all EDGE_EH.
8199 2018-06-20 Maya Rashish <coypu@sdf.org>
8201 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8203 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8205 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8206 * config/aarch64/aarch64.c (xgene1_tunings): Add
8207 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8208 (aarch64_mode_valid_for_sched_fusion_p):
8209 Allow 16-byte modes.
8210 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8211 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8213 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8215 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8216 * config/aarch64/iterators.md (VQ2): New mode iterator.
8218 2018-06-20 Martin Liska <mliska@suse.cz>
8220 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8221 Change default ratio from 10 to 8.
8223 2018-06-20 Martin Liska <mliska@suse.cz>
8225 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8227 (bit_test_cluster::find_bit_tests): Likewise.
8228 (switch_decision_tree::analyze_switch_statement): Find clusters.
8229 * tree-switch-conversion.h (struct jump_table_cluster): Document
8232 2018-06-20 Martin Liska <mliska@suse.cz>
8234 * tree-switch-conversion.c (switch_conversion::collect):
8235 Record m_uniq property.
8236 (switch_conversion::expand): Bail out for special conditions.
8237 (group_cluster::~group_cluster): New.
8238 (group_cluster::group_cluster): Likewise.
8239 (group_cluster::dump): Likewise.
8240 (jump_table_cluster::emit): New.
8241 (switch_decision_tree::fix_phi_operands_for_edges): New.
8242 (struct case_node): Remove struct.
8243 (jump_table_cluster::can_be_handled): New.
8244 (case_values_threshold): Moved to header.
8245 (reset_out_edges_aux): Likewise.
8246 (jump_table_cluster::is_beneficial): New.
8247 (bit_test_cluster::can_be_handled): Likewise.
8248 (add_case_node): Remove.
8249 (bit_test_cluster::is_beneficial): New.
8250 (case_bit_test::cmp): New.
8251 (bit_test_cluster::emit): New.
8252 (expand_switch_as_decision_tree_p): Remove.
8253 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8254 (fix_phi_operands_for_edge): Likewise.
8255 (switch_decision_tree::analyze_switch_statement): New.
8256 (compute_cases_per_edge): Move ...
8257 (switch_decision_tree::compute_cases_per_edge): ... here.
8258 (try_switch_expansion): Likewise.
8259 (switch_decision_tree::try_switch_expansion): Likewise.
8260 (record_phi_operand_mapping): Likewise.
8261 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8262 (emit_case_decision_tree): Likewise.
8263 (switch_decision_tree::emit): Likewise.
8264 (balance_case_nodes): Likewise.
8265 (switch_decision_tree::balance_case_nodes): Likewise.
8266 (dump_case_nodes): Likewise.
8267 (switch_decision_tree::dump_case_nodes): Likewise.
8268 (emit_jump): Likewise.
8269 (switch_decision_tree::emit_jump): Likewise.
8270 (emit_cmp_and_jump_insns): Likewise.
8271 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8272 (emit_case_nodes): Likewise.
8273 (switch_decision_tree::emit_case_nodes): Likewise.
8274 (conditional_probability): Remove.
8275 * tree-switch-conversion.h (enum cluster_type): New.
8277 (struct cluster): Likewise.
8278 (cluster::cluster): Likewise.
8279 (struct simple_cluster): Likewise.
8280 (simple_cluster::simple_cluster): Likewise.
8281 (struct group_cluster): Likewise.
8282 (struct jump_table_cluster): Likewise.
8283 (struct bit_test_cluster): Likewise.
8284 (struct min_cluster_item): Likewise.
8285 (struct case_tree_node): Likewise.
8286 (case_tree_node::case_tree_node): Likewise.
8287 (jump_table_cluster::case_values_threshold): Likewise.
8288 (struct case_bit_test): Likewise.
8289 (struct switch_decision_tree): Likewise.
8290 (struct switch_conversion): Likewise.
8291 (switch_decision_tree::reset_out_edges_aux): Likewise.
8293 2018-06-20 Martin Liska <mliska@suse.cz>
8295 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8296 (hoist_edge_and_branch_if_true): Likewise.
8297 (expand_switch_using_bit_tests_p): Likewise.
8298 (struct case_bit_test): Likewise.
8299 (case_bit_test_cmp): Likewise.
8300 (emit_case_bit_tests): Likewise.
8301 (switch_conversion::switch_conversion): New class.
8302 (struct switch_conv_info): Remove old struct.
8303 (collect_switch_conv_info): More to ...
8304 (switch_conversion::collect): ... this.
8305 (check_range): Likewise.
8306 (switch_conversion::check_range): Likewise.
8307 (check_all_empty_except_final): Likewise.
8308 (switch_conversion::check_all_empty_except_final): Likewise.
8309 (check_final_bb): Likewise.
8310 (switch_conversion::check_final_bb): Likewise.
8311 (create_temp_arrays): Likewise.
8312 (switch_conversion::create_temp_arrays): Likewise.
8313 (free_temp_arrays): Likewise.
8314 (gather_default_values): Likewise.
8315 (switch_conversion::gather_default_values): Likewise.
8316 (build_constructors): Likewise.
8317 (switch_conversion::build_constructors): Likewise.
8318 (constructor_contains_same_values_p): Likewise.
8319 (switch_conversion::contains_same_values_p): Likewise.
8320 (array_value_type): Likewise.
8321 (switch_conversion::array_value_type): Likewise.
8322 (build_one_array): Likewise.
8323 (switch_conversion::build_one_array): Likewise.
8324 (build_arrays): Likewise.
8325 (switch_conversion::build_arrays): Likewise.
8326 (gen_def_assigns): Likewise.
8327 (switch_conversion::gen_def_assigns): Likewise.
8328 (prune_bbs): Likewise.
8329 (switch_conversion::prune_bbs): Likewise.
8330 (fix_phi_nodes): Likewise.
8331 (switch_conversion::fix_phi_nodes): Likewise.
8332 (gen_inbound_check): Likewise.
8333 (switch_conversion::gen_inbound_check): Likewise.
8334 (process_switch): Use the newly created class.
8335 (switch_conversion::expand): New.
8336 (switch_conversion::~switch_conversion): New.
8337 * tree-switch-conversion.h: New file.
8339 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8341 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8342 tree-vect-patterns.c.
8343 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8344 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8345 (vect_recog_sad_pattern): Likewise.
8346 (vect_recog_widen_sum_pattern): Likewise.
8347 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8348 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8349 (vect_recog_rotate_pattern): Likewise.
8350 (vect_recog_mult_pattern): Likewise.
8351 (vect_recog_vector_vector_shift_pattern): Likewise.
8352 (vect_recog_divmod_pattern): Likewise.
8353 (vect_recog_mixed_size_cond_pattern): Likewise.
8354 (vect_recog_bool_pattern): Likewise.
8355 (vect_recog_mask_conversion_pattern): Likewise.
8356 (vect_try_gather_scatter_pattern): Likewise.
8357 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8358 (vect_recog_over_widening_pattern): Likewise.
8359 (vect_recog_gather_scatter_pattern): Likewise.
8360 (vect_recog_func_ptr): Move from tree-vectorizer.h
8361 (vect_vect_recog_func_ptrs): Move further down the file.
8362 (vect_recog_func): Likewise. Remove the third argument.
8363 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8364 (vect_pattern_recog_1): Expect the pattern function to do any
8365 necessary target tests. Also expect it to provide a vector type.
8366 Remove the type_in handling.
8368 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8370 * tree-vect-patterns.c (vect_pattern_detected): New function.
8371 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8372 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8373 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8374 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8375 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8376 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8377 (vect_recog_mask_conversion_pattern)
8378 (vect_try_gather_scatter_pattern): Likewise.
8380 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8382 * tree-vect-patterns.c (vect_get_internal_def): New function.
8383 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8384 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8385 (search_type_for_mask_1): Use it.
8387 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8389 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8390 redundant WIDEN_SUM_EXPR handling.
8391 (vect_recog_sad_pattern): Likewise.
8393 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8395 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8396 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8397 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8398 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8399 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8401 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8403 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8404 use the stmt_vec_info of the original bb statement for the
8405 new zero assignment, even if the call is part of a pattern.
8407 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8409 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8410 that the sequence is attached to the original statement rather
8411 than the pattern statement.
8412 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8413 PATTERN_DEF_SEQ from the original statement rather than
8414 the main pattern statement.
8415 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8416 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8417 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8419 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8421 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8422 definition statements before the early exit for statements that aren't
8424 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8426 (vect_transform_loop): ...here. Process pattern definition
8427 statements without first checking whether the main pattern
8428 statement is live or relevant.
8430 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8432 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
8433 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
8435 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
8437 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
8438 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
8439 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
8440 (expand_block_compare): Change select_block_compare_mode call.
8441 (expand_strncmp_align_check): Use new functions, fix comment.
8442 (emit_final_str_compare_gpr): New function.
8443 (expand_strn_compare): Refactor and clean up code.
8444 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
8446 2018-06-19 Tony Reix <tony.reix@atos.com>
8447 Damien Bergamini <damien.bergamini@atos.com>
8448 David Edelsohn <dje.gcc@gmail.com>
8450 * collect2.c (static_obj): New variable.
8451 (static_libs): New variable.
8452 (is_in_list): Uncomment declaration.
8453 (main): Track AIX libraries linked statically.
8454 (is_in_list): Uncomment definition.
8455 (scan_prog_file): Don't add AIX shared libraries initializer
8456 to constructor list if linking statically.
8458 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
8460 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
8462 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
8464 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
8466 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
8469 2018-06-19 Martin Liska <mliska@suse.cz>
8471 * config/i386/i386.c (ix86_can_inline_p): Do not use
8472 ipa_fn_summaries::get_create.
8473 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
8475 (devirtualization_time_bonus): Likewise.
8476 (ipcp_propagate_stage): Likewise.
8477 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8478 (edge_set_predicate): Likewise.
8479 (evaluate_conditions_for_known_args): Likewise.
8480 (evaluate_properties_for_edge): Likewise.
8481 (ipa_call_summary::reset): Tranform to ...
8482 (ipa_call_summary::~ipa_call_summary): ... this.
8483 (ipa_fn_summary::reset): Transform to ...
8484 (ipa_fn_summary::~ipa_fn_summary): ... this.
8485 (ipa_fn_summary_t::remove): Rename to ...
8486 (ipa_fn_summary_t::remove_callees): ... this.
8487 (ipa_fn_summary_t::duplicate): Use placement new
8488 instead of memory copy.
8489 (ipa_call_summary_t::duplicate): Likewise.
8490 (ipa_call_summary_t::remove): Remove.
8491 (dump_ipa_call_summary): Change get_create to get.
8492 (ipa_dump_fn_summary): Dump only when summary exists.
8493 (analyze_function_body): Use symbol_summary::get instead
8495 (compute_fn_summary): Likewise.
8496 (estimate_edge_devirt_benefit): Likewise.
8497 (estimate_edge_size_and_time): Likewise.
8498 (inline_update_callee_summaries): Likewise.
8499 (remap_edge_change_prob): Likewise.
8500 (remap_edge_summaries): Likewise.
8501 (ipa_merge_fn_summary_after_inlining): Likewise.
8502 (write_ipa_call_summary): Likewise.
8503 (ipa_fn_summary_write): Likewise.
8504 (ipa_free_fn_summary): Likewise.
8505 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
8506 (struct ipa_call_summary): Likewise.
8507 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
8509 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
8510 (estimate_size_after_inlining): Likewise.
8511 (estimate_growth): Likewise.
8512 (growth_likely_positive): Likewise.
8513 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8514 (inline_call): Likewise.
8515 * ipa-inline.c (caller_growth_limits): Likewise.
8516 (can_inline_edge_p): Likewise.
8517 (can_inline_edge_by_limits_p): Likewise.
8518 (compute_uninlined_call_time): Likewise.
8519 (compute_inlined_call_time): Likewise.
8520 (want_inline_small_function_p): Likewise.
8521 (edge_badness): Likewise.
8522 (update_caller_keys): Likewise.
8523 (update_callee_keys): Likewise.
8524 (inline_small_functions): Likewise.
8525 (inline_to_all_callers_1): Likewise.
8526 (dump_overall_stats): Likewise.
8527 (early_inline_small_functions): Likewise.
8528 (early_inliner): Likewise.
8529 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8530 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8531 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8532 * ipa-split.c (execute_split_functions): Likewise.
8533 * symbol-summary.h: Likewise.
8534 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8536 2018-06-19 Richard Biener <rguenther@suse.de>
8538 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
8539 (vectorize_loops): ... here. Fix dbgcnt handling.
8540 (try_vectorize_loop): Wrap try_vectorize_loop_1.
8542 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
8545 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
8546 ieee128 argument takes up only one (vector) register, not two (floating
8549 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8551 * gimplify.c (gimplify_init_constructor): Really never clear for an
8552 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
8554 2018-06-19 Richard Biener <rguenther@suse.de>
8556 PR tree-optimization/86179
8557 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
8558 after failed recognition.
8560 2018-06-18 Martin Sebor <msebor@redhat.com>
8563 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
8564 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
8565 Handle integer subtraction.
8566 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
8567 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
8569 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8571 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
8572 param from rtx to rtx_insn *.
8573 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
8575 (frv_ifcvt_modify_insn): Likwise.
8576 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
8577 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
8578 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
8579 as_a <rtx_insn *> cast to local "unprotected_region" once
8580 it's been established that it's not NULL or pc_rtx.
8581 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
8582 param "sethi" from rtx to rtx_insn *.
8583 (nds32_group_float_insns): Likewise for param "insn".
8584 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
8586 (vax_output_int_subtract): Likewise.
8587 * config/vax/vax.c (vax_output_int_add): Likewise for param
8589 (vax_output_int_subtract): Likewise.
8590 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
8591 (emit_pattern_after): Likewise for param "after".
8592 (emit_insn_after): Likewise.
8593 (emit_jump_insn_after): Likewise.
8594 (emit_call_insn_after): Likewise.
8595 (emit_debug_insn_after): Likewise.
8596 (emit_pattern_before): Likewise for param "before".
8597 (emit_insn_before): Likewise.
8598 (emit_jump_insn_before): Likewise.
8599 * final.c (get_insn_template): Likewise for param "insn", removing
8601 * output.h (get_insn_template): Likewise for 2nd param.
8602 * rtl.h (emit_insn_before): Likewise.
8603 (emit_jump_insn_before): Likewise.
8604 (emit_debug_insn_before_noloc): Likewise.
8605 (emit_insn_after): Likewise.
8606 (emit_jump_insn_after): Likewise.
8607 (emit_call_insn_after): Likewise.
8608 (emit_debug_insn_after): Likewise.
8609 (set_insn_deleted): Likewise for param.
8611 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
8614 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8615 floating point modes, so that IFmode is numerically greater than
8616 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8617 to declare the ordering. This prevents IFmode from being
8618 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8619 machine. Include rs6000-modes.h to share the fractional values
8620 between genmodes* and the rest of the compiler.
8624 * config/rs6000/rs6000-modes.h: New file.
8625 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8626 meaning of rs6000_long_double_size so that 126..128 selects an
8627 appropriate 128-bit floating point type.
8628 (rs6000_option_override_internal): Likewise.
8629 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8630 (TARGET_LONG_DOUBLE_128): Change the meaning of
8631 rs6000_long_double_size so that 126..128 selects an appropriate
8632 128-bit floating point type.
8633 (LONG_DOUBLE_TYPE_SIZE): Update comment.
8634 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8635 source and destination to match the standard usage.
8636 (truncifkf2): Likewise.
8637 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8638 ISA 2.07 to use an explicit clobber, instead of passing in a
8640 (copysign<mode>3_soft): Likewise.
8642 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8644 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8645 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8646 (vect_slp_analyze_instance_dependence): Likewise.
8647 (vect_enhance_data_refs_alignment): Likewise.
8648 (vect_analyze_data_refs_alignment): Likewise.
8649 (vect_slp_analyze_and_verify_instance_alignment
8650 (vect_analyze_data_ref_accesses): Likewise.
8651 (vect_prune_runtime_alias_test_list): Likewise.
8652 (vect_analyze_data_refs): Likewise.
8653 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8654 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8655 (vect_analyze_scalar_cycles_1): Likewise.
8656 (vect_get_loop_niters): Likewise.
8657 (vect_analyze_loop_form_1): Likewise.
8658 (vect_update_vf_for_slp): Likewise.
8659 (vect_analyze_loop_operations): Likewise.
8660 (vect_analyze_loop): Likewise.
8661 (vectorizable_induction): Likewise.
8662 (vect_transform_loop): Likewise.
8663 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8664 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8665 (vect_make_slp_decision): Likewise.
8666 (vect_detect_hybrid_slp): Likewise.
8667 (vect_slp_analyze_operations): Likewise.
8668 (vect_slp_bb): Likewise.
8669 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8670 (vectorizable_bswap): Likewise.
8671 (vectorizable_call): Likewise.
8672 (vectorizable_simd_clone_call): Likewise.
8673 (vectorizable_conversion): Likewise.
8674 (vectorizable_assignment): Likewise.
8675 (vectorizable_shift): Likewise.
8676 (vectorizable_operation): Likewise.
8677 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8679 2018-06-18 Martin Sebor <msebor@redhat.com>
8681 PR tree-optimization/81384
8682 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8683 * builtins.c (expand_builtin_strnlen): New function.
8684 (expand_builtin): Call it.
8685 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8686 * builtins.def (BUILT_IN_STRNLEN): New.
8687 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8688 Warn for bounds in excess of maximum object size.
8689 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8690 single-value ranges. Handle strnlen.
8691 (handle_builtin_strlen): Handle strnlen.
8692 (strlen_check_and_optimize_stmt): Same.
8693 * doc/extend.texi (Other Builtins): Document strnlen.
8695 2018-06-18 Maya Rashish <coypu@sdf.org>
8697 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8698 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8699 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8701 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8703 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8705 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8707 * tree.c (escaped_string::escape): Replace cast to char * by
8708 const_cast<char *> (unescaped).
8710 2018-06-18 Nick Clifton <nickc@redhat.com>
8713 * tree.c (escaped_string): New class. Converts an unescaped
8714 string into its escaped equivalent.
8715 (warn_deprecated_use): Use the new class to convert the
8716 deprecation message, if present.
8717 (test_escaped_strings): New self test.
8718 (test_c_tests): Add test_escaped_strings.
8719 * doc/extend.texi (deprecated): Add a note that the
8720 deprecation message is affected by the -fmessage-length
8721 option, and that control characters will be escaped.
8722 (#pragma GCC error): Document this pragma.
8723 (#pragma GCC warning): Likewise.
8724 * doc/invoke.texi (-fmessage-length): Document this option's
8725 effect on the #warning and #error preprocessor directives and
8726 the deprecated attribute.
8728 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
8730 * tree.c (decl_value_expr_lookup): Revert latest change.
8731 (decl_value_expr_insert): Likewise.
8733 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
8735 * gimplify.c (nonlocal_vlas): Delete.
8736 (nonlocal_vla_vars): Likewise.
8737 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
8739 (gimplify_body): Do not create and destroy nonlocal_vlas.
8740 * tree-nested.c: Include diagnostic.h.
8741 (use_pointer_in_frame): Tweak.
8742 (lookup_field_for_decl): Add assertion and declare the transformation.
8743 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
8744 internal error when the reference is in a wrong context. Do not
8745 create a debug decl by default.
8746 (note_nonlocal_block_vlas): Delete.
8747 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
8748 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
8749 create a debug decl by default.
8750 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
8751 call is in a wrong context.
8752 (fixup_vla_decls): New function.
8753 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
8754 debug variables were created.
8755 * tree.c (decl_value_expr_lookup): Add checking assertion.
8756 (decl_value_expr_insert): Likewise.
8758 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8761 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
8762 * tree-scalar-evolution.c (interpret_expr): Likewise.
8763 (expression_expensive_p): Likewise.
8764 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
8765 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
8766 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
8767 (ssa_defined_by_minus_one_stmt_p): New.
8769 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8772 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
8773 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
8774 * dojump.c (do_jump): Likewise.
8775 * expr.c (expand_expr_real_2): Check operand type's sign.
8776 * fold-const.c (const_unop): Handle ABSU_EXPR.
8777 (fold_abs_const): Likewise.
8778 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
8779 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
8780 (strip_sign_op_1): Likesise.
8781 * match.pd: Add new pattern to generate ABSU_EXPR.
8782 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
8783 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8784 * tree-eh.c (operation_could_trap_helper_p): Likewise.
8785 * tree-inline.c (estimate_operator_cost): Likewise.
8786 * tree-pretty-print.c (dump_generic_node): Likewise.
8787 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
8788 * tree.def (ABSU_EXPR): New.
8790 2018-06-16 Jakub Jelinek <jakub@redhat.com>
8793 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
8794 documented as preserved for backward compatibility only.
8795 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
8797 PR rtl-optimization/86108
8798 * bb-reorder.c (create_forwarder_block): Renamed to ...
8799 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
8800 jump from new landing pad to the second part.
8801 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
8804 2018-06-15 Jakub Jelinek <jakub@redhat.com>
8807 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
8808 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
8809 Only call store_expr for halves if the mode is the same.
8812 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
8813 Fix up comment formatting.
8815 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
8817 * typed-splay-tree.h (typed_splay_tree::remove): New function.
8818 (typed_splay_tree::closure,
8819 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
8820 (typed_splay_tree::typed_splay_tree,
8821 typed_splay_tree::operator =): Declared private.
8822 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
8823 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
8824 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
8825 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
8826 typed_splay_tree::splay_tree_splay,
8827 typed_splay_tree::splay_tree_foreach_helper,
8828 typed_splay_tree::splay_tree_insert,
8829 typed_splay_tree::splay_tree_remove,
8830 typed_splay_tree::splay_tree_lookup,
8831 typed_splay_tree::splay_tree_predecessor,
8832 typed_splay_tree::splay_tree_successor,
8833 typed_splay_tree::splay_tree_min,
8834 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
8835 (typed_splay_tree::root, typed_splay_tree::comp,
8836 typed_splay_tree::delete_key,
8837 typed_splay_tree::delete_value): New data members.
8838 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
8839 typed_splay_tree::remove.
8841 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
8843 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
8844 -mginv and -mno-ginv to the assembler.
8845 * config/mips/mips.opt (-mcrc): New option.
8847 * doc/invoke.text (-mcrc): Document.
8850 2018-06-15 Nick Clifton <nickc@redhat.com>
8853 * tree.c (escaped_string): New class. Converts an unescaped
8854 string into its escaped equivalent.
8855 (warn_deprecated_use): Use the new class to convert the
8856 deprecation message, if present.
8857 (test_escaped_strings): New self test.
8858 (test_c_tests): Add test_escaped_strings.
8859 * doc/extend.texi (deprecated): Add a note that the
8860 deprecation message is affected by the -fmessage-length
8861 option, and that control characters will be escaped.
8862 (#pragma GCC error): Document this pragma.
8863 (#pragma GCC warning): Likewise.
8864 * doc/invoke.texi (-fmessage-length): Document this option's
8865 effect on the #warning and #error preprocessor directives and
8866 the deprecated attribute.
8868 2018-06-15 Richard Biener <rguenther@suse.de>
8870 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
8871 here, also noting vector size used.
8872 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
8873 size used in MSG_OPTIMIZED_LOCATIONS dump.
8874 (pass_slp_vectorize::execute): Adjust.
8876 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
8879 * config/arc/arc.c (arc_return_address_register): Fix
8882 2018-06-15 Richard Biener <rguenther@suse.de>
8885 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
8886 leave useless conversion stripping to force_gimple_operand_gsi.
8887 (gimplify_build2): Likewise.
8888 (gimplify_build1): Likewise.
8890 2018-06-15 Richard Biener <rguenther@suse.de>
8893 * tree-cfg.c (move_stmt_op): unshare invariant addresses
8894 before adjusting their block.
8896 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
8898 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
8899 multilibs for *-*-rtems*.
8900 * config/riscv/t-rtems: New file.
8902 2018-06-14 Jakub Jelinek <jakub@redhat.com>
8905 * match.pd ((A +- CST1) +- CST2): Punt if last resort
8906 unsigned_type_for returns NULL.
8909 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
8910 subregs of multi-word pseudos unless the float mode has word size.
8912 2018-06-14 Richard Biener <rguenther@suse.de>
8915 * tree-vect-generic.c (build_word_mode_vector_type): Remove
8916 duplicate and harmful type_hash_canon.
8917 * tree.c (type_hash_canon): Assert we didn't find ourselves.
8919 2018-06-14 Richard Biener <rguenther@suse.de>
8922 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
8925 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
8927 * config/rtems.h (STDINT_LONG32): Define.
8929 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
8930 Prachi Godbole <prachi.godbole@imgtec.com>
8932 * config/mips/mips-cpus.def: Define P6600.
8933 * config/mips/mips-tables.opt: Regenerate.
8934 * config/mips/mips.c (mips_ucbranch_type): New enum.
8935 (mips_rtx_cost_data): Add support for P6600.
8936 (mips_issue_rate): Likewise.
8937 (mips_multipass_dfa_lookahead): Likewise.
8938 (mips_avoid_hazard): Likewise.
8939 (mips_reorg_process_insns): Likewise.
8940 (mips_classify_branch_p6600): New function.
8941 * config/mips/mips.h (TUNE_P6600): New define.
8942 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
8943 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
8944 * config/mips/mips.md: Include p6600.md.
8945 (processor): Add p6600.
8946 * config/mips/p6600.md: New file.
8947 * doc/invoke.texi: Add p6600 to supported architectures.
8949 2018-06-13 Martin Sebor <msebor@redhat.com>
8951 PR tree-optimization/86114
8952 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
8954 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
8956 2018-06-13 Richard Biener <rguenther@suse.de>
8958 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
8959 Properly set vector type of the intermediate stmt.
8960 * tree-vect-stmts.c (vectorizable_operation): The destination
8961 var always has vectype_out type.
8963 2018-06-13 Jeff Law <law@redhat.com>
8965 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
8966 integer 0 for argument to print_rtl_with_bb.
8967 (rl78_reorg): Likewise.
8969 2018-06-13 David Malcolm <dmalcolm@redhat.com>
8971 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
8972 from rtx to rtx_insn *.
8973 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
8975 (add_sched_insns_for_speculation): Likewise for local "target",
8976 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8977 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
8978 from rtx_insn ** to rtx_code_label **.
8979 (reorg_emit_nops): Likewise.
8980 (c6x_reorg): Likewise for local "call_labels".
8981 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
8983 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
8984 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
8985 the loops over LABEL_REFS.
8986 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
8988 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8989 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
8990 (split_branches): Strengthen local "olabel" from rtx to
8991 rtx_insn *, adding a safe_as_a cast.
8992 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
8994 (add_insn_after): Likewise for first two params.
8995 (add_insn_before): Likewise.
8996 (remove_insn): Likewise for param.
8997 (emit_pattern_before_noloc): Likewise for second and third params.
8998 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
8999 (emit_call_insn_before_noloc): Likewise.
9000 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9002 (emit_barrier_before): Likewise.
9003 (emit_label_before): Strengthen "label" param from "rtx" to
9004 "rtx_code_label *". Strengthen "before" param from "rtx" to
9006 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9008 (emit_pattern_after_noloc): Likewise.
9009 (emit_insn_after_noloc): Likewise.
9010 (emit_jump_insn_after_noloc): Likewise.
9011 (emit_call_insn_after_noloc): Likewise.
9012 (emit_debug_insn_after_noloc): Likewise.
9013 (emit_barrier_after): Likewise.
9014 (emit_label_after): Likewise for both params.
9015 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9016 "loc" param from "int" to "location_t".
9017 (emit_insn_after_setloc): Likewise.
9018 (emit_jump_insn_after_setloc): Likewise.
9019 (emit_call_insn_after_setloc): Likewise.
9020 (emit_debug_insn_after_setloc): Likewise.
9021 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9022 "loc" param from "int" to "location_t".
9023 (emit_pattern_before): Convert NULL_RTX to NULL.
9024 (emit_insn_before_setloc): Convert "loc" param from "int" to
9026 (emit_jump_insn_before_setloc): Likewise.
9027 (emit_call_insn_before_setloc): Likewise.
9028 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9029 rtx_insn *. Convert "loc" param from "int" to "location_t".
9030 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9031 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9032 Convert 3rd param from "int" to "location_t".
9033 (emit_barrier_before, emit_barrier_after, next_real_insn):
9034 Strengthen param from rtx to rtx_insn *.
9035 (emit_label_before): Strengthen 1st param from "rtx" to
9036 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9038 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9039 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9040 Strengthen 2nd param from "rtx" to "rtx_insn *".
9041 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9042 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9043 Likewise. Convert 3rd param from "int" to "location_t".
9044 (emit_label_after): Strengthen 1st param from "rtx" to
9046 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9048 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9049 from "rtx" to "rtx_insn *".
9051 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9053 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9054 bodies streamed in with -Q.
9055 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9056 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9057 * gimple-streamer-out.c: Include gimple-pretty-print.h
9058 (output_bb): Dump stmts streamed.
9059 * lto-section-out.c: Include print-tree.h
9060 (lto_begin_section): Dump sections created.
9061 (lto_output_decl_index): Dump decl encoded.
9062 * lto-streamer-out.c: Include print-tree.h
9063 (create_output_block): Dump output block created.
9064 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9065 (output_function): Dump function output.
9066 (output_constructor): Dump constructor streamed.
9067 (write_global_stream): Output indexes encoded.
9068 (produce_asm_for_decls): Dump streams encoded.
9069 * lto-streamer.c (streamer_dump_file): New global var.
9070 * lto-streamer.h (streamer_dump_file): Declare.
9071 * passes.c (ipa_write_summaries): Initialize streamer dump.
9072 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9075 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9078 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9079 offsets for register save directives. Emit a second batch of save
9080 directives, if need be, when the function accesses prior frames.
9082 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9084 * config/arc/fpu.md (fmasf4): Force operand to register.
9085 (fnmasf4): Likewise.
9087 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9089 * config/arc/arc-protos.h (arc_pad_return): Remove.
9090 * config/arc/arc.c (machine_function): Remove force_short_suffix
9092 (arc_print_operand): Adjust printing of '&'.
9093 (arc_verify_short): Remove conditional printing of short suffix.
9094 (arc_final_prescan_insn): Remove reference to size_reason.
9095 (pad_return): New function.
9096 (arc_reorg): Call pad_return.
9097 (arc_pad_return): Remove.
9098 (arc_init_machine_status): Remove reference to force_short_suffix.
9099 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9100 (attr length): When attribute iscompact is true force to 2
9101 regardless; in the case of maybe check if we want to force the
9102 instruction to have 4 bytes length.
9103 (nopv): Change it to generate 4 byte long nop as well.
9104 (blockage): New pattern.
9105 (simple_return): Remove call to arc_pad_return.
9106 (p_return_i): Likewise.
9108 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9110 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9112 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9114 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9117 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9119 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9120 for ARC700 and ARCv2.
9122 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9125 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9126 operands[2] instead of operands[1].
9129 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9131 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9132 case, check whether the outer register overlaps an unallocatable
9133 register, not just whether it fits the required class.
9135 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9137 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9139 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9140 (memrefs_conflict_p): Likewise.
9141 (init_alias_analysis): Likewise.
9142 * cfgexpand.c (expand_debug_expr): Likewise.
9143 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9144 * cse.c (fold_rtx): Likewise.
9145 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9146 * expr.c (emit_block_move_hints): Likewise.
9147 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9148 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9149 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9150 (emit_group_store): Likewise.
9151 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9152 to read the PRE/POST_MODIFY increment.
9153 * calls.c (store_one_arg): Use strip_offset.
9154 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9156 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9158 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9159 (simplify_binary_operation_1): Extend CONST_INT handling to
9161 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9162 than a HOST_WIDE_INT.
9163 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9165 (adjust_mems, add_stores): Update accodingly.
9166 (vt_canonicalize_addr): Track polynomial offsets.
9167 (emit_note_insn_var_location): Likewise.
9168 (vt_add_function_parameter): Likewise.
9169 (vt_initialize): Likewise.
9171 2018-06-12 Jeff Law <law@redhat.com>
9173 * config.gcc (alpha*-*-freebsd*): Remove.
9174 * config/alpha/freebsd.h: Remove.
9176 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9179 * spellcheck-tree.c (levenshtein_distance): Rename to...
9180 (get_edit_distance): ...this, and update for underlying renaming.
9181 * spellcheck-tree.h (levenshtein_distance): Rename to...
9182 (get_edit_distance): ...this.
9183 * spellcheck.c (levenshtein_distance): Rename to...
9184 (get_edit_distance): ...this. Convert from Levenshtein distance
9185 to Damerau-Levenshtein distance by supporting transpositions of
9186 adjacent characters. Rename "v1" to "v_next" and "v0" to
9188 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9189 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9190 update for underlying renaming.
9191 (selftest::levenshtein_distance_unit_test): Rename to...
9192 (selftest::test_get_edit_distance_unit): ...this, and update for
9193 underlying renaming.
9194 (selftest::test_find_closest_string): Add example from PR 69968
9195 where transposition helps
9196 (selftest::test_metric_conditions): Update for renaming.
9197 (selftest::test_metric_conditions): Likewise.
9198 (selftest::spellcheck_c_tests): Likewise.
9199 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9200 (get_edit_distance): ...this.
9201 (best_match::consider): Update for renaming.
9203 2018-06-12 Martin Sebor <msebor@redhat.com>
9205 PR tree-optimization/85259
9206 * builtins.c (compute_objsize): Handle constant offsets.
9207 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9208 true iff a warning has been issued.
9209 * gimple.h (gimple_nonartificial_location): New function.
9210 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9211 gimple_nonartificial_location and handle -Wno-system-headers.
9212 (handle_builtin_stxncpy): Same.
9214 2018-06-12 Martin Sebor <msebor@redhat.com>
9217 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9219 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9221 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9222 BUILTIN_VEC_XST entries for pointer to double and long long.
9224 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9227 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9229 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9232 2018-06-12 Martin Liska <mliska@suse.cz>
9234 * doc/options.texi: Document IntegerRange.
9236 2018-06-12 Martin Liska <mliska@suse.cz>
9238 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9239 * opt-functions.awk: Handle Deprecated flag.
9240 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9242 (read_cmdline_option): Report warning for a deprecated option.
9243 * opts.h (struct cl_option): Add new field cl_deprecated.
9244 (CL_ERR_DEPRECATED): New.
9246 2018-06-12 Martin Liska <mliska@suse.cz>
9248 * doc/options.texi: Document Deprecated option flag.
9250 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9252 * config/arc/arc-arch.h (arc_extras): New enum.
9253 (arc_cpu_t):Add field extra.
9254 (arc_cpu_types): Consider the extras.
9255 * config/arc/arc-cpus.def: Add extras info.
9256 * config/arc/arc-opts.h (processor_type): Consider extra field.
9257 * config/arc/arc.c (arc_override_options): Handle extra field.
9259 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9261 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9262 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9264 * config/arc/arc.c (arc_init): Update pic warning.
9265 (irq_range): Update irq range parsing warnings.
9266 (arc_override_options): Update various warning messages.
9267 (arc_handle_aux_attribute): Likewise.
9269 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9271 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9273 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9275 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9276 compiler messages directives.
9278 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9280 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9281 * config/mips/mips-tables.opt: Regenerate.
9282 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9284 * doc/invoke.texi: Document -march=i6500.
9286 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9288 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9289 (i6400_gpmul): Add cpu_unit.
9290 (i6400_gpdiv): Likewise.
9291 (i6400_msa_add_d): Update reservations.
9292 (i6400_msa_int_add) Likewise.
9293 (i6400_msa_short_logic3) Likewise.
9294 (i6400_msa_short_logic2) Likewise.
9295 (i6400_msa_short_logic) Likewise.
9296 (i6400_msa_move) Likewise.
9297 (i6400_msa_cmp) Likewise.
9298 (i6400_msa_short_float2) Likewise.
9299 (i6400_msa_div_d) Likewise.
9300 (i6400_msa_long_logic1) Likewise.
9301 (i6400_msa_long_logic2) Likewise.
9302 (i6400_msa_mult) Likewise.
9303 (i6400_msa_long_float2) Likewise.
9304 (i6400_msa_long_float4) Likewise.
9305 (i6400_msa_long_float5) Likewise.
9306 (i6400_msa_long_float8) Likewise.
9307 (i6400_fpu_fadd): Include frint type.
9308 (i6400_fpu_store): New define_insn_reservation.
9309 (i6400_fpu_load): Likewise.
9310 (i6400_fpu_move): Likewise.
9311 (i6400_fpu_fcmp): Likewise.
9312 (i6400_fpu_fmadd): Likewise.
9313 (i6400_int_mult): Include imul3nc type and update reservation.
9314 (i6400_int_div): Include idiv3 type and update reservation.
9315 (i6400_int_load): Update to check type not move_type.
9316 (i6400_int_store): Likewise.
9317 (i6400_int_prefetch): Set zero latency.
9319 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9321 * gcc.c: Document new %@{...} sequence.
9322 (LINK_COMMAND_SPEC): Use it for the -L switches.
9323 (cpp_unique_options): Use it for the -I switches.
9324 (at_file_argbuf): New global variable.
9325 (in_at_file): Likewise.
9326 (alloc_args): Create at_file_argbuf.
9327 (clear_args): Truncate at_file_argbuf.
9328 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9329 (open_at_file): New function.
9330 (close_at_file): Likewise.
9331 (create_at_file): Delete.
9332 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9335 (validate_switches_from_spec): Deal with %@{...} sequence.
9336 (validate_switches): Likewise.
9337 (driver::finalize): Call clear_args.
9339 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9341 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9343 2018-06-11 Martin Sebor <msebor@redhat.com>
9345 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9346 (Wno-class-memaccess): Add @opindex.
9347 (Wno-templates, Wno-multiple-inheritance): Same.
9348 (Wno-virtual-inheritance, Wno-namespaces): Same.
9349 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9350 (Wno-format-overflow, Wno-format-truncation): Same.
9351 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9352 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9353 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9354 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9355 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9356 (Wno-misspelled-isr): Same.
9358 2018-06-11 Martin Sebor <msebor@redhat.com>
9360 * PR tree-optimization/86083
9361 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9363 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9365 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9367 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9370 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9371 on the correct operand.
9372 (*movdi_internal64): Ditto.
9374 2018-06-11 Martin Liska <mliska@suse.cz>
9376 PR tree-optimization/86089
9377 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9379 2018-06-11 Julia Koval <julia.koval@intel.com>
9381 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9382 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9383 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9385 2018-06-11 Olivier Hainque <hainque@adacore.com>
9387 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9388 for Ada with strict dwarf2.
9390 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9393 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9396 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9398 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9400 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9402 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9405 2018-06-08 Martin Liska <mliska@suse.cz>
9407 * tree-cfg.h (debug_function): Fix argument type to match
9410 2018-06-08 Martin Liska <mliska@suse.cz>
9412 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9413 Remove usage of MPX-related (and removed) fields.
9414 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9416 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9418 * cfg.c (debug): Use TDF_NONE rather than 0.
9419 * cfghooks.c (debug): Likewise.
9420 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9421 (struct dump_option_value_info): Convert to...
9422 (struct kv_pair): ...this template type.
9423 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9425 (optinfo_verbosity_options): Likewise.
9426 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9428 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
9429 than int for "optgroup_flags" param.
9430 (dump_generic_expr_loc): Use dump_flags_t rather than int for
9432 (dump_dec): Likewise.
9433 (dump_finish): Use TDF_NONE rather than 0.
9434 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
9435 rather than int for "optgroup_flags" param. Use TDF_NONE rather
9436 than 0. Update for change to option_ptr.
9437 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
9438 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
9439 0. Update for changes to optinfo_verbosity_options and
9441 (opt_info_switch_p): Convert optgroup_flags from int to
9443 (dump_basic_block): Use dump_flags_t rather than int
9444 for "dump_kind" param.
9445 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
9446 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
9447 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
9448 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
9449 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
9450 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
9451 TDF_NONE): Convert from macros to...
9452 (enum dump_flag): ...this new enum.
9453 (dump_flags_t): Update to use enum.
9454 (operator|, operator&, operator~, operator|=, operator&=):
9455 Implement for dump_flags_t.
9456 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
9457 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
9458 Convert from macros to...
9459 (enum optgroup_flag): ...this new enum.
9460 (optgroup_flags_t): New typedef.
9461 (operator|, operator|=): Implement for optgroup_flags_t.
9462 (struct dump_file_info): Convert field "alt_flags" to
9463 dump_flags_t. Convert field "optgroup_flags" to
9465 (dump_basic_block): Use dump_flags_t rather than int for param.
9466 (dump_generic_expr_loc): Likewise.
9467 (dump_dec): Likewise.
9468 (dump_register): Convert param "optgroup_flags" to
9470 (opt_info_enable_passes): Likewise.
9471 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
9473 * gimple-pretty-print.c (debug): Likewise.
9474 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
9475 (merged_store_group::apply_stores): Likewise.
9476 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
9477 * gimple.c (verify_gimple_pp): Likewise.
9478 * graphite-poly.c (print_pbb_body): Likewise.
9479 * passes.c (pass_manager::register_one_dump_file): Convert
9480 local "optgroup_flags" to optgroup_flags_t.
9481 * print-tree.c (print_node): Use TDF_NONE rather than 0.
9483 (debug_body): Likewise.
9484 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
9485 to optgroup_flags_t.
9486 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
9488 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
9489 (convert_mult_to_fma): Likewise.
9490 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
9491 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
9492 * tree-vect-data-refs.c (dump_lower_bound): Convert param
9493 "dump_kind" to dump_flags_t.
9495 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9497 * config/rs6000/rs6000.c (min, max): Delete.
9499 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9501 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
9502 -mabi=spe and -mabi=no-spe.
9504 2018-06-08 Martin Liska <mliska@suse.cz>
9506 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
9507 where we expect an existing summary.
9509 2018-06-08 Martin Liska <mliska@suse.cz>
9511 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
9512 * ipa-inline.h (estimate_edge_growth): Likewise.
9514 2018-06-08 Martin Liska <mliska@suse.cz>
9516 * cgraph.c (function_version_hasher::hash): Use
9517 cgraph_node::get_uid ().
9518 (function_version_hasher::equal):
9519 * cgraph.h (cgraph_node::get_uid): New method.
9520 * ipa-inline.c (update_caller_keys): Use
9521 cgraph_node::get_uid ().
9522 (update_callee_keys): Likewise.
9523 * ipa-utils.c (searchc): Likewise.
9524 (ipa_reduced_postorder): Likewise.
9525 * lto-cgraph.c (input_node): Likewise.
9526 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
9527 * symbol-summary.h (symtab_insertion): Likewise.
9528 (symtab_removal): Likewise.
9529 (symtab_duplication): Likewise.
9530 * tree-pretty-print.c (dump_function_header): Likewise.
9531 * tree-sra.c (convert_callers_for_node): Likewise.
9533 2018-06-08 Martin Liska <mliska@suse.cz>
9535 * cgraph.c (symbol_table::create_edge): Always assign a new
9537 (symbol_table::free_edge): Do not recycle numbers.
9538 * cgraph.h (cgraph_edge::get): New method.
9539 * symbol-summary.h (symtab_removal): Use it.
9540 (symtab_duplication): Likewise.
9541 (call_summary::hashable_uid): Remove.
9543 2018-06-08 Martin Liska <mliska@suse.cz>
9545 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
9546 (initialize_growth_caches): Remove.
9547 (free_growth_caches): Likewise.
9548 (do_estimate_edge_time): Use edge_growth_cache.
9549 (do_estimate_edge_size): Likewise.
9550 (do_estimate_edge_hints): Likewise.
9551 * ipa-inline.c (reset_edge_caches): Likewise.
9552 (recursive_inlining): Likewise.
9553 (inline_small_functions): Likewise.
9554 * ipa-inline.h (initialize_growth_caches): Remove.
9555 (estimate_edge_size): Likewise.
9556 (estimate_edge_time): Likewise.
9557 (estimate_edge_hints): Likewise.
9558 (reset_edge_growth_cache): Likewise.
9559 * symbol-summary.h (call_summary::remove): New method.
9561 2018-06-08 Martin Liska <mliska@suse.cz>
9563 * ipa-cp.c (class edge_clone_summary): New summary.
9564 (grow_edge_clone_vectors): Remove.
9565 (ipcp_edge_duplication_hook): Remove.
9566 (class edge_clone_summary_t): New call_summary class.
9567 (ipcp_edge_removal_hook): Remove.
9568 (edge_clone_summary_t::duplicate): New function.
9569 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
9570 (create_specialized_node): Likewise.
9571 (ipcp_driver): Initialize edge_clone_summaries and do not
9574 2018-06-08 Martin Liska <mliska@suse.cz>
9576 * symbol-summary.h (get): New function.
9577 (call_summary::m_initialize_when_cloning): New class member.
9579 2018-06-08 Martin Liska <mliska@suse.cz>
9581 * cgraph.c (cgraph_node::remove): Do not recycle uid.
9582 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
9583 (symbol_table::allocate_cgraph_symbol): Do not set uid.
9584 * passes.c (uid_hash_t): Record removed_nodes by their uids.
9585 (remove_cgraph_node_from_order): Use the removed_nodes set.
9586 (do_per_function_toporder): Likwise.
9587 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
9588 instead of summary_uid.
9589 (symtab_removal): Likewise.
9590 (symtab_duplication): Likewise.
9592 2018-06-08 Martin Liska <mliska@suse.cz>
9594 * ipa-cp.c (ipcp_store_bits_results): Use
9595 ipcp_transformation_sum.
9596 (ipcp_store_vr_results): Likewise.
9597 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
9599 (ipcp_transformation_initialize): ... this.
9600 (ipa_set_node_agg_value_chain):
9601 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9602 (write_ipcp_transformation_info): Likewise.
9603 (read_ipcp_transformation_info): Likewise.
9604 (ipcp_update_bits): Likewise.
9605 (ipcp_update_vr): Likewise.
9606 (ipcp_transform_function): Likewise.
9607 * ipa-prop.h: Rename ipcp_transformation_summary to
9608 ipcp_transformation.
9609 (class ipcp_transformation_t): New function summary.
9610 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9611 (ipa_get_agg_replacements_for_node): Likewise.
9613 2018-06-08 Martin Liska <mliska@suse.cz>
9615 * ipa-pure-const.c (struct funct_state_d): Do it class instead
9617 (class funct_state_summary_t): New function_summary class.
9618 (has_function_state): Remove.
9619 (get_function_state): Likewise.
9620 (set_function_state): Likewise.
9621 (add_new_function): Likewise.
9622 (funct_state_summary_t::insert): New function.
9623 (duplicate_node_data): Remove.
9624 (remove_node_data): Remove.
9625 (funct_state_summary_t::duplicate): New function.
9626 (register_hooks): Create new funct_state_summaries.
9627 (pure_const_generate_summary): Use it.
9628 (pure_const_write_summary): Likewise.
9629 (pure_const_read_summary): Likewise.
9630 (propagate_pure_const): Likewise.
9631 (propagate_nothrow): Likewise.
9632 (dump_malloc_lattice): Likewise.
9633 (propagate_malloc): Likewise.
9634 (execute): Do not register hooks, just remove summary
9636 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9639 2018-06-08 Martin Liska <mliska@suse.cz>
9641 * ipa-reference.c (remove_node_data): Remove.
9642 (duplicate_node_data): Likewise.
9643 (class ipa_ref_var_info_summary_t): New class.
9644 (class ipa_ref_opt_summary_t): Likewise.
9645 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9646 (get_reference_optimization_summary): Use
9647 ipa_ref_opt_sum_summaries.
9648 (set_reference_vars_info): Remove.
9649 (set_reference_optimization_summary): Likewise.
9650 (ipa_init): Create summaries.
9651 (init_function_info): Use function summary.
9652 (ipa_ref_opt_summary_t::duplicate): New function.
9653 (ipa_ref_opt_summary_t::remove): New function.
9654 (get_read_write_all_from_node): Fix GNU coding style.
9655 (propagate): Use function summary.
9656 (write_node_summary_p): Fix GNU coding style.
9657 (stream_out_bitmap): Likewise.
9658 (ipa_reference_read_optimization_summary): Use function summary.
9659 (ipa_reference_c_finalize): Do not release hooks.
9661 2018-06-08 Martin Liska <mliska@suse.cz>
9663 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9664 (analyze_function_body): Extract multiple calls of get_create.
9665 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9666 * ipa-inline.c (recursive_inlining): Use ::get method.
9667 * ipa-inline.h (estimate_edge_growth): Likewise.
9669 2018-06-08 Martin Liska <mliska@suse.cz>
9671 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9673 (hsa_function_summary::hsa_function_summary): Use the new enum
9675 (hsa_gpu_implementation_p): Use hsa_summaries::get.
9676 * hsa-gen.c (hsa_get_host_function): Likewise.
9677 (get_brig_function_name): Likewise.
9678 * ipa-hsa.c (process_hsa_functions): Likewise.
9679 (ipa_hsa_write_summary): Likewise.
9680 * symbol-summary.h (symtab_duplication): Use ::get function/
9681 (get): New function.
9683 2018-06-08 Martin Liska <mliska@suse.cz>
9685 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9687 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9688 (hsa_register_kernel): Likewise.
9689 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9690 * hsa-gen.c (hsa_get_host_function): Likewise.
9691 (get_brig_function_name): Likewise.
9692 (generate_hsa): Likewise.
9693 (pass_gen_hsail::execute): Likewise.
9694 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9695 (devirtualization_time_bonus): Likewise.
9696 (ipcp_propagate_stage): Likewise.
9697 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9698 (edge_set_predicate): Likewise.
9699 (evaluate_conditions_for_known_args): Likewise.
9700 (evaluate_properties_for_edge): Likewise.
9701 (ipa_fn_summary::reset): Likewise.
9702 (ipa_fn_summary_t::duplicate): Likewise.
9703 (dump_ipa_call_summary): Likewise.
9704 (ipa_dump_fn_summary): Likewise.
9705 (analyze_function_body): Likewise.
9706 (compute_fn_summary): Likewise.
9707 (estimate_edge_devirt_benefit): Likewise.
9708 (estimate_edge_size_and_time): Likewise.
9709 (estimate_calls_size_and_time): Likewise.
9710 (estimate_node_size_and_time): Likewise.
9711 (inline_update_callee_summaries): Likewise.
9712 (remap_edge_change_prob): Likewise.
9713 (remap_edge_summaries): Likewise.
9714 (ipa_merge_fn_summary_after_inlining): Likewise.
9715 (ipa_update_overall_fn_summary): Likewise.
9716 (read_ipa_call_summary): Likewise.
9717 (inline_read_section): Likewise.
9718 (write_ipa_call_summary): Likewise.
9719 (ipa_fn_summary_write): Likewise.
9720 (ipa_free_fn_summary): Likewise.
9721 * ipa-hsa.c (process_hsa_functions): Likewise.
9722 (ipa_hsa_write_summary): Likewise.
9723 (ipa_hsa_read_section): Likewise.
9724 * ipa-icf.c (sem_function::merge): Likewise.
9725 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9726 (do_estimate_edge_time): Likewise.
9727 (estimate_size_after_inlining): Likewise.
9728 (estimate_growth): Likewise.
9729 (growth_likely_positive): Likewise.
9730 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9731 (inline_call): Likewise.
9732 * ipa-inline.c (caller_growth_limits): Likewise.
9733 (can_inline_edge_p): Likewise.
9734 (can_inline_edge_by_limits_p): Likewise.
9735 (compute_uninlined_call_time): Likewise.
9736 (compute_inlined_call_time): Likewise.
9737 (want_inline_small_function_p): Likewise.
9738 (edge_badness): Likewise.
9739 (update_caller_keys): Likewise.
9740 (update_callee_keys): Likewise.
9741 (recursive_inlining): Likewise.
9742 (inline_small_functions): Likewise.
9743 (inline_to_all_callers_1): Likewise.
9744 (dump_overall_stats): Likewise.
9745 (early_inline_small_functions): Likewise.
9746 (early_inliner): Likewise.
9747 * ipa-inline.h (estimate_edge_growth): Likewise.
9748 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9749 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9750 * ipa-prop.h (IPA_NODE_REF): Likewise.
9751 (IPA_EDGE_REF): Likewise.
9752 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9753 (propagate_malloc): Likewise.
9754 * ipa-split.c (execute_split_functions): Likewise.
9755 * symbol-summary.h: Rename get to get_create.
9757 (get_create): Likewise.
9758 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9760 2018-06-08 Martin Liska <mliska@suse.cz>
9762 * symbol-summary.h (release): Move definition out of class
9764 (symtab_removal): Likewise.
9765 (symtab_duplication): Likewise.
9767 2018-06-08 Martin Liska <mliska@suse.cz>
9769 * symbol-summary.h (function_summary): Move constructor
9770 implementation out of class declaration.
9771 (release): Likewise.
9772 (symtab_insertion): Likewise.
9773 (symtab_removal): Likewise.
9774 (symtab_duplication): Likewise.
9777 2018-06-08 Martin Liska <mliska@suse.cz>
9779 * Makefile.in: Remove support for MPX (macros, related functions,
9780 fields in cgraph_node, ...).
9781 * builtin-types.def (BT_BND): Likewise.
9782 (BT_FN_BND_CONST_PTR): Likewise.
9783 (BT_FN_CONST_PTR_BND): Likewise.
9784 (BT_FN_VOID_PTR_BND): Likewise.
9785 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
9786 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
9787 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
9788 (expand_builtin_mempcpy_with_bounds): Likewise.
9789 (expand_builtin_memset_with_bounds): Likewise.
9790 (expand_builtin_memset_args): Likewise.
9791 (std_expand_builtin_va_start): Likewise.
9792 (expand_builtin): Likewise.
9793 (expand_builtin_with_bounds): Likewise.
9794 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
9795 (DEF_LIB_BUILTIN_CHKP): Likewise.
9796 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
9797 (DEF_CHKP_BUILTIN): Likewise.
9798 (BUILT_IN_MEMCPY): Likewise.
9799 (BUILT_IN_MEMMOVE): Likewise.
9800 (BUILT_IN_MEMPCPY): Likewise.
9801 (BUILT_IN_MEMSET): Likewise.
9802 (BUILT_IN_STPCPY): Likewise.
9803 (BUILT_IN_STRCAT): Likewise.
9804 (BUILT_IN_STRCHR): Likewise.
9805 (BUILT_IN_STRCPY): Likewise.
9806 (BUILT_IN_STRLEN): Likewise.
9807 (BUILT_IN_MEMCPY_CHK): Likewise.
9808 (BUILT_IN_MEMMOVE_CHK): Likewise.
9809 (BUILT_IN_MEMPCPY_CHK): Likewise.
9810 (BUILT_IN_MEMSET_CHK): Likewise.
9811 (BUILT_IN_STPCPY_CHK): Likewise.
9812 (BUILT_IN_STRCAT_CHK): Likewise.
9813 (BUILT_IN_STRCPY_CHK): Likewise.
9814 * calls.c (store_bounds): Likewise.
9815 (emit_call_1): Likewise.
9816 (special_function_p): Likewise.
9817 (maybe_warn_nonstring_arg): Likewise.
9818 (initialize_argument_information): Likewise.
9819 (finalize_must_preallocate): Likewise.
9820 (compute_argument_addresses): Likewise.
9821 (expand_call): Likewise.
9822 * cfgexpand.c (expand_call_stmt): Likewise.
9823 (expand_return): Likewise.
9824 (expand_gimple_stmt_1): Likewise.
9825 (pass_expand::execute): Likewise.
9826 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
9827 (cgraph_node::remove): Likewise.
9828 (cgraph_node::dump): Likewise.
9829 (cgraph_node::verify_node): Likewise.
9830 * cgraph.h (chkp_function_instrumented_p): Likewise.
9831 (symtab_node::get_alias_target): Likewise.
9832 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
9833 (cgraph_local_p): Likewise.
9834 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
9835 (cgraph_edge::rebuild_references): Likewise.
9836 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
9837 (walk_polymorphic_call_targets): Likewise.
9838 (cgraph_node::expand_thunk): Likewise.
9839 (symbol_table::output_weakrefs): Likewise.
9840 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
9841 (ix86_handle_option): Likewise.
9842 * config/i386/constraints.md: Likewise.
9843 * config/i386/i386-builtin-types.def (BND): Likewise.
9847 * config/i386/i386-builtin.def (BDESC_END): Likewise.
9848 (BDESC_FIRST): Likewise.
9850 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
9851 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
9852 * config/i386/i386.c (enum reg_class): Likewise.
9853 (ix86_target_string): Likewise.
9854 (ix86_option_override_internal): Likewise.
9855 (ix86_conditional_register_usage): Likewise.
9856 (ix86_valid_target_attribute_inner_p): Likewise.
9857 (ix86_set_indirect_branch_type): Likewise.
9858 (ix86_set_current_function): Likewise.
9859 (ix86_function_arg_regno_p): Likewise.
9860 (init_cumulative_args): Likewise.
9861 (ix86_function_arg_advance): Likewise.
9862 (ix86_function_arg): Likewise.
9863 (ix86_pass_by_reference): Likewise.
9864 (ix86_function_value_regno_p): Likewise.
9865 (ix86_function_value_1): Likewise.
9866 (ix86_function_value_bounds): Likewise.
9867 (ix86_return_in_memory): Likewise.
9868 (ix86_setup_incoming_vararg_bounds): Likewise.
9869 (ix86_va_start): Likewise.
9870 (indirect_thunk_need_prefix): Likewise.
9871 (print_reg): Likewise.
9872 (ix86_print_operand): Likewise.
9873 (ix86_expand_call): Likewise.
9874 (ix86_output_function_return): Likewise.
9875 (reg_encoded_number): Likewise.
9876 (BDESC_VERIFYS): Likewise.
9877 (ix86_init_mpx_builtins): Likewise.
9878 (ix86_init_builtins): Likewise.
9879 (ix86_emit_cmove): Likewise.
9880 (ix86_emit_move_max): Likewise.
9881 (ix86_expand_builtin): Likewise.
9882 (ix86_builtin_mpx_function): Likewise.
9883 (ix86_get_arg_address_for_bt): Likewise.
9884 (ix86_load_bounds): Likewise.
9885 (ix86_store_bounds): Likewise.
9886 (ix86_load_returned_bounds): Likewise.
9887 (ix86_store_returned_bounds): Likewise.
9888 (ix86_class_likely_spilled_p): Likewise.
9889 (ix86_hard_regno_mode_ok): Likewise.
9890 (x86_order_regs_for_local_alloc): Likewise.
9891 (ix86_mitigate_rop): Likewise.
9892 (ix86_bnd_prefixed_insn_p): Likewise.
9893 (ix86_mpx_bound_mode): Likewise.
9894 (ix86_make_bounds_constant): Likewise.
9895 (ix86_initialize_bounds): Likewise.
9896 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
9897 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
9898 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
9899 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
9900 (TARGET_CHKP_BOUND_MODE): Likewise.
9901 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
9902 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
9903 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
9904 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
9905 * config/i386/i386.h (TARGET_MPX): Likewise.
9906 (TARGET_MPX_P): Likewise.
9907 (VALID_BND_REG_MODE): Likewise.
9908 (FIRST_BND_REG): Likewise.
9909 (LAST_BND_REG): Likewise.
9910 (enum reg_class): Likewise.
9911 (BND_REG_P): Likewise.
9912 (BND_REGNO_P): Likewise.
9913 (BNDmode): Likewise.
9914 (ADJUST_INSN_LENGTH): Likewise.
9915 * config/i386/i386.md: Likewise.
9916 * config/i386/i386.opt: Likewise.
9917 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
9918 (defined): Likewise.
9919 (LINK_MPX): Likewise.
9920 (MPX_SPEC): Likewise.
9921 (LIBMPX_SPEC): Likewise.
9922 (LIBMPXWRAPPERS_SPEC): Likewise.
9923 (CHKP_SPEC): Likewise.
9924 * config/i386/predicates.md: Likewise.
9925 * dbxout.c (dbxout_type): Likewise.
9926 * doc/extend.texi: Likewise.
9927 * doc/invoke.texi: Likewise.
9928 * doc/md.texi: Likewise.
9929 * doc/tm.texi: Likewise.
9930 * doc/tm.texi.in: Likewise.
9931 * dwarf2out.c (is_base_type): Likewise.
9932 (gen_formal_types_die): Likewise.
9933 (gen_subprogram_die): Likewise.
9934 (gen_type_die_with_usage): Likewise.
9935 (gen_decl_die): Likewise.
9936 (dwarf2out_late_global_decl): Likewise.
9937 * expr.c (expand_assignment): Likewise.
9938 (emit_storent_insn): Likewise.
9939 (store_expr_with_bounds): Likewise.
9940 (store_expr): Likewise.
9941 (expand_expr_real_1): Likewise.
9942 * expr.h (store_expr_with_bounds): Likewise.
9943 * function.c (use_register_for_decl): Likewise.
9944 (struct bounds_parm_data): Likewise.
9945 (assign_parms_augmented_arg_list): Likewise.
9946 (assign_parm_find_entry_rtl): Likewise.
9947 (assign_parm_is_stack_parm): Likewise.
9948 (assign_parm_load_bounds): Likewise.
9949 (assign_bounds): Likewise.
9950 (assign_parms): Likewise.
9951 (expand_function_start): Likewise.
9952 * gcc.c (CHKP_SPEC): Likewise.
9953 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
9954 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
9955 (wrestrict_dom_walker::check_call): Likewise.
9956 * gimple.c (gimple_build_call_from_tree): Likewise.
9957 * gimple.h (enum gf_mask): Likewise.
9958 (gimple_call_with_bounds_p): Likewise.
9959 (gimple_call_set_with_bounds): Likewise.
9960 * gimplify.c (gimplify_init_constructor): Likewise.
9961 * ipa-cp.c (initialize_node_lattices): Likewise.
9962 (propagate_constants_across_call): Likewise.
9963 (find_more_scalar_values_for_callers_subset): Likewise.
9964 * ipa-hsa.c (process_hsa_functions): Likewise.
9965 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
9966 * ipa-icf.c (sem_function::merge): Likewise.
9967 * ipa-inline.c (early_inliner): Likewise.
9968 * ipa-pure-const.c (warn_function_noreturn): Likewise.
9969 (warn_function_cold): Likewise.
9970 (propagate_pure_const): Likewise.
9971 * ipa-ref.h (enum GTY): Likewise.
9972 * ipa-split.c (find_retbnd): Likewise.
9973 (consider_split): Likewise.
9974 (split_function): Likewise.
9975 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
9976 * ipa.c (walk_polymorphic_call_targets): Likewise.
9977 (symbol_table::remove_unreachable_nodes): Likewise.
9978 (process_references): Likewise.
9979 (cgraph_build_static_cdtor_1): Likewise.
9980 * lto-cgraph.c (lto_output_node): Likewise.
9981 (output_refs): Likewise.
9982 (compute_ltrans_boundary): Likewise.
9983 (input_overwrite_node): Likewise.
9984 (input_node): Likewise.
9985 (input_cgraph_1): Likewise.
9986 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
9987 * passes.c (pass_manager::execute_early_local_passes): Likewise.
9988 (class pass_chkp_instrumentation_passes): Likewise.
9989 (make_pass_chkp_instrumentation_passes): Likewise.
9990 * passes.def: Likewise.
9991 * rtl.h (struct GTY): Likewise.
9992 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
9993 * stor-layout.c (layout_type): Likewise.
9994 * symtab.c: Likewise.
9995 * target.def: Likewise.
9996 * targhooks.c (default_chkp_bound_type): Likewise.
9997 (default_chkp_bound_mode): Likewise.
9998 (default_builtin_chkp_function): Likewise.
9999 (default_chkp_function_value_bounds): Likewise.
10000 (default_chkp_make_bounds_constant): Likewise.
10001 (default_chkp_initialize_bounds): Likewise.
10002 * targhooks.h (default_chkp_bound_type): Likewise.
10003 (default_chkp_bound_mode): Likewise.
10004 (default_builtin_chkp_function): Likewise.
10005 (default_chkp_function_value_bounds): Likewise.
10006 (default_chkp_make_bounds_constant): Likewise.
10007 (default_chkp_initialize_bounds): Likewise.
10008 * toplev.c (compile_file): Likewise.
10009 (process_options): Likewise.
10010 * tree-core.h (DEF_BUILTIN): Likewise.
10011 (DEF_BUILTIN_CHKP): Likewise.
10012 * tree-inline.c (declare_return_variable): Likewise.
10013 (remap_gimple_stmt): Likewise.
10014 (copy_bb): Likewise.
10015 (initialize_inlined_parameters): Likewise.
10016 (expand_call_inline): Likewise.
10017 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10018 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10019 (make_pass_ipa_chkp_produce_thunks): Likewise.
10020 (make_pass_chkp): Likewise.
10021 (make_pass_chkp_opt): Likewise.
10022 (make_pass_chkp_instrumentation_passes): Likewise.
10023 * tree-pretty-print.c (dump_generic_node): Likewise.
10024 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10025 * tree-ssa-dce.c (propagate_necessity): Likewise.
10026 (eliminate_unnecessary_stmts): Likewise.
10027 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10028 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10029 * tree-ssa-sccvn.h: Likewise.
10030 * tree-ssa-strlen.c (get_string_length): Likewise.
10031 (valid_builtin_call): Likewise.
10032 (adjust_last_stmt): Likewise.
10033 (handle_builtin_strchr): Likewise.
10034 (handle_builtin_strcpy): Likewise.
10035 (handle_builtin_stxncpy): Likewise.
10036 (handle_builtin_memcpy): Likewise.
10037 (handle_builtin_strcat): Likewise.
10038 (strlen_check_and_optimize_stmt): Likewise.
10039 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10040 * tree-streamer-in.c: Likewise.
10041 * tree-streamer.c (record_common_node): Likewise.
10042 * tree.c (tree_code_size): Likewise.
10043 (wide_int_to_tree_1): Likewise.
10044 (type_contains_placeholder_1): Likewise.
10045 (build_common_tree_nodes): Likewise.
10046 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10047 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10048 (POINTER_BOUNDS_P): Likewise.
10049 (BOUNDED_TYPE_P): Likewise.
10050 (BOUNDED_P): Likewise.
10051 (CALL_WITH_BOUNDS_P): Likewise.
10052 (pointer_bounds_type_node): Likewise.
10053 * value-prof.c (gimple_ic): Likewise.
10054 * var-tracking.c (vt_add_function_parameters): Likewise.
10055 * varasm.c (make_decl_rtl): Likewise.
10056 (assemble_start_function): Likewise.
10057 (output_constant): Likewise.
10058 (maybe_assemble_visibility): Likewise.
10059 * varpool.c (ctor_for_folding): Likewise.
10060 * chkp-builtins.def: Remove.
10061 * ipa-chkp.c: Remove.
10062 * ipa-chkp.h: Remove.
10063 * rtl-chkp.c: Remove.
10064 * rtl-chkp.h: Remove.
10065 * tree-chkp-opt.c: Remove.
10066 * tree-chkp.c: Remove.
10067 * tree-chkp.h: Remove.
10069 2018-06-07 Carl Love <cel@us.ibm.com>
10071 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10072 vextract_fp_from_shortl): Add BE support.
10074 2018-06-07 Paul Koning <ni1d@arrl.net>
10076 * compare-elim.c (try_merge_compare): Don't merge compare if
10077 address contains a side effect.
10078 (try_eliminate_compare): Likewise.
10080 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10082 * config.gcc: Support "tremont".
10083 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10084 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10086 * config/i386/i386.c (m_TREMONT): Define.
10087 (processor_target_table): Add "tremont".
10088 (PTA_TREMONT): Define.
10089 (ix86_lea_outperforms): Add TARGET_TREMONT.
10090 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10091 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10092 and M_INTEL_GOLDMONT_PLUS.
10093 (fold_builtin_cpu): Add "tremont".
10094 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10095 (ix86_option_override_internal): Add "tremont".
10096 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10097 (processor_type): Add PROCESSOR_TREMONT.
10098 * config/i386/x86-tune.def: Add m_TREMONT.
10099 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10101 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10103 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10104 symbol defined for msp430i* devices to be lower case.
10106 2018-06-07 Richard Biener <rguenther@suse.de>
10108 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10109 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10110 Properly wrap signed arithmetic if overflow wraps.
10112 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10114 PR tree-optimization/69615
10115 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10116 of a cast from a same precision integral SSA_NAME in a bb dominated
10117 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10118 cast to utype if rhs2 has already a compatible type.
10120 2018-06-07 Richard Biener <rguenther@suse.de>
10122 PR tree-optimization/85935
10123 * graphite-scop-detection.c (find_params_in_bb): Analyze
10124 condition operands with respect to the correct loop. Assert
10125 the analysis doesn't fail.
10127 2018-06-04 Carl Love <cel@us.ibm.com>
10129 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10130 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10131 as it is slightly cheaper.
10132 (first_match_or_eos_index_<mode>):
10133 Calculate index using natural element order.
10134 (first_match_index_<mode>):
10135 Calculate index using natural element order.
10136 (first_match_or_eos_index_<mode>):
10137 Calculate index using natural order.
10138 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10139 for BE and LE modes.
10140 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10141 P9V_BUILTIN_VCLZLSBB_V16QI.
10142 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10145 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10147 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10148 indentation and line wrap for many prototypes. Add missing
10149 @smallexample directives around block of prototypes for vec_xl and
10152 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10154 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10155 track if we pass or return IEEE 128-bit floating point.
10156 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10157 C++ mangling that is compatible with GCC 8.1.
10158 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10159 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10160 floating point types.
10161 (rs6000_function_arg_advance_1): Likewise.
10162 (rs6000_mangle_type): Optionally generate mangled names that match
10163 what GCC 8.1 generated for IEEE 128-bit floating point types.
10164 (rs6000_globalize_decl_name): If we have an external function that
10165 passes or returns IEEE 128-bit types, generate a weak reference
10166 from the mangled name used in GCC 8.1 to the current mangled
10168 (rs6000_init_builtins): Make __ibm128 use the long double type if
10169 long double is IBM extended double. Make __float128 use the long
10170 double type if long double is IEEE 128-bit.
10173 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10174 macro for __ibm128 built-in functions.
10175 (PACK_IF): Add __ibm128 pack/unpack functions.
10176 (UNPACK_IF): Likewise.
10177 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10178 enable long double built-in functions if long double is IEEE
10179 128-bit floating point.
10180 (rs6000_invalid_builtin): Update long double built-in function
10182 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10183 functions, adjust the built-in function to use the long double
10184 built-in function if __ibm128 and long double are the same type.
10185 * doc/extend.texi (PowerPC builtins): Update documention for
10186 __builtin_{,un}pack_longdouble. Add documentation for
10187 __builtin_{,un}pack_ibm128.
10189 2018-06-06 Jim Wilson <jimw@sifive.com>
10191 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10192 (struct machine_function): New field interrupt_mode.
10193 (riscv_handle_type_attribute): New function. Add forward declaration.
10194 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10195 (riscv_expand_epilogue): Check interrupt_mode field.
10196 (riscv_set_current_function): Check interrupt attribute args and
10197 set interrupt_mode field.
10198 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10199 (riscv_sret, riscv_uret): New.
10200 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10201 new arguments to interrupt attribute.
10203 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10206 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10207 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10209 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10211 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10212 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10213 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10214 several redundant entries.
10216 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10218 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10219 type from "rtx" to "rtx_insn *".
10220 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10221 for local "call_insn", removing cast.
10222 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10224 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10226 PR tree-optimization/86066
10227 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10228 for BIT_INSERT_EXPR stores.
10230 2018-06-06 Richard Biener <rguenther@suse.de>
10232 PR tree-optimization/86062
10233 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10234 component refs ontop
10235 of to be offsetted base.
10237 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10239 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10240 to be static and remove check on interrupt attribute name.
10242 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10244 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10245 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10247 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10250 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10252 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10253 Remove second argument, change how error is called.
10254 (aarch64_layout_arg): Remove second argument from
10255 aarch64_err_no_fpadvsimd call.
10256 (aarch64_init_cumulative_args): Ditto.
10257 (aarch64_gimplify_va_arg_expr): Ditto.
10258 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10260 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10262 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10263 (*simple_return_indirect_internal<mode>): Rename from
10264 simple_return_indirect_internal. Use W mode iterator.
10265 (rstorssp): New expander.
10266 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10267 (clrssbsy): New expander.
10268 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10270 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10272 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10274 (cmse_check_pointed_object): Likewise.
10276 2018-06-05 Martin Liska <mliska@suse.cz>
10278 PR gcov-profile/47618
10279 * doc/invoke.texi: Document how -fprofile-dir format
10282 2018-06-05 Richard Biener <rguenther@suse.de>
10284 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10285 removal pretend DOM info isn't available so we do not update
10286 it and only remove edges, not dominated blocks. Actually free
10287 DOM info in case we removed something. Remove unreachable blocks.
10288 (mfb_keep_latches): Work with either DOM info or marked backedges.
10289 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10290 first. Mark backedges if DOM info isn't available.
10291 (Re-)compute DOM info after cleanup_control_flow_pre.
10293 2018-06-05 Richard Biener <rguenther@suse.de>
10295 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10296 (locus_discrim_hasher::hash): Adjust.
10297 (locus_discrim_hasher::equal): Likewise.
10298 (next_discriminator_for_locus): Work on line directly.
10299 (same_line_p): Pass in expanded locus1 as well.
10300 (assign_discriminators): Avoid redundant location expansions.
10302 2018-06-05 Richard Biener <rguenther@suse.de>
10304 PR tree-optimization/86046
10305 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10306 if required after clearing TREE_ADDRESSABLE.
10308 2018-06-05 Richard Biener <rguenther@suse.de>
10310 PR tree-optimization/86047
10311 * tree-ssa-loop.c (for_each_index): Glob handling of all
10312 decls and constants and really handle all of them.
10314 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10317 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10318 qualifier_void_pointer and qualifier_const_void_pointer.
10319 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10320 (arm_init_builtins): Handle the above.
10321 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10322 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10323 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10326 2018-06-05 Martin Liska <mliska@suse.cz>
10328 * auto-profile.c (read_autofdo_file): Do not use
10329 gcov_ctr_summary struct.
10330 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10331 * coverage.c (struct counts_entry): Likewise.
10332 (read_counts_file): Read just single summary entry.
10333 (get_coverage_counts): Use gcov_summary struct.
10334 * coverage.h (get_coverage_counts): Likewise.
10335 * gcov-dump.c (dump_working_sets): Likewise.
10336 (tag_summary): Dump just single summary.
10337 * gcov-io.c (gcov_write_summary): Write just histogram
10339 (gcov_read_summary): Read just single summary.
10340 (compute_working_sets): Use gcov_summary struct.
10341 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10342 of GCOV_COUNTERS_SUMMABLE.
10343 (GCOV_COUNTERS_SUMMABLE): Remove.
10344 (GCOV_FIRST_VALUE_COUNTER): Replace with
10345 GCOV_COUNTER_V_INTERVAL.
10346 (struct gcov_ctr_summary): Remove.
10347 (struct gcov_summary): Directly use fields of former
10349 (compute_working_sets): Use gcov_summary struct.
10350 * gcov.c (read_count_file): Do not use ctrs fields.
10351 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10353 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10355 * profile.c: Likewise.
10356 * profile.h: Likewise.
10358 2018-06-05 Martin Liska <mliska@suse.cz>
10360 PR gcov-profile/84846
10361 * gcov.c (output_lines): Print working directory only
10362 in intermediate format.
10364 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10366 * config/s390/s390-builtin-types.def: Add void function type.
10367 * config/s390/s390-builtins.def: Use the function type for the
10370 2018-06-04 Jim Wilson <jimw@sifive.com>
10372 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10374 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10375 and maybe_eh_return. Change regno to unsigned int. Use new args to
10376 handle EH_RETURN_DATA_REGNO registers properly.
10377 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10378 (riscv_expand_epilogue): Update comment. Change argument name and
10379 type. Update code to use new name and type. Pass new args to
10380 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10382 * config/riscv/riscv.md (NORMAL_RETURN): New.
10383 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10384 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10385 (eh_return): Call gen_eh_return_internal and emit barrier.
10386 (eh_return_internal): Call riscv_expand_epilogue.
10388 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10390 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10391 bit_insertion field and declare can_be_merged_into method.
10392 (merged_store_group::can_be_merged_into): New method.
10393 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10394 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10395 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10397 2018-06-04 Richard Biener <rguenther@suse.de>
10399 PR tree-optimization/85955
10400 * builtins.c (fold_builtin_sincos): Convert pointers to
10401 destination to appropriate type before dereferencing.
10403 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10405 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10407 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10409 * expr.c (expand_expr_real_1): Force the operand into memory if
10410 its TYPE_MODE is BLKmode and if there is no integer mode for
10411 the number of bits being extracted.
10413 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10417 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10418 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10420 2018-06-04 Richard Biener <rguenther@suse.de>
10422 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10423 (cleanup_tree_cfg_noloop): ... single caller. Do
10424 start_recording_case_labels later.
10426 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
10428 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10429 to _IMMINTRIN_H_INCLUDED.
10430 * config/i386/pconfigintrin.h: Ditto.
10431 * config/i386/waitpkgintrin.h: Ditto.
10432 * config/i386/immintrin.h: Add includes for sgxintrin.h,
10433 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
10434 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
10435 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
10436 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
10437 waitpkgintrin.h and cldemoteintrin.h.
10439 2018-06-04 Richard Biener <rguenther@suse.de>
10441 PR tree-optimization/86038
10442 * tracer.c (find_best_successor): Check probability for
10443 being initialized, bail out if not.
10445 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
10448 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
10449 of bits to ignore when comparing architectures.
10451 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10453 PR tree-optimization/69615
10454 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
10455 maximum or minimum of the type, try to merge it also as if
10456 range1 is + [-, x - 1] or + [x + 1, -].
10459 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
10461 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
10463 PR tree-optimization/86034
10464 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
10465 the unsigned bitfield type in a bit insertion sequence if it does not
10466 have a larger precision than the bitfield size.
10467 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
10469 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
10471 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
10473 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10475 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
10476 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
10477 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
10478 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
10480 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10482 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
10483 Disable -fdelete-null-pointer-checks for ELF toolchain.
10485 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10486 Kito Cheng <kito.cheng@gmail.com>
10488 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
10489 (nds32le-*-*, nds32be-*-*): Integrate checking process.
10490 (nds32*-*-*): Add glibc and uclibc conditions.
10491 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
10492 (TARGET_EXCEPT_UNWIND_INFO): Define.
10493 * config/nds32/elf.h: New file.
10494 * config/nds32/linux.h: New file.
10495 * config/nds32/nds32-elf.opt: New file.
10496 * config/nds32/nds32-linux.opt: New file.
10497 * config/nds32/nds32-fp-as-gp.c
10498 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
10499 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
10501 (nds32_asm_file_end): Ditto.
10502 (nds32_print_operand): Ditto.
10503 (nds32_insert_attributes): Ditto.
10504 (nds32_init_libfuncs): New function.
10505 (TARGET_HAVE_TLS): Define.
10506 (TARGET_INIT_LIBFUNCS): Define.
10507 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
10509 (TARGET_ELF): Apply different mcmodel setting.
10510 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
10511 been migrated into elf.h and linux.h files.
10512 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
10513 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
10514 (mcmodel): The content has been migrated into nds32-elf.opt and
10515 nds32-linux.opt files.
10516 * config/nds32/t-elf: New file.
10517 * config/nds32/t-linux: New file.
10519 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10520 Shiva Chen <shiva0217@gmail.com>
10522 * config/nds32/constants.md (unspec_volatile_element): Add
10523 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
10524 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
10526 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
10527 (make_pass_nds32_fp_as_gp): Declare.
10528 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
10530 (nds32_asm_function_end_prologue): Remove unused asm output.
10531 (nds32_asm_function_begin_epilogue): Remove unused asm output.
10532 (nds32_asm_file_start): Output necessary fp_as_gp information.
10533 (nds32_option_override): Adjust register usage.
10534 (nds32_expand_prologue): Consider fp_as_gp situation.
10535 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
10536 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
10539 (simple_return): Ditto.
10540 (omit_fp_begin): Output special directive for fp_as_gp.
10541 (omit_fp_end): Output special directive for fp_as_gp.
10542 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
10543 mforbid-fp-as-gp): New options.
10545 2018-06-01 Mark Wielaard <mark@klomp.org>
10547 * dwarf2out.c (dwarf2out_finish): Remove generation of
10548 DW_AT_loclists_base.
10550 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
10552 * gimple-ssa-store-merging.c: Include gimple-fold.h.
10553 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
10554 (struct merged_store_group): Add bit_insertion field.
10555 (dump_char_array): Use standard hexadecimal format.
10556 (merged_store_group::merged_store_group): Set bit_insertion to false.
10557 (merged_store_group::apply_stores): Use optimal buffer size. Deal
10558 with BIT_INSERT_EXPR stores. Move up code updating the mask and
10559 also print the mask in the dump file.
10560 (pass_store_merging::gate): Minor tweak.
10561 (imm_store_chain_info::coalesce_immediate): Fix wrong association
10562 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
10563 stores with INTEGER_CST stores.
10564 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
10565 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
10566 and use it throughout. Generate bit insertion sequences if need be.
10567 (pass_store_merging::process_store): Remove redundant condition.
10568 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
10570 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
10572 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
10573 the 128-bit floating point types. Fix function comment.
10575 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10577 * config/aarch64/aarch64-simd.md
10578 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
10580 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
10583 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10585 PR tree-optimization/85989
10586 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
10588 (backprop::intersect_uses): Check it when deciding whether this
10589 is a backedge reference.
10590 (backprop::process_block): Add each phi to m_visited_phis
10591 after visiting it, then clear it at the end.
10593 2018-06-01 Richard Biener <rguenther@suse.de>
10595 * tree-vectorizer.h (vect_dr_stmt): New function.
10596 (vect_get_load_cost): Adjust.
10597 (vect_get_store_cost): Likewise.
10598 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
10599 Use vect_dr_stmt instead of DR_SMTT.
10600 (vect_record_base_alignments): Likewise.
10601 (vect_calculate_target_alignment): Likewise.
10602 (vect_compute_data_ref_alignment): Likewise and make static.
10603 (vect_update_misalignment_for_peel): Likewise.
10604 (vect_verify_datarefs_alignment): Likewise.
10605 (vector_alignment_reachable_p): Likewise.
10606 (vect_get_data_access_cost): Likewise. Pass down
10607 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10608 (vect_get_peeling_costs_all_drs): Likewise.
10609 (vect_peeling_hash_get_lowest_cost): Likewise.
10610 (vect_enhance_data_refs_alignment): Likewise.
10611 (vect_find_same_alignment_drs): Likewise.
10612 (vect_analyze_data_refs_alignment): Likewise.
10613 (vect_analyze_group_access_1): Likewise.
10614 (vect_analyze_group_access): Likewise.
10615 (vect_analyze_data_ref_access): Likewise.
10616 (vect_analyze_data_ref_accesses): Likewise.
10617 (vect_vfa_segment_size): Likewise.
10618 (vect_small_gap_p): Likewise.
10619 (vectorizable_with_step_bound_p): Likewise.
10620 (vect_prune_runtime_alias_test_list): Likewise.
10621 (vect_analyze_data_refs): Likewise.
10622 (vect_supportable_dr_alignment): Likewise.
10623 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10624 (vect_gen_prolog_loop_niters): Likewise.
10625 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10626 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10628 (vect_recog_mask_conversion_pattern): Likewise.
10629 (vect_try_gather_scatter_pattern): Likewise.
10630 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10631 to vect_get_store_cost.
10632 (vect_get_store_cost): Get stmt_info instead of DR.
10633 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10634 (vect_get_load_cost): Get stmt_info instead of DR.
10636 2018-06-01 Richard Biener <rguenther@suse.de>
10638 PR middle-end/86017
10639 * gimple-fold.c (var_decl_component_p): Also allow offsetted
10640 vars wrapped in MEM_REFs.
10642 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10644 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10645 Fix subreg tests so that we only return a choice between
10646 GENERAL_REGS and FP_REGS if the original classes included both.
10648 2018-06-01 Richard Biener <rguenther@suse.de>
10651 * tree-ssa-structalias.c (get_function_part_constraint):
10652 Handle NULL fi->decl.
10653 (find_func_aliases_for_call): Properly handle indirect
10654 fi from direct call.
10655 (find_func_clobbers): Likewise.
10656 (ipa_pta_execute): Likewise.
10657 (create_variable_info_for): For functions that are ifunc_resolver
10658 resolve to a varinfo that contains the result of the resolver call.
10659 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10662 2018-05-31 Michael Collison <michael.collison@arm.com>
10664 * config/aarch64/aarch64.md:
10665 (*fix_to_zero_extenddfdi2): New pattern.
10666 * gcc.target/aarch64/fix_extend1.c: New testcase.
10668 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
10670 PR middle-end/78809
10671 PR middle-end/83026
10672 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10673 and BUILT_IN_STRNCMP_EQ.
10674 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10675 BUILT_IN_STRNCMP_EQ.
10676 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10677 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10678 (gimple_fold_builtin): Likewise.
10679 * tree-ssa-strlen.c (compute_string_length): New function.
10680 (determine_min_obsize): New function.
10681 (handle_builtin_string_cmp): New function to handle calls to
10682 string compare functions.
10683 (strlen_optimize_stmt): Add handling to builtin string compare
10685 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10686 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10687 * tree.c (build_common_builtin_nodes): Add new defines of
10688 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10690 2018-05-31 Jakub Jelinek <jakub@redhat.com>
10693 * bb-reorder.c (pass_partition_blocks::gate): Return false for
10694 functions with naked attribute.
10696 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
10698 * config/i386/sse.md (avx_vec_concat<mode>):
10699 Substitute concat_tg_mode mode attribute with xtg_mode.
10700 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10701 (concat_tg_mode): Remove mode attribute.
10703 2018-05-31 Martin Sebor <msebor@redhat.com>
10706 * calls.c (alloc_max_size): Correct a logic error/typo.
10707 Treat excessive arguments as infinite. Warn for invalid arguments.
10708 * doc/invoke.texi (-Walloc-size-larger-than): Update.
10710 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10713 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10714 and movx for Haswell.
10716 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
10717 Cesar Philippidis <cesar@codesourcery.com>
10719 PR middle-end/85879
10720 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10721 when emitting error on private/firstprivate reductions.
10722 * omp-low.c (lower_omp_target): Avoid reference-type processing
10723 on pointers for firstprivate clause.
10725 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10727 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10730 * config/aarch64/aarch64-simd.md
10731 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
10732 (aarch64_ld1_x3_<mode>): Likewise
10733 (aarch64_st1x2<VALLDIF:mode>): Likewise
10734 (aarch64_st1_x2_<mode>): Likewise
10735 (aarch64_st1x3<VALLDIF:mode>): Likewise
10736 (aarch64_st1_x3_<mode>): Likewise
10737 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
10738 (vld1_s8_x3): Likewise.
10739 (vld1_u16_x3): Likewise.
10740 (vld1_s16_x3): Likewise.
10741 (vld1_u32_x3): Likewise.
10742 (vld1_s32_x3): Likewise.
10743 (vld1_u64_x3): Likewise.
10744 (vld1_s64_x3): Likewise.
10745 (vld1_f16_x3): Likewise.
10746 (vld1_f32_x3): Likewise.
10747 (vld1_f64_x3): Likewise.
10748 (vld1_p8_x3): Likewise.
10749 (vld1_p16_x3): Likewise.
10750 (vld1_p64_x3): Likewise.
10751 (vld1q_u8_x3): Likewise.
10752 (vld1q_s8_x3): Likewise.
10753 (vld1q_u16_x3): Likewise.
10754 (vld1q_s16_x3): Likewise.
10755 (vld1q_u32_x3): Likewise.
10756 (vld1q_s32_x3): Likewise.
10757 (vld1q_u64_x3): Likewise.
10758 (vld1q_s64_x3): Likewise.
10759 (vld1q_f16_x3): Likewise.
10760 (vld1q_f32_x3): Likewise.
10761 (vld1q_f64_x3): Likewise.
10762 (vld1q_p8_x3): Likewise.
10763 (vld1q_p16_x3): Likewise.
10764 (vld1q_p64_x3): Likewise.
10765 (vst1_s64_x2): Likewise.
10766 (vst1_u64_x2): Likewise.
10767 (vst1_f64_x2): Likewise.
10768 (vst1_s8_x2): Likewise.
10769 (vst1_p8_x2): Likewise.
10770 (vst1_s16_x2): Likewise.
10771 (vst1_p16_x2): Likewise.
10772 (vst1_s32_x2): Likewise.
10773 (vst1_u8_x2): Likewise.
10774 (vst1_u16_x2): Likewise.
10775 (vst1_u32_x2): Likewise.
10776 (vst1_f16_x2): Likewise.
10777 (vst1_f32_x2): Likewise.
10778 (vst1_p64_x2): Likewise.
10779 (vst1q_s8_x2): Likewise.
10780 (vst1q_p8_x2): Likewise.
10781 (vst1q_s16_x2): Likewise.
10782 (vst1q_p16_x2): Likewise.
10783 (vst1q_s32_x2): Likewise.
10784 (vst1q_s64_x2): Likewise.
10785 (vst1q_u8_x2): Likewise.
10786 (vst1q_u16_x2): Likewise.
10787 (vst1q_u32_x2): Likewise.
10788 (vst1q_u64_x2): Likewise.
10789 (vst1q_f16_x2): Likewise.
10790 (vst1q_f32_x2): Likewise.
10791 (vst1q_f64_x2): Likewise.
10792 (vst1q_p64_x2): Likewise.
10793 (vst1_s64_x3): Likewise.
10794 (vst1_u64_x3): Likewise.
10795 (vst1_f64_x3): Likewise.
10796 (vst1_s8_x3): Likewise.
10797 (vst1_p8_x3): Likewise.
10798 (vst1_s16_x3): Likewise.
10799 (vst1_p16_x3): Likewise.
10800 (vst1_s32_x3): Likewise.
10801 (vst1_u8_x3): Likewise.
10802 (vst1_u16_x3): Likewise.
10803 (vst1_u32_x3): Likewise.
10804 (vst1_f16_x3): Likewise.
10805 (vst1_f32_x3): Likewise.
10806 (vst1_p64_x3): Likewise.
10807 (vst1q_s8_x3): Likewise.
10808 (vst1q_p8_x3): Likewise.
10809 (vst1q_s16_x3): Likewise.
10810 (vst1q_p16_x3): Likewise.
10811 (vst1q_s32_x3): Likewise.
10812 (vst1q_s64_x3): Likewise.
10813 (vst1q_u8_x3): Likewise.
10814 (vst1q_u16_x3): Likewise.
10815 (vst1q_u32_x3): Likewise.
10816 (vst1q_u64_x3): Likewise.
10817 (vst1q_f16_x3): Likewise.
10818 (vst1q_f32_x3): Likewise.
10819 (vst1q_f64_x3): Likewise.
10820 (vst1q_p64_x3): Likewise.
10822 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10824 * config/msp430/msp430.c (msp430_output_labelref): Prepend
10825 user_label_prefix to name.
10827 * tree-core.h: Update comment about the format of NAME string
10828 passed to handler in attribute_spec.
10830 * config/msp430/msp430.md: Remove erroneous subreg expression from
10831 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
10832 zero_extend{q,h}isi2.
10834 2018-05-30 Borislav Petkov <bp@suse.de>
10836 * doc/extend.texi: Document some architecture specific
10837 constraints and sort entries.
10839 2018-05-30 Martin Sebor <msebor@redhat.com>
10841 PR middle-end/85369
10842 * builtins.c (expand_builtin_stpcpy_1): New function.
10843 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
10844 only if the former succeeds.
10846 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10848 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
10851 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10853 * doc/invoke.texi (-flinker-output): Document
10855 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10857 * passes.c (ipa_write_summaries): Only modify statements if body
10859 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
10860 incrementally linking.
10861 (ipa_passes): Likewise.
10862 * lto-cgraph.c (lto_output_node): When incrementally linking do not
10863 pass down resolution info.
10864 * common.opt (flag_incremental_link): Update info.
10865 * gcc.c (plugin specs): Turn flinker-output=* to
10866 -plugin-opt=-linker-output-known
10867 * toplev.c (compile_file): Also cut compilation when doing incremental
10869 * flag-types. (enum lto_partition_model): Add
10870 LTO_LINKER_OUTPUT_NOLTOREL.
10871 (invoke.texi): Add -flinker-output docs.
10872 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
10873 link same way as WPA; do not stream in dead initializers.
10875 * dwarf2out.c (dwarf2out_die_ref_for_decl,
10876 darf2out_register_external_decl): Support incremental link.
10878 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10880 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
10882 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10884 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
10885 it down to simple_object_copy_lto_debug_sections.
10886 (run_gcc): Determine incremental LTO link time and configure
10887 lto1 into non-wpa mode, disable renaming of debug sections.
10889 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
10891 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10892 descriptions of various incorrectly documented functions.
10894 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
10897 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10898 address check not strict.
10900 2018-05-30 Richard Biener <rguenther@suse.de>
10902 PR tree-optimization/85964
10903 * tracer.c (better_p): Drop initialized count check, we only
10904 call the function with initialized counts now.
10905 (find_best_successor): Do find a best edge if one
10906 has uninitialized count.
10907 (find_best_predecessor): Likewise. Do BB frequency check only
10908 if count is initialized.
10910 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
10912 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
10913 (aarch64_ldrstr_offset_compare): New.
10914 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
10915 load/store orderings.
10916 (aarch64_gen_adjusted_ldpstp): Likewise.
10918 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
10920 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10921 Check for subset of GENERAL_REGS and FP_REGS.
10922 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
10925 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
10927 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
10928 and wi::to_poly_offset. Add the current offset and then check
10929 whether the sum fits, rather than using an unchecked addition of
10930 a checked term. Check for a shwi rather than a uhwi.
10931 * expr.c (get_bit_range): Use tree_to_poly_uint64.
10932 (store_constructor): Use poly_int_tree_p.
10933 (expand_expr_real_1): Likewise.
10934 * function.c (assign_temp): Likewise.
10935 * fold-const.c (const_binop): Use poly_int_tree_p and
10936 wi::to_poly_offset.
10937 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
10939 * ipa-icf-gimple.c (func_checker::compare_operand): Use
10940 to_poly_offset for MEM offsets.
10941 * ipa-icf.c (sem_variable::equals): Likewise.
10942 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
10943 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
10944 wi::to_poly_offset for BIT_FIELD_REF offsets.
10945 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
10946 wi::to_poly_offset.
10947 * var-tracking.c (emit_note_insn_var_location): Use
10948 tree_to_poly_uint64.
10950 2018-05-29 Jim Wilson <jimw@sifive.com>
10952 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
10954 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
10957 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
10958 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
10960 (sse4_1_round<mode>2): Use nonimmediate_operand
10961 for operand 1 predicate.
10963 2018-05-29 Martin Sebor <msebor@redhat.com>
10964 Richard Biener <rguenther@suse.de>
10967 * calls.c (get_size_range): Call determine_value_range instead
10968 of get_value_range..
10969 * tree-vrp.h (determine_value_range): Declared new function.
10970 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
10972 2018-05-29 Richard Biener <rguenther@suse.de>
10974 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
10975 sure to use non-pattern stmts for get_earlier_stmt arguments.
10976 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
10977 called on pattern stmts.
10978 (get_later_stmt): Likewise.
10980 2018-05-29 Martin Liska <mliska@suse.cz>
10982 PR gcov-profile/85759
10983 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
10986 2018-05-29 Jakub Jelinek <jakub@redhat.com>
10988 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
10990 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
10994 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10995 VEC_PACK_FLOAT_EXPR): New tree codes.
10996 * tree-pretty-print.c (op_code_prio): Handle
10997 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10998 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10999 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11000 * tree-inline.c (estimate_operator_cost): Likewise.
11001 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11002 * fold-const.c (const_binop): Likewise.
11003 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11004 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11005 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11006 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11007 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11008 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11009 * expr.c (expand_expr_real_2): Likewise.
11010 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11011 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11012 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11014 * optabs.c (expand_widen_pattern_expr): For
11015 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11016 sign from result type rather than operand's type.
11017 (expand_binop_directly): For vec_packu_float_optab and
11018 vec_packs_float_optab allow result type to be different from operand's
11020 * optabs-tree.c (optab_for_tree_code): Handle
11021 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11022 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11023 * tree-vect-generic.c (expand_vector_operations_1): Handle
11024 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11025 VEC_PACK_FLOAT_EXPR.
11026 * tree-vect-stmts.c (supportable_widening_operation): Handle
11028 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11029 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11030 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11031 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11032 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11034 (vec_pack<floatprefix>_float_<mode>): New expander.
11035 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11037 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11038 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11039 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11040 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11041 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11043 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11044 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11045 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11046 VEC_PACK_FLOAT_EXPR): Document.
11048 2018-05-29 Richard Biener <rguenther@suse.de>
11050 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11052 (stmt_vec_info_vec): Make pointer.
11053 (init_stmt_vec_info_vec): Remove.
11054 (free_stmt_vec_info_vec): Likewise.
11055 (set_stmt_vec_info_vec): New function.
11056 (free_stmt_vec_infos): Likewise.
11057 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11058 (set_vinfo_for_stmt): Likewise.
11059 (get_earlier_stmt): Likewise.
11060 (get_later_stmt): Likewise.
11061 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11062 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11063 (vec_info::~vec_info): Free stmt_vec_infos.
11064 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11065 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11066 (pass_slp_vectorize::execute): Likewise.
11067 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11068 (free_stmt_vec_info_vec): Likewise.
11069 (set_stmt_vec_info_vec): New function.
11070 (free_stmt_vec_infos): Likewise.
11071 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11072 the global stmt_vec_info_vec.
11073 * tree-parloops.c (gather_scalar_reductions): Use
11074 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11077 2018-05-29 Richard Biener <rguenther@suse.de>
11079 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11081 2018-05-29 Martin Liska <mliska@suse.cz>
11082 David Malcolm <dmalcolm@redhat.com>
11084 * vec.c (test_reverse): New.
11085 (vec_c_tests): Add new test.
11086 * vec.h (vl_ptr>::reverse): New function.
11088 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11090 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11092 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11095 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11097 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11099 2018-05-28 Richard Biener <rguenther@suse.de>
11101 PR tree-optimization/85933
11102 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11103 look at stmts marked as vectorizable.
11105 2018-05-28 Richard Biener <rguenther@suse.de>
11107 PR tree-optimization/85934
11108 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11109 vector boolean check before scalar optimization.
11111 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11113 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11116 2018-05-28 Mark Wielaard <mark@klomp.org>
11118 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11119 if it is an expression containing a minus sign.
11121 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11123 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11125 2018-05-27 Paul Koning <ni1d@arrl.net>
11127 * config/pdp11/pdp11.md (truncsihi2): Remove.
11129 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11130 Chung-Ju Wu <jasonwucj@gmail.com>
11132 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11134 (unaligned_store_dw): Ditto.
11135 * config/nds32/nds32-memory-manipulation.c
11136 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11137 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11138 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11139 (emit_setmem_word_loop): Rename to ...
11140 (emit_setmem_doubleword_loop): ... this.
11141 (nds32_gen_dup_4_byte_to_word_value): New function.
11142 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11143 (nds32_expand_setmem_loop): Refine implementation.
11144 (nds32_expand_setmem_loop_v3m): Ditto.
11145 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11148 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11150 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11152 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11154 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11155 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11156 machine->attr_no_prologue_p.
11157 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11158 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11159 (nds32_expand_epilogue): Consider attr_naked_p.
11160 (nds32_expand_epilogue_v3pop): Likewise.
11161 (nds32_can_use_return_insn): Likewise.
11162 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11163 attr_no_prologue_p fields.
11164 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11166 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11169 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11171 * config/i386/sse.md
11172 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11174 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11176 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11178 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11180 (*<floatsuffix>floatv2div2sf2): Rename to ...
11181 (*float<floatunssuffix>v2div2sf2): ... this.
11182 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11183 (float<floatunssuffix>v2div2sf2_mask): ... this.
11184 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11185 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11186 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11188 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11190 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11192 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11194 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11196 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11198 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11199 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11200 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11201 gen_ufix_truncv8dfv8si2.
11202 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11203 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11204 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11205 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11206 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11207 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11208 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11209 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11211 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11215 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11217 2018-05-25 Jim Wilson <jimw@sifive.com>
11219 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11220 * config/riscv/riscv.c (struct machine_function): Add
11221 interrupt_handler_p and attribute_checked_p fields.
11222 (riscv_attribute_table): Add interrupt.
11223 (riscv_interrupt_type_p): New.
11224 (riscv_save_reg_p): Save extra regs for interrupt handler.
11225 (riscv_use_save_libcall): Return false for interrupt handler.
11226 (riscv_first_stack_step): Add forward declaration.
11227 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11228 for interrupt handler with large frame. Use it for saved reg list.
11229 (riscv_expand_prologue): Move flag_stack_usage_info support to
11230 eliminate duplication.
11231 (riscv_expand_epilogue): Generate mret for interrupt handler.
11232 (riscv_epilogue_uses): New.
11233 (riscv_can_use_return_insn): Return false for interrupt handler.
11234 (riscv_function_ok_for_sibcall): Likewise.
11235 (riscv_set_current_function): Add interrupt handler support.
11236 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11237 * config/riscv/riscv.md (UNSPECV_MRET): New.
11239 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11241 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11243 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11245 PR tree-optimization/85712
11246 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11247 this candidate has already been replaced in-situ by a copy.
11249 2018-05-25 Jason Merrill <jason@redhat.com>
11251 PR c++/80485 - inline function non-zero address.
11252 * symtab.c (nonzero_address): Check DECL_COMDAT.
11254 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11257 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11258 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11259 extension of SImode operation. Use const123_operand predicate.
11260 (*saddsi_1): Remove.
11261 (*saddl_se_1): Ditto.
11262 (*ssubsi_1): Ditto.
11263 (*ssubl_se_1): Ditto.
11264 * config/alpha/predicates.md (const123_operand): New predicate.
11265 * config/alpha/constraints.md (P): Use IN_RANGE.
11267 2018-05-25 Richard Biener <rguenther@suse.de>
11269 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11271 (ref_maybe_used_by_stmt_p): Likewise.
11272 (stmt_may_clobber_ref_p): Likewise.
11273 (stmt_may_clobber_ref_p_1): Likewise.
11274 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11276 (ref_maybe_used_by_stmt_p): Likewise.
11277 (stmt_may_clobber_ref_p): Likewise.
11278 (stmt_may_clobber_ref_p_1): Likewise.
11279 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11280 the alias oracle to disambiguate DRs with stmts DR analysis
11282 (vect_analyze_data_refs): Do not give up on not analyzable
11283 DRs for BB vectorization. Remove code truncating the dataref
11286 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11289 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11290 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11291 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11293 2018-05-25 Richard Biener <rguenther@suse.de>
11295 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11296 function, combining stmt data ref gathering and fatal analysis
11298 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11299 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11300 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11301 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11302 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11304 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11306 PR tree-optimization/85720
11307 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11308 SCC if all partitions are builtins.
11309 (version_loop_by_alias_check): New parameter. Generate cancelable
11310 runtime alias check if all partitions are builtins.
11311 (distribute_loop): Update call to above function.
11313 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11315 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11316 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11317 (parm_default_def_partition_arg): Ditto.
11318 (set_parm_default_def_partition): Ditto.
11319 (get_parm_default_def_partitions): Ditto and make it static.
11320 (get_undefined_value_partitions): Ditto and make it static.
11321 (remove_ssa_form): Refactor call to init_var_map here.
11322 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11323 computation for loop region.
11324 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11325 (register_default_def): Delete.
11326 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11327 (parm_default_def_partition_arg): Ditto.
11328 (set_parm_default_def_partition): Ditto.
11329 (get_parm_default_def_partitions): Ditto and make it static.
11330 (get_undefined_value_partitions): Ditto and make it static.
11331 (coalesce_with_default, coalesce_with_default): Update comment.
11332 (create_coalesce_list_for_region): New func factored out from
11333 create_outofssa_var_map.
11334 (populate_coalesce_list_for_outofssa): New func factored out from
11335 create_outofssa_var_map and coalesce_ssa_name.
11336 (create_outofssa_var_map): Delete.
11337 (coalesce_ssa_name): Refactor to support live range computation.
11338 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11339 (get_parm_default_def_partitions): Delete.
11340 (get_undefined_value_partitions): Ditto.
11341 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11342 computation for loop region.
11343 (new_tree_live_info, loe_visit_block): Ditto.
11344 (live_worklist, set_var_live_on_entry): Ditto.
11345 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11346 * tree-ssa-live.h (struct _var_map): New fields.
11347 (init_var_map): Change decl.
11348 (region_contains_p): New.
11350 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11352 * tree-ssa-live.h (live_merge_and_clear): Delete.
11354 2018-05-25 Richard Biener <rguenther@suse.de>
11357 * tree-dump.c (dequeue_and_dump): Remove access to removed
11358 operand 2 of a SWITCH_EXPR.
11360 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11362 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11363 multiplication and division.
11364 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11365 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11366 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11367 (cond_udiv_optab, cond_umod_optab): New optabs.
11368 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11369 (IFN_COND_RDIV): New internal functions.
11370 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11371 TRUNC_MOD_EXPR and RDIV_EXPR.
11372 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11373 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11375 (SVE_INT_BINARY): Include mult.
11376 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11377 (optab, sve_int_op): Handle mult.
11378 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11380 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11381 for SVE_INT_BINARY_SD.
11383 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11385 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11386 (optab, sve_int_op): Handle div and udiv.
11387 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11388 for SVE_INT_BINARY_SD.
11389 (*<optab><mode>3): New insn for the same.
11391 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11393 * tree-vect-patterns.c: Include predict.h.
11394 (vect_recog_divmod_pattern): Restrict check for division support
11395 to when optimizing for size.
11397 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11399 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11400 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11401 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11402 (gimple_match_op::set_op): Likewise.
11403 (gimple_resimplify4): Declare.
11404 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11405 (expr::gen_transform): Likewise.
11406 (decision_tree::gen): Generate a simplification routine for 4 operands.
11407 * gimple-match-head.c (gimple_simplify): Add an overload for
11408 4 operands. In the top-level function, handle up to 4 call
11409 arguments and call gimple_resimplify4.
11410 (gimple_resimplify4): New function.
11411 (build_call_internal): Pass a fourth operand.
11412 (maybe_push_to_seq): Likewise.
11413 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11414 Fold VEC_COND_EXPRs of an operation and a default value into
11415 an IFN_COND_* function if possible.
11416 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11418 (SVE_COND_FP_BINARY): Include them.
11419 (optab, sve_fp_op): Handle them.
11420 (SVE_INT_BINARY_REV): New code iterator.
11421 (SVE_COND_FP_BINARY_REV): New int iterator.
11422 (commutative): New int attribute.
11423 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11425 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11427 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11428 (*cond_<optab><mode>): New patterns for reversed operands.
11430 2018-05-25 Richard Biener <rguenther@suse.de>
11432 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
11433 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
11434 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
11435 (STMT_VINFO_GROUPED_ACCESS): Adjust.
11436 * tree-vect-data-refs.c (everywhere): Adjust users.
11437 * tree-vect-loop.c (everywhere): Likewise.
11438 * tree-vect-slp.c (everywhere): Likewise.
11439 * tree-vect-stmts.c (everywhere): Likewise.
11440 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11442 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11444 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
11446 (gcc_cv_as_section_exclude): ... this.
11447 Try Solaris as #exclude syntax.
11448 * configure: Regenerate.
11449 * config.in: Regenerate.
11450 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11452 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11453 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
11455 * varasm.c (default_elf_asm_named_section): Don't check if
11456 HAVE_GAS_SECTION_EXCLUDE is defined.
11458 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11460 * doc/md.texi: Update the documentation of the cond_* optabs
11461 to mention the new final operand. Fix GET_MODE_NUNITS call.
11462 Describe the scalar case too.
11463 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
11464 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
11466 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
11467 (get_conditional_internal_fn): Update comment.
11468 * tree-vect-loop.c (vectorizable_reduction): Pass the original
11469 accumulator value as a final argument to conditional functions.
11470 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
11471 a define_expand and add an "else" operand. Assert for now that
11472 the else operand is equal to operand 2. Use SVE_INT_BINARY and
11473 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
11474 (*cond_<optab><mode>): New patterns.
11475 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
11476 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11477 (UNSPEC_COND_EOR): Delete.
11478 (optab): Remove associated mappings.
11479 (SVE_INT_BINARY): New code iterator.
11480 (sve_int_op): Remove int attribute and add "minus" to the code
11482 (SVE_COND_INT_OP): Delete.
11483 (SVE_COND_FP_OP): Rename to...
11484 (SVE_COND_FP_BINARY): ...this.
11486 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11488 * optabs.c (can_reuse_operands_p): New function.
11489 (maybe_legitimize_operands): Try to reuse the results for
11492 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11494 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
11495 Add {q} suffix to insn mnemonic.
11497 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11499 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
11500 (msp430_warn_func_return): New.
11502 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
11504 * fold-const.c (tree_nonzero_bits): New function.
11505 * fold-const.h (tree_nonzero_bits): Likewise.
11506 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
11507 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
11509 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11513 * varasm.c (assemble_alias): Check ifunc_resolver only on
11516 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11519 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
11520 when memory input operand is handled.
11522 2018-05-24 Luis Machado <luis.machado@linaro.org>
11524 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
11526 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
11528 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11530 * match.pd: Delay FMA folds until after vectorization.
11532 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
11535 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11536 address check not strict.
11538 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11540 * gimple-match.h (gimple_match_op): New class.
11541 (mprts_hook): Replace parameters with a gimple_match_op *.
11542 (maybe_build_generic_op): Likewise.
11543 (gimple_simplified_result_is_gimple_val): Replace parameters with
11544 a const gimple_match_op *.
11545 (gimple_simplify): Replace code_helper * and tree * parameters with
11546 a gimple_match_op * parameter.
11547 (gimple_resimplify1): Replace code_helper *, tree and tree *
11548 parameters with a gimple_match_op * parameter.
11549 (gimple_resimplify2): Likewise.
11550 (gimple_resimplify3): Likewise.
11551 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
11552 parameters with a gimple_match_op * parameter.
11553 * gimple-match-head.c (gimple_simplify): Change prototypes of
11554 auto-generated functions to take a gimple_match_op * instead of
11555 separate code_helper * and tree * parameters. Make the same
11556 change in the top-level overload and update calls to the
11557 gimple_resimplify routines. Update calls to the auto-generated
11558 functions and to maybe_push_res_to_seq in the publicly-facing
11559 operation-specific gimple_simplify overloads.
11560 (gimple_match_op::MAX_NUM_OPS): Define.
11561 (gimple_resimplify1): Replace rcode and ops with a single res_op
11562 parameter. Update call to gimple_simplify.
11563 (gimple_resimplify2): Likewise.
11564 (gimple_resimplify3): Likewise.
11565 (mprts_hook): Replace parameters with a gimple_match_op *.
11566 (maybe_build_generic_op): Likewise.
11567 (build_call_internal): Replace type, nargs and ops with
11568 a gimple_match_op *.
11569 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
11570 with a single gimple_match_op *. Update calls to mprts_hook,
11571 build_call_internal and gimple_simplified_result_is_gimple_val.
11572 Factor out code that is common to the tree_code and combined_fn cases.
11573 * genmatch.c (expr::gen_transform): Replace tem_code and
11574 tem_ops with a gimple_match_op called tem_op. Update calls
11575 to the gimple_resimplify functions and maybe_push_res_to_seq.
11576 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
11577 res_ops. Update call to the gimple_resimplify functions.
11578 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
11579 (decision_tree::gen): Make the functions take a gimple_match_op *
11580 called res_op instead of separate res_code and res_ops parameters.
11581 Update call accordingly.
11582 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
11583 and ops with a single res_op parameter. Update calls to
11584 maybe_build_generic_op and maybe_push_res_to_seq.
11585 (fold_stmt_1): Update calls to gimple_simplify and
11586 replace_stmt_with_simplification.
11587 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
11588 and gimple_simplified_result_is_gimple_val.
11589 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
11591 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
11592 with a gimple_match_op *.
11593 (vn_nary_build_or_lookup): Likewise. Update call to
11594 vn_nary_build_or_lookup_1.
11595 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
11596 gimple_match_op *. Update calls to the gimple_resimplify routines
11597 and to gimple_simplified_result_is_gimple_val.
11598 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
11599 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
11600 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11601 (visit_nary_op): Likewise.
11602 (visit_reference_op_load): Likewise.
11604 2018-05-23 Luis Machado <luis.machado@linaro.org>
11606 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11607 modifier for printing the step amount.
11609 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
11612 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11615 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
11617 * doc/sourcebuild.texi (Endianness): New subsubsection.
11619 2018-05-23 Luis Machado <luis.machado@linaro.org>
11621 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11622 <prefetch_dynamic_strides>: New const bool field.
11623 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11624 prefetch_dynamic_strides.
11625 (exynosm1_prefetch_tune): Likewise.
11626 (thunderxt88_prefetch_tune): Likewise.
11627 (thunderx_prefetch_tune): Likewise.
11628 (thunderx2t99_prefetch_tune): Likewise.
11629 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11631 (aarch64_override_options_internal): Update to set
11632 PARAM_PREFETCH_DYNAMIC_STRIDES.
11633 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11634 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11635 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11636 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11637 prefetch-dynamic-strides setting.
11639 2018-05-23 Luis Machado <luis.machado@linaro.org>
11641 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11642 <minimum_stride>: New const int field.
11643 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11644 minimum_stride field defaulting to -1.
11645 (exynosm1_prefetch_tune): Likewise.
11646 (thunderxt88_prefetch_tune): Likewise.
11647 (thunderx_prefetch_tune): Likewise.
11648 (thunderx2t99_prefetch_tune): Likewise.
11649 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11650 <default_opt_level>: Set to 3.
11651 (aarch64_override_options_internal): Update to set
11652 PARAM_PREFETCH_MINIMUM_STRIDE.
11653 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11654 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11655 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11656 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11657 stride is constant and is below the minimum stride threshold.
11659 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11661 * config/arm/arm-cpus.in (mode26): Delete.
11662 (armv4): Delete mode26 reference.
11663 * config/arm/arm.c (arm_configure_build_target): Delete use of
11666 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11668 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11670 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11671 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
11673 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
11674 (floatunsdidf2): Ditto.
11676 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11678 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11679 (fixuns_trunc<mode>si2_avx512f): Ditto.
11680 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11681 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11682 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11684 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11686 PR rtl-optimization/79985
11687 * df-scan.c (df_insn_refs_collect): Remove special case for
11688 global registers and asm statements.
11690 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11692 * extend.texi (Global Register Variables): Rewrite the bullet list.
11693 Note that the register is available for allocation. Note that access
11694 via inline asm must use constraints. Add note about async-signal
11695 handlers. Remove paragraph about automagic register selection.
11697 2018-05-23 Richard Biener <rguenther@suse.de>
11699 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11700 of fixed offset from memset VN.
11702 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
11704 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11705 first_interp field.
11706 (alloc_cand_and_find_basis): Initialize first_interp field.
11707 (slsr_process_mul): Modify first_interp field.
11708 (slsr_process_add): Likewise.
11709 (slsr_process_cast): Modify first_interp field for each new
11711 (slsr_process_copy): Likewise.
11712 (dump_candidate): Dump first_interp field.
11713 (replace_mult_candidate): Process all interpretations, not just
11715 (replace_rhs_if_not_dup): Likewise.
11716 (replace_one_candidate): Likewise.
11718 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
11720 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11722 (aarch64_needs_frame_chain): New function.
11723 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11725 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11728 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11729 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11730 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11731 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
11732 as true for strict-align.
11733 (aarch64_can_inline_p): Perform checks even when callee has no
11734 attributes to check for strict alignment.
11735 * doc/extend.texi (AArch64 Function Attributes): Document
11737 * doc/invoke.texi: (AArch64 Options): Likewise.
11739 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
11741 PR tree-optimization/85853
11742 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
11743 the handling of the root of the node to...
11744 (vect_slp_analyze_node_operations_1): ...this new function,
11745 and run the whole thing with the child nodes' def types
11746 set according to their SLP node's def type.
11748 2018-05-23 Richard Biener <rguenther@suse.de>
11750 PR middle-end/85874
11751 * tree-data-ref.c (create_runtime_alias_checks): Defer
11752 and ignore overflow warnings.
11754 2018-05-23 Yury Gribov <tetra2005@gmail.com>
11756 PR tree-optimization/85822
11757 * tree-vrp.c (is_masked_range_test): Fix handling of negative
11760 2018-05-23 Richard Biener <rguenther@suse.de>
11762 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11763 memset constants via native_interpret_expr.
11765 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
11768 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
11770 (cgraph_node::create_alias): Likewise.
11771 (cgraph_node::get_availability): Check ifunc_resolver instead
11772 of looking up ifunc attribute.
11773 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11774 * varasm.c (do_assemble_alias): Likewise.
11775 (assemble_alias): Likewise.
11776 (default_binds_local_p_3): Likewise.
11777 * cgraph.h (cgraph_node): Add ifunc_resolver.
11778 (cgraph_node::only_called_directly_or_aliased_p): Return false
11779 for IFUNC resolver.
11780 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
11782 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
11783 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
11784 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
11785 instead of looking up ifunc attribute.
11787 2018-05-22 Luis Machado <luis.machado@linaro.org>
11789 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
11791 2018-05-22 Martin Sebor <msebor@redhat.com>
11793 PR middle-end/85359
11794 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
11795 only when expasion succeeds.
11796 (expand_builtin_strcmp): Same.
11797 (expand_builtin_strncmp): Same.
11799 2018-05-22 Martin Sebor <msebor@redhat.com>
11801 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
11803 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11804 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11806 * config/aarch64/aarch64-ldpstp.md: Replace uses of
11807 aarch64_mem_pair_operand with memory_operand and delete operand swapping
11809 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11810 Add check for legitimate_address.
11811 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
11812 (aarch64_swap_ldrstr_operands): New.
11813 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
11816 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11817 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11819 * config/aarch64/aarch64.md: New patterns to generate stp
11821 (store_pair_sw, store_pair_dw): New patterns to generate stp for
11822 single words and double words.
11823 (load_pair_sw, load_pair_dw): Likewise.
11824 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
11826 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
11828 * config/aarch64/aarch64-ldpstp.md: Modify peephole
11829 for different mode ldpstp and add peephole for merged zero stores.
11830 Likewise for loads.
11831 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11833 (aarch64_gen_store_pair): Rename calls to match new patterns.
11834 (aarch64_gen_load_pair): Rename calls to match new patterns.
11835 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
11836 (load_pair<DREG:mode><DREG2:mode>): ... This.
11837 (store_pair<mode>): Rename to...
11838 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
11839 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
11840 New mode iterators.
11841 (V_INT_EQUIV): Handle SImode.
11842 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
11845 2018-05-22 Martin Sebor <msebor@redhat.com>
11848 * calls.c (maybe_warn_nonstring_arg): Use string length to set
11849 or ajust the presumed bound on an operation to avoid unnecessary
11852 2018-05-22 Martin Sebor <msebor@redhat.com>
11854 PR tree-optimization/85826
11855 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
11856 assuming that a DECL necesarily has a constant size.
11858 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
11860 PR middle-end/85862
11861 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
11863 2018-05-22 Richard Biener <rguenther@suse.de>
11865 PR tree-optimization/85834
11866 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
11867 non-constant and non-zero memset arguments.
11869 2018-05-22 Martin Liska <mliska@suse.cz>
11872 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
11874 2018-05-22 Richard Biener <rguenther@suse.de>
11876 PR tree-optimization/85863
11877 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
11878 comparisons when vectype is specified.
11879 (vectorizable_condition): Do not specify vectype for
11880 vect_is_simple_cond when SLP vectorizing.
11882 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
11885 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
11886 define __ibm128 as long double.
11887 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
11888 as a distinct type when IEEE 128-bit support is enabled.
11889 (init_float128_ieee): Fix up conversions between IFmode and IEEE
11890 128-bit types to use the correct functions.
11891 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
11892 convert between 128-bit floating point types that have different
11893 modes but the same representation, instead of using gen_lowpart to
11895 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
11897 (IFKF_reg): New attributes to give the register constraints for
11899 (extend<mode>tf2_internal): New insns to mark an explicit
11900 conversion between 128-bit floating point types that have a
11901 different mode but share the same representation.
11903 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
11905 PR tree-optimization/85814
11906 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
11907 a null return from get_strinfo when unsharing the next
11908 strinfo in the chain.
11910 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11913 * varasm.c (weak_finish): Clean up weak_decls.
11915 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11917 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
11918 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
11919 UNSPEC_UADALP values.
11920 * config/aarch64/iterators.md (ABAL): New int iterator.
11923 (sur): Add mappings for the above.
11924 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
11926 (aarch64_<sur>abal<mode>_4): Likewise.
11927 (aarch64_<sur>adalp<mode>_3): Likewise.
11928 (<sur>sadv16qi): New define_expand.
11930 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
11932 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
11933 (*movdf_internal): Ditto.
11934 (*rcpsf2_sse): Ditto.
11935 (*rsqrtsf2_sse): Ditto.
11936 (*sqrt<mode>2_sse): Ditto.
11938 2018-05-21 Tamar Christina <tamar.christina@arm.com>
11940 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
11942 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
11943 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
11944 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
11945 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
11947 * config/aarch64/arm_neon.h: Likewise.
11948 * config/aarch64/iterators.md (VQ_I): New.
11950 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
11952 * config.gcc: Add arc/t-multilib-linux to tmake_file for
11954 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
11957 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11959 * config/nds32/constraints.md (S): New constraint.
11960 * config/nds32/nds32.md (call_internal): Use constraint S.
11961 (call_value_internal): Likewise.
11962 (sibcall_internal): Likewise.
11963 (sibcall_value_internal): Likewise.
11965 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11966 Chung-Ju Wu <jasonwucj@gmail.com>
11968 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
11969 into consideration.
11971 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11972 Chung-Ju Wu <jasonwucj@gmail.com>
11974 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
11975 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
11976 (nds32_rtx_costs_impl): Simplify.
11977 (nds32_address_cost_impl): Simplify.
11978 (nds32_init_rtx_costs): New function.
11979 (nds32_rtx_costs_speed_prefer): Likewise.
11980 (nds32_rtx_costs_size_prefer): Likewise.
11981 (nds32_address_cost_speed_prefer): Likewise.
11982 (nds32_address_cost_speed_fwprop): Likewise.
11983 (nds32_address_cost_size_prefer): Likewise.
11984 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
11985 * config/nds32/nds32.c (nds32_option_override): Use
11986 nds32_init_rtx_costs function.
11988 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11990 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
11991 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
11992 (TARGET_PIPELINE_N8): Likewise.
11993 (TARGET_PIPELINE_N10): Likewise.
11994 (TARGET_PIPELINE_N13): Likewise.
11995 (TARGET_PIPELINE_GRAYWOLF): Likewise.
11997 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
11999 * config/nds32/nds32-fpu.md: Update copyright year.
12001 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12003 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12005 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12007 * config/nds32/nds32.c
12008 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12009 * config/nds32/nds32.opt (minline-asm-r15): New option.
12011 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12013 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12015 * config/nds32/nds32.md (abssi2): New pattern.
12017 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12019 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12020 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12021 Merge insn pattern from sse_cvtsi2ss<round_name> and
12022 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12023 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12024 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12025 using SWI48 mode iterator.
12026 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12027 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12028 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12029 pattern from sse_cvttss2si<round_saeonly_name>
12030 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12031 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12032 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12033 using SWI48 mode iterator.
12034 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12035 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12036 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12037 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12038 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12039 using SWI48 mode iterator.
12040 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12041 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12042 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12043 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12044 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12045 SWI48 mode iterator.
12046 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12047 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12048 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12049 pattern from sse_cvttsd2si<round_saeonly_name>
12050 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12052 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12054 * config/nds32/nds32-md-auxiliary.c
12055 (nds32_valid_smw_lwm_base_p): Refine.
12056 (nds32_output_smw_single_word): Refine.
12057 (nds32_output_smw_double_word): New.
12058 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12060 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12062 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12063 (nds32_output_stack_pop): Refine.
12064 (nds32_expand_unaligned_load): Refine.
12065 (nds32_expand_unaligned_store): Refine.
12067 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12068 Chung-Ju Wu <jasonwucj@gmail.com>
12070 * config/nds32/constants.md: Add TP_REGNUM constant.
12071 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12072 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12074 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12075 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12076 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12077 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12078 and PIC code generation.
12079 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12081 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12083 * config/nds32/nds32.md: Support TLS and PIC.
12084 * config/nds32/nds32.c: Support TLS and PIC.
12085 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12086 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12089 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12091 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12092 mode with E_ prefix.
12094 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12095 Chung-Ju Wu <jasonwucj@gmail.com>
12097 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12098 * config/nds32/nds32-md-auxiliary.c
12099 (symbolic_reference_mentioned_p): New.
12100 (nds32_legitimize_ict_address): New.
12101 (nds32_expand_ict_move): New.
12102 (nds32_indirect_call_referenced_p): New.
12103 (nds32_symbol_binds_local_p): Delete.
12104 (nds32_long_call_p): Modify.
12105 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12106 * config/nds32/nds32-protos.h
12107 (symbolic_reference_mentioned_p): Declare.
12108 (nds32_legitimize_ict_address): Declare.
12109 (nds32_expand_ict_move): Declare.
12110 (nds32_indirect_call_referenced_p): Declare.
12111 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12112 (nds32_relax_group): Use nds32_ict_const_p as condition.
12113 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12114 (nds32_asm_file_start): Output ict_model directive in asm code.
12115 (nds32_legitimate_address_p): Consider indirect call.
12116 (nds32_print_operand): Consider indirect call.
12117 (nds32_print_operand_address): Consider indirect call.
12118 (nds32_insert_attributes): Handle "indirect_call" attribute.
12119 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12120 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12121 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12122 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12123 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12124 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12125 (TARGET_ICT_MODEL_SMALL): Define.
12126 (TARGET_ICT_MODEL_LARGE): Define.
12127 * config/nds32/nds32.md (movsi): Consider ict model.
12128 (call, call_value): Consider ict model.
12129 (sibcall, sibcall_value): Consider ict model.
12130 * config/nds32/nds32.opt (mict-model): New option.
12131 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12134 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12135 Monk Chiang <sh.chiang04@gmail.com>
12136 Jim Wilson <jimw@sifive.com>
12138 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12139 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12140 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12141 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12142 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12143 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12144 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12145 compute save_libcall_adjustment properly.
12146 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12147 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12148 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12149 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12150 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12151 (ABI_SPEC): Handle mabi=ilp32e.
12152 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12153 (RVE): Add RVE mask.
12154 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12155 <-march>: Add rv32e as an example.
12157 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12160 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12161 (intra_create_variable_infos): Handle C++ constructors.
12163 2018-05-18 Martin Liska <mliska@suse.cz>
12165 * passes.def: Remove a redundant pass.
12167 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12170 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12172 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12174 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12176 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12178 (armv2, armv2a, armv3, armv3m): Delete architectures.
12179 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12180 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12181 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12183 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12184 (*mulsidi3adddi): Likewise.
12185 (mulsidi3): Likewise.
12186 (*mulsidi3_nov6): Likewise.
12187 (umulsidi3): Likewise.
12188 (umulsidi3_nov6): Likewise.
12189 (umaddsidi4): Likewise.
12190 (*umulsidi3adddi): Likewise.
12191 (smulsi3_highpart): Likewise.
12192 (*smulsi3_highpart_nov6): Likewise.
12193 (umulsi3_highpart): Likewise.
12194 (*umulsi3_highpart_nov6): Likewise.
12195 * config/arm/arm.h (arm_arch3m): Delete.
12196 * config/arm/arm.c (arm_arch3m): Delete.
12197 (arm_option_override_internal): Update armv3-related comment.
12198 (arm_configure_build_target): Delete use of isa_bit_mode32.
12199 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12200 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12201 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12202 (mulsa3): Likewise.
12203 (mulusa3): Likewise.
12204 * config/arm/arm-protos.h (arm_arch3m): Delete.
12205 * config/arm/arm-tables.opt: Regenerate.
12206 * config/arm/arm-tune.md: Likewise.
12207 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12208 deleted architectures.
12210 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12212 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12213 (armv5t, armv5te): New features.
12214 (ARMv5, ARMv5e): Delete fgroups.
12215 (ARMv5t, ARMv5te): Adjust for above changes.
12216 (ARMv6m): Likewise.
12217 (armv5, armv5e): Delete arches.
12218 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12220 (*call_reg_arm): Likewise.
12221 (*call_value_reg_armv5): Likewise.
12222 (*call_value_reg_arm): Likewise.
12223 (*call_symbol): Likewise.
12224 (*call_value_symbol): Likewise.
12225 (*sibcall_insn): Likewise.
12226 (*sibcall_value_insn): Likewise.
12227 (clzsi2): Likewise.
12228 (prefetch): Likewise.
12229 (define_split and define_peephole2 dependent on arm_arch5):
12231 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12233 (TARGET_ARM_QBIT): Likewise.
12234 (TARGET_DSP_MULTIPLY): Likewise.
12235 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12236 (arm_arch5, arm_arch5e): Delete.
12237 (arm_arch5t, arm_arch5te): Declare.
12238 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12239 (arm_arch5t): Declare.
12240 (arm_option_reconfigure_globals): Update for the above.
12241 (arm_options_perform_arch_sanity_checks): Update comment, replace
12242 use of arm_arch5 with arm_arch5t.
12243 (use_return_insn): Likewise.
12244 (arm_emit_call_insn): Likewise.
12245 (output_return_instruction): Likewise.
12246 (arm_final_prescan_insn): Likewise.
12247 (arm_coproc_builtin_available): Likewise.
12248 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12249 arm_arch5e with arm_arch5t and arm_arch5te.
12250 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12251 (arm_arch5t, arm_arch5te): Declare.
12252 * config/arm/arm-tables.opt: Regenerate.
12253 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12254 * config/arm/t-multilib: Likewise.
12255 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12256 instead of arm_arch5.
12257 (*call_reg_thumb1): Likewise.
12258 (*call_value_reg_thumb1_v5): Likewise.
12259 (*call_value_reg_thumb1): Likewise.
12260 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12262 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12264 2018-05-18 Martin Liska <mliska@suse.cz>
12266 PR gcov-profile/84846
12267 * doc/gcov.texi: Document -t option of gcov tool.
12269 2018-05-18 Martin Liska <mliska@suse.cz>
12271 PR gcov-profile/84846
12272 * gcov.c (print_usage): Add new -t option.
12273 (process_args): Handle the option.
12274 (generate_results): Use stdout as output when requested by
12277 2018-05-18 Martin Liska <mliska@suse.cz>
12279 PR gcov-profile/84846
12280 * coverage.c (coverage_init): Write PWD to .gcno file.
12281 * doc/gcov.texi: Document how working directory is printed.
12282 * gcov-dump.c (dump_gcov_file): Print PWD.
12283 * gcov.c (output_intermediate_file): Likewise.
12284 (read_graph_file): Read PWD string.
12285 (output_lines): Print PWD.
12287 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12289 PR middle-end/85817
12290 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12291 for retval and return false if all args to phi are zero.
12293 2018-05-18 Richard Biener <rguenther@suse.de>
12295 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12297 (evrp_dom_walker::before_dom_children): Call it.
12299 2018-05-18 Richard Biener <rguenther@suse.de>
12301 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12302 results when processing array refs with variable index.
12304 2018-05-18 Toon Moene <toon@moene.org>
12306 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12307 directly after that of -floop-interchange. Indicate that both
12308 options are enabled by default when specifying -O3.
12310 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12312 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12313 iterator. Delete separate integer-mode vec_set<mode> expander.
12314 (aarch64_simd_vec_setv2di): Delete.
12315 (vec_setv2di): Delete.
12316 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12317 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12318 the "w, r" alternative.
12320 2018-05-18 Martin Liska <mliska@suse.cz>
12322 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12323 * tree-pass.h (make_pass_lower_switch_O0): New function.
12324 * tree-switch-conversion.c (node_has_low_bound): Remove.
12325 (node_has_high_bound): Likewise.
12326 (node_is_bounded): Likewise.
12327 (class pass_lower_switch): Make it a template type and create
12329 (pass_lower_switch::execute): Add template argument.
12330 (make_pass_lower_switch): New function.
12331 (make_pass_lower_switch_O0): New function.
12332 (do_jump_if_equal): Remove.
12333 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12334 all the hard work to tree optimization passes.
12336 2018-05-18 Martin Liska <mliska@suse.cz>
12338 * dbgcnt.c (limit_low): Renamed from limit.
12339 (limit_high): New variable.
12340 (dbg_cnt_is_enabled): Check for upper limit.
12341 (dbg_cnt): Adjust dumping.
12342 (dbg_cnt_set_limit_by_index): Add new argument for high
12344 (dbg_cnt_set_limit_by_name): Likewise.
12345 (dbg_cnt_process_single_pair): Parse new format.
12346 (dbg_cnt_process_opt): Use strtok.
12347 (dbg_cnt_list_all_counters): Remove 'value' and add
12349 * doc/invoke.texi: Document changes.
12351 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12353 * doc/sourcebuild.texi (scalar_all_fma): Document.
12354 * tree.def (FMA_EXPR): Delete.
12355 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12356 * internal-fn.c (ternary_direct): New macro.
12357 (expand_ternary_optab_fn): Likewise.
12358 (direct_ternary_optab_supported_p): Likewise.
12359 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12360 * builtins.c (fold_builtin_fma): Delete.
12361 (fold_builtin_3): Don't call it.
12362 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12363 * expr.c (expand_expr_real_2): Likewise.
12364 * fold-const.c (operand_equal_p): Likewise.
12365 (fold_ternary_loc): Likewise.
12366 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12367 * gimple.c (DEFTREECODE): Likewise.
12368 * gimplify.c (gimplify_expr): Likewise.
12369 * optabs-tree.c (optab_for_tree_code): Likewise.
12370 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12371 * tree-eh.c (operation_could_trap_p): Likewise.
12372 (stmt_could_throw_1_p): Likewise.
12373 * tree-inline.c (estimate_operator_cost): Likewise.
12374 * tree-pretty-print.c (dump_generic_node): Likewise.
12375 (op_code_prio): Likewise.
12376 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12377 * tree-ssa-operands.c (get_expr_operands): Likewise.
12378 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12379 * fold-const-call.h (fold_fma): Delete.
12380 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12381 CFN_FNMA and CFN_FNMS.
12382 (fold_fma): Delete.
12383 * genmatch.c (combined_fn): New enum.
12384 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12385 (commutative_op): New function.
12386 (commutate): Use it. Handle more than 2 operands.
12387 (dt_operand::gen_gimple_expr): Use commutative_op.
12388 (parser::parse_expr): Allow :c to be used with non-binary
12389 operators if the commutative operand is known.
12390 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12391 CFN_FMS, CFN_FNMA and CFN_FNMS.
12392 (backprop::process_assign_use): Remove FMA_EXPR handling.
12393 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12394 (gen_hsa_fma): New function.
12395 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12396 IFN_FNMA and IFN_FNMS.
12397 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12398 * gimple-fold.h (follow_all_ssa_edges): Declare.
12399 * gimple-fold.c (follow_all_ssa_edges): New function.
12400 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12401 gimple_build interface and use follow_all_ssa_edges to fold the result.
12402 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12403 instead of checking for optabs directly.
12404 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12405 rather than FMA_EXPRs.
12406 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12407 call to IFN_FMA instead of an FMA_EXPR.
12409 2018-05-17 Jim Wilson <jimw@sifive.com>
12411 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12412 sign extended promoted subreg, and the range does not have the sign bit
12413 set, then do a sign extend.
12415 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12416 test, check for sign extended subreg and/or constant operands, and
12417 do a sign extend in that case.
12419 2018-05-17 Steve Ellcey <sellcey@cavium.com>
12421 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12422 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12424 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12425 Change logics_shift_reg to logics_shift_imm.
12426 (thunderx2t99_fp_loadpair_basic): Delete.
12427 (thunderx2t99_fp_storepair_basic): Delete.
12428 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
12429 (thunderx2t99_asimd_polynomial): Delete.
12430 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
12431 and neon_fp_mul_d_scalar_q.
12432 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
12433 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
12434 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
12435 (thunderx2t99_asimd_lut): Add missing tbl types.
12436 (thunderx2t99_asimd_ext): Delete.
12437 (thunderx2t99_asimd_load1_1_mult): Delete.
12438 (thunderx2t99_asimd_load1_2_mult): Delete.
12439 (thunderx2t99_asimd_load1_ldp): New.
12440 (thunderx2t99_asimd_load1): New.
12441 (thunderx2t99_asimd_load2): Add missing *load2* types.
12442 (thunderx2t99_asimd_load3): New.
12443 (thunderx2t99_asimd_load4): New.
12444 (thunderx2t99_asimd_store1_1_mult): Delete.
12445 (thunderx2t99_asimd_store1_2_mult): Delete.
12446 (thunderx2t99_asimd_store2_mult): Delete.
12447 (thunderx2t99_asimd_store2_onelane): Delete.
12448 (thunderx2t99_asimd_store_stp): New.
12449 (thunderx2t99_asimd_store1): New.
12450 (thunderx2t99_asimd_store2): New.
12451 (thunderx2t99_asimd_store3): New.
12452 (thunderx2t99_asimd_store4): New.
12454 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
12456 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
12457 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
12459 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
12460 Segher Boessenkool <segher@kernel.crashing.org>
12463 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
12466 2018-05-17 Richard Biener <rguenther@suse.de>
12468 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
12469 for pruning loop and prune defs feeding only already visited PHIs.
12471 2018-05-17 Richard Biener <rguenther@suse.de>
12473 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
12475 2018-05-17 Bin Cheng <bin.cheng@arm.com>
12476 Richard Biener <rguenther@suse.de>
12478 PR tree-optimization/85793
12479 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
12480 for VMAT_ELEMENTWISE.
12482 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12484 * internal-fn.h (lookup_internal_fn): Declare
12485 * internal-fn.c (lookup_internal_fn): New function.
12486 * gimple.c (gimple_build_call_from_tree): Handle calls to
12487 internal functions.
12488 * gimple-pretty-print.c (dump_gimple_call): Print "." before
12489 internal function names.
12490 * tree-pretty-print.c (dump_generic_node): Likewise.
12491 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
12493 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12495 * gimple-fold.h (gimple_build): Make the function forms take
12496 combined_fn rather than built_in_function.
12497 (gimple_simplify): Likewise.
12498 * gimple-match-head.c (gimple_simplify): Likewise.
12499 * gimple-fold.c (gimple_build): Likewise.
12500 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
12501 rather than gimple_build_call_internal.
12502 (get_initial_defs_for_reduction): Likewise.
12503 (vect_create_epilog_for_reduction): Likewise.
12504 (vectorizable_live_operation): Likewise.
12506 2018-05-17 Martin Liska <mliska@suse.cz>
12508 * gimple-ssa-sprintf.c (format_directive): Do not use
12509 space in between 'G_' and '('.
12511 2018-05-17 Jakub Jelinek <jakub@redhat.com>
12514 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
12515 even if the mask is not all ones.
12518 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
12520 (ix86_gimple_fold_builtin): Likewise.
12523 * config/i386/i386.c: Include tree-vector-builder.h.
12524 (ix86_vector_shift_count): New function.
12525 (ix86_fold_builtin): Fold shift builtins by scalar count.
12526 (ix86_gimple_fold_builtin): Likewise.
12528 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
12529 _mm512_setzero): New intrinsics.
12531 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
12532 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12534 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
12535 code generation for cases where splatting a value is not useful.
12536 * simplify-rtx.c (simplify_ternary_operation): Simplify
12537 vec_merge across a vec_duplicate and a paradoxical subreg forming
12538 a vector mode to a vec_concat.
12540 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
12542 * config.gcc: Support "goldmont-plus".
12543 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12545 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12546 PROCESSOR_GOLDMONT_PLUS.
12547 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
12548 (processor_target_table): Add "goldmont-plus".
12549 (PTA_GOLDMONT_PLUS): Define.
12550 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
12551 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
12552 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
12553 (fold_builtin_cpu): Add "goldmont-plus".
12554 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
12555 (ix86_option_override_internal): Add "goldmont-plus".
12556 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
12557 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
12558 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
12559 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
12561 2018-05-17 Richard Biener <rguenther@suse.de>
12563 PR tree-optimization/85757
12564 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
12565 remove defs that only feed that PHI from further processing.
12567 2018-05-16 Jim Wilson <jimw@sifive.com>
12569 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
12571 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
12573 2018-05-16 Mark Wielaard <mark@klomp.org>
12575 * dwarf2out.c (count_index_strings): New function.
12576 (output_indirect_strings): Call count_index_strings and generate
12577 header for dwarf_version >= 5.
12579 2018-05-16 Mark Wielaard <mark@klomp.org>
12581 * dwarf2out.c (dwarf_FORM): New function.
12582 (set_indirect_string): Use dwarf_FORM.
12583 (reset_indirect_string): Likewise.
12584 (size_of_die): Likewise.
12585 (value_format): Likewise.
12586 (output_die): Likewise.
12587 (add_skeleton_AT_string): Likewise.
12588 (output_macinfo_op): Likewise.
12589 (index_string): Likewise.
12590 (output_index_string_offset): Likewise.
12591 (output_index_string): Likewise.
12592 (count_index_strings): Likewise.
12594 2018-05-16 Carl Love <cel@us.ibm.com>
12596 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12597 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
12599 2018-05-16 Martin Jambor <mjambor@suse.cz>
12601 * ipa-prop.c (ipa_free_all_edge_args): Remove.
12602 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12604 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
12606 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12607 (fnma<mode>4): Likewise.
12608 (fms<mode>4): Likewise.
12609 (fnms<mode>4): Likewise.
12610 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12611 (aarch64_fnma<mode>4): Likewise.
12612 (aarch64_fms<mode>4): Likewise.
12613 (aarch64_fnms<mode>4): Likewise.
12614 (aarch64_fnmadd<mode>4): Likewise.
12616 2018-05-16 Jason Merrill <jason@redhat.com>
12618 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
12620 2018-05-16 Richard Biener <rguenther@suse.de>
12622 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12623 (dump_stmt_cost): Declare.
12624 (add_stmt_cost): Dump cost we add.
12625 (add_stmt_costs): New function.
12626 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12627 No longer exported.
12628 (vect_analyze_stmt): Adjust prototype.
12629 (vectorizable_condition): Likewise.
12630 (vectorizable_live_operation): Likewise.
12631 (vectorizable_reduction): Likewise.
12632 (vectorizable_induction): Likewise.
12633 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12634 cost vector to pass to vectorizable_ and record afterwards.
12635 (vect_model_reduction_cost): Take cost vector argument and adjust.
12636 (vect_model_induction_cost): Likewise.
12637 (vectorizable_reduction): Likewise.
12638 (vectorizable_induction): Likewise.
12639 (vectorizable_live_operation): Likewise.
12640 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12641 SLP_TREE_NUMBER_OF_VEC_STMTS.
12642 (vect_analyze_slp_cost_1): Remove.
12643 (vect_analyze_slp_cost): Likewise.
12644 (vect_slp_analyze_node_operations): Take visited args and
12645 a target cost vector. Avoid processing already visited stmt sets.
12646 (vect_slp_analyze_operations): Use a local cost vector to gather
12647 costs and register those of non-discarded instances.
12648 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12649 (vect_schedule_slp_instance): Remove copying of
12650 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
12652 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12653 adding cost. Record cost entry location.
12654 (vect_prologue_cost_for_slp_op): Function to compute cost of
12655 a constant or invariant generated for SLP vect in the prologue,
12656 split out from vect_analyze_slp_cost_1.
12657 (vect_model_simple_cost): Make static. Adjust for SLP costing.
12658 (vect_model_promotion_demotion_cost): Likewise.
12659 (vect_model_store_cost): Likewise, make static.
12660 (vect_model_load_cost): Likewise.
12661 (vectorizable_bswap): Add cost vector arg and adjust.
12662 (vectorizable_call): Likewise.
12663 (vectorizable_simd_clone_call): Likewise.
12664 (vectorizable_conversion): Likewise.
12665 (vectorizable_assignment): Likewise.
12666 (vectorizable_shift): Likewise.
12667 (vectorizable_operation): Likewise.
12668 (vectorizable_store): Likewise.
12669 (vectorizable_load): Likewise.
12670 (vectorizable_condition): Likewise.
12671 (vectorizable_comparison): Likewise.
12672 (can_vectorize_live_stmts): Likewise.
12673 (vect_analyze_stmt): Likewise.
12674 (vect_transform_stmt): Adjust calls to vectorizable_*.
12675 * tree-vectorizer.c: Include gimple-pretty-print.h.
12676 (dump_stmt_cost): New function.
12678 2018-05-16 Richard Biener <rguenther@suse.de>
12680 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12681 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12682 * tree-ssa-dse.c: Include tree-ssa-loop.h.
12683 (check_name): New callback.
12684 (dse_classify_store): Track cycles via a visited bitmap of PHI
12685 defs and simplify handling of in-loop and across loop dead stores
12686 and properly fail for loop-variant refs. Handle byte-tracking with
12687 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12690 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
12692 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12693 (vect_get_mask_type_for_stmt): Likewise.
12694 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12696 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
12697 to determine the statement's vector type and the vector type that
12698 should be used for calculating nunits. Deal with cases in which
12699 the type has to be deferred.
12700 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12701 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12702 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12703 (vect_determine_vf_for_stmt): New functions, split out from...
12704 (vect_determine_vectorization_factor): ...here.
12705 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12706 (vect_get_mask_type_for_stmt): New functions, split out from
12707 vect_determine_vectorization_factor.
12709 2018-05-16 Richard Biener <rguenther@suse.de>
12711 * tree-cfg.c (verify_gimple_assign_ternary): Properly
12712 verify the [VEC_]COND_EXPR embedded comparison.
12714 2018-05-15 Martin Sebor <msebor@redhat.com>
12716 PR tree-optimization/85753
12717 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12718 RECORD_TYPE in addition to ARRAY_TYPE.
12720 2018-05-15 Martin Sebor <msebor@redhat.com>
12722 PR middle-end/85643
12723 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12725 2018-05-15 Richard Biener <rguenther@suse.de>
12727 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12728 add by_clobber_p one. Change algorithm to collect all defs
12729 representing uses we need to walk and try reducing them to
12730 a single one before failing.
12731 (dse_dom_walker::dse_optimize_stmt): Adjust.
12733 2018-05-13 Mark Wielaard <mark@klomp.org>
12735 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12736 (size_of_loc_descr): Likewise.
12737 (output_loc_operands): Likewise.
12738 (output_loc_operands_raw): Likewise.
12739 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
12740 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
12741 (hash_loc_operands): Likewise.
12742 (compare_loc_operands): Likewise.
12744 2018-05-14 Mark Wielaard <mark@klomp.org>
12746 * dwarf2out.c (count_index_addrs): New function.
12747 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
12749 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12751 PR tree-optimization/83648
12752 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
12753 return value as malloc candidate.
12755 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12758 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
12759 param as true in call to suggest_attribute.
12761 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
12763 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
12764 -mreadonly-in-sdata.
12766 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12768 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
12770 (aarch64_crypto_aesd_fused): Likewise.
12772 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
12774 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
12775 (movsi_aarch64): Likewise.
12776 (load_pairsi): Likewise.
12777 (load_pairdi): Likewise.
12778 (store_pairsi): Likewise.
12779 (store_pairdi): Likewise.
12780 (load_pairsf): Likewise.
12781 (load_pairdf): Likewise.
12782 (store_pairsf): Likewise.
12783 (store_pairdf): Likewise.
12784 (zero_extend): Likewise.
12785 (trunc): Swap alternatives.
12786 (fcvt_target): Add '?' to prefer w over r.
12788 2018-05-14 Jakub Jelinek <jakub@redhat.com>
12791 * config/i386/i386.md: Disallow non-commutative arithmetics in
12792 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
12793 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
12794 in the peephole2 before it.
12796 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
12798 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
12799 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
12800 (ix86_handle_option): Handle -mcldemote.
12801 * config.gcc: New header.
12802 * config/i386/cldemoteintrin.h: New file.
12803 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
12804 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12806 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12807 OPTION_MASK_ISA_CLDEMOTE.
12808 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
12809 (ix86_valid_target_attribute_inner_p): Ditto.
12810 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
12811 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
12812 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
12813 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
12814 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
12816 * config/i386/i386.opt: Add -mcldemote.
12817 * config/i386/x86intrin.h: New header.
12818 * doc/invoke.texi: Add -mcldemote.
12820 2018-05-14 Richard Biener <rguenther@suse.de>
12822 * doc/match-and-simplify.texi: Adjust :s documentation.
12824 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
12826 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
12827 intended memcpy size.
12828 (REORDER_45): Likewise.
12830 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
12832 * sort.cc: New file.
12833 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
12834 * vec.c (qsort_chk): Use gcc_qsort.
12835 * Makefile.in (OBJS-libcommon): Add sort.o.
12836 (build/sort.o): New target. Use it...
12837 (BUILD_RTL): ... here, and...
12838 (build/gencfn-macros): ... here, and...
12839 (build/genmatch): ... here.
12841 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12842 Chung-Ju Wu <jasonwucj@gmail.com>
12844 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
12845 * config/nds32/nds32-graywolf.md: New file.
12846 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
12847 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
12849 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
12850 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
12851 * config/nds32/nds32.md (pipeline_model): Add graywolf.
12852 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
12853 * config/nds32/pipelines.md: Include n15 settings.
12855 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12856 Chung-Ju Wu <jasonwucj@gmail.com>
12858 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
12859 * config/nds32/nds32-n13.md: New file.
12860 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
12861 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
12863 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
12864 * config/nds32/nds32.md (pipeline_model): Add n13.
12865 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
12866 * config/nds32/pipelines.md: Include n13 settings.
12868 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12869 Chung-Ju Wu <jasonwucj@gmail.com>
12871 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
12872 * config/nds32/nds32-n10.md: New file.
12873 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
12874 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
12876 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
12877 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
12878 * config/nds32/nds32.md (pipeline_model): Add n10.
12879 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
12880 * config/nds32/pipelines.md: Include n10 settings.
12882 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
12883 Kito Cheng <kito.cheng@gmail.com>
12884 Chung-Ju Wu <jasonwucj@gmail.com>
12886 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
12887 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12888 Add enum values for DSP extension instructions.
12889 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
12891 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
12892 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
12893 New code iterators.
12894 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
12895 * config/nds32/nds32-dspext.md: New file for DSP implementation.
12896 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
12897 * config/nds32/nds32-intrinsic.md: Likewise.
12898 * config/nds32/nds32_intrinsic.h: Likewise.
12899 * config/nds32/nds32-md-auxiliary.c: Likewise.
12900 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
12901 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
12902 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
12903 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
12904 * config/nds32/nds32-protos.h: New declarations for DSP extension.
12905 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
12906 TYPE_DMAC in switch statement.
12907 * config/nds32/nds32.c: New checking and implementation for DSP
12908 extension instructions.
12909 * config/nds32/nds32.h: Likewise.
12910 * config/nds32/nds32.md: Likewise.
12911 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
12912 * config/nds32/predicates.md: Implement new predicates for DSP
12915 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12917 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
12918 Reformat alternatives and attributes so it is easier to identify
12919 which constraints/attributes go with which instruction.
12920 (mov<mode>_hardfloat32, FMOVE64): Likewise.
12921 (mov<mode>_softfloat32, FMOVE64): Likewise.
12922 (mov<mode>_hardfloat64, FMOVE64): Likewise.
12923 (mov<mode>_softfloat64, FMOVE64): Likewise.
12925 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
12927 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12929 (Basic PowerPC Built-in Functions): The new name of the
12930 subsection previously known as "PowerPC Built-in Functions".
12931 (Basic PowerPC Built-in Functions Available on all Configurations):
12933 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
12934 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
12935 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
12936 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
12938 2018-05-11 Martin Jambor <mjambor@suse.cz>
12941 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
12944 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12947 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
12949 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
12951 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
12952 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
12953 (ix86_handle_option): Handle -mwaitpkg.
12954 * config.gcc: New header.
12955 * config/i386/cpuid.h (bit_WAITPKG): New bit.
12956 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
12957 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
12959 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12960 OPTION_MASK_ISA_WAITPKG.
12961 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
12962 (ix86_option_override_internal): Add PTA_WAITPKG.
12963 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
12964 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
12965 IX86_BUILTIN_TPAUSE.
12966 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
12967 __builtin_ia32_umwait and __builtin_ia32_tpause.
12968 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
12969 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
12970 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
12971 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
12972 UNSPECV_TPAUSE): New.
12973 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
12974 * config/i386/i386.opt: Add -mwaitpkg.
12975 * config/i386/waitpkgintrin.h: New file.
12976 * config/i386/x86intrin.h: New header.
12977 * doc/invoke.texi: Add -mwaitpkg.
12979 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12982 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
12984 (cortex-m0): Use armv6s-m isa.
12985 (cortex-m0plus): Likewise.
12986 (cortex-m1): Likewise.
12987 (cortex-m0.small-multiply): Likewise.
12988 (cortex-m0plus.small-multiply): Likewise.
12989 (cortex-m1.small-multiply): Likewise.
12991 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
12992 Jakub Jelinek <jakub@redhat.com>
12994 PR tree-optimization/85692
12995 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
12996 source permute as well.
12998 2018-05-11 Martin Liska <mliska@suse.cz>
13001 * doc/extend.texi: Document LLVM style format for no_sanitize
13004 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13006 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13007 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13008 (mode_supports_vsx_dform_quad): Likewise.
13009 (mode_supports_vmx_dform): Move these functions to be next to the
13010 other mode_supports functions.
13011 (mode_supports_dq_form): Likewise.
13012 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13013 mode_supports_dq_form.
13014 (reg_offset_addressing_ok_p): Likewise.
13015 (offsettable_ok_by_alignment): Likewise.
13016 (rs6000_legitimate_offset_address_p): Likewise.
13017 (legitimate_lo_sum_address_p): Likewise.
13018 (rs6000_legitimize_address): Likewise.
13019 (rs6000_legitimize_reload_address): Likewise.
13020 (rs6000_secondary_reload_inner): Likewise.
13021 (rs6000_preferred_reload_class): Likewise.
13022 (rs6000_output_move_128bit): Likewise.
13024 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13026 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13027 Generate SImode target register for null target.
13028 <case IX86_BUILTIN_XGETBV>: Ditto.
13029 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13030 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13032 2018-05-10 Carl Love <cel@us.ibm.com>
13034 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13035 dcbtt and dcbtstt if operands[2] is 0.
13037 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13040 * config/i386/sse.md (usadv64qi): New expander.
13042 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13044 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13045 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13046 -maltivec=be support.
13047 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13048 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13049 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13050 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13051 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13052 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13053 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13054 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13055 altivec_vsumsws): Adjust.
13056 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13057 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13058 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13060 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13061 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13062 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13063 (altivec_lve<VI_char>x): Delete expand.
13064 (*altivec_lve<VI_char>x_internal): Rename to...
13065 (altivec_lve<VI_char>x): ... this.
13066 (altivec_lvxl_<mode>): Delete expand.
13067 (*altivec_lvxl_<mode>_internal): Rename to ...
13068 (altivec_lvxl_<mode>): ... this.
13069 (altivec_stvxl_<mode>): Delete expand.
13070 (*altivec_stvxl_<mode>_internal): Rename to ...
13071 (altivec_stvxl_<mode>): ... this.
13072 (altivec_stve<VI_char>x): Delete expand.
13073 (*altivec_stve<VI_char>x_internal): Rename to ...
13074 (altivec_stve<VI_char>x): ... this.
13075 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13076 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13077 reduc_plus_scal_<mode>): Adjust.
13078 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13080 (rs6000_cpu_cpp_builtins): Adjust.
13081 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13082 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13083 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13084 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13085 -maltivec=be support.
13086 (rs6000_split_vec_extract_var): Adjust.
13087 (rs6000_split_v4si_init): Adjust.
13088 (swap_selector_for_mode): Delete.
13089 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13090 altivec_expand_stvex_be): Delete.
13091 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13092 -maltivec=be support.
13093 (rs6000_gimple_fold_builtin): Ditto.
13094 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13096 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13097 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13098 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13099 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13100 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13101 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13102 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13103 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13104 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13105 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13106 anonymous split): Adjust.
13107 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13108 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13110 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13112 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13113 when --with-gxx-include-dir is also specified.
13114 * configure: Regenerate.
13116 2018-05-09 Jim Wilson <jimw@sifive.com>
13119 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13120 * config/riscv/t-withmultilib: New.
13121 * config/riscv/withmultilib.h: New.
13122 * doc/install.texi: Document RISC-V --with-multilib-list support.
13124 2018-05-09 Richard Biener <rguenther@suse.de>
13126 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13128 (vect_bb_vectorization_profitable_p): Adjust. Compute
13129 actual scalar cost using the cost vector and the add_stmt_cost
13132 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13134 PR rtl-optimization/85645
13135 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13136 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13138 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13140 PR rtl-optimization/85645
13141 * shrink-wrap.c (spread_components): Return a boolean saying if
13142 anything was changed.
13143 (try_shrink_wrapping_separate): Iterate spread_components until
13144 nothing changes anymore.
13146 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13148 PR rtl-optimization/85645
13149 * regrename.c (build_def_use): Also kill the chains that include the
13150 destination of a REG_CFA_REGISTER note.
13152 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13154 PR rtl-optimization/85645
13155 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13156 insn that has a REG_CFA_REGISTER note.
13158 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13160 * cfgexpand.c (expand_clobber): New function.
13161 (expand_gimple_stmt_1): Use it.
13162 * tree-vect-stmts.c (vect_clobber_variable): New function,
13164 (vectorizable_simd_clone_call): ...here.
13165 (vectorizable_store): Emit a clobber either side of an
13166 IFN_STORE_LANES sequence.
13167 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13169 2018-05-09 Tom de Vries <tom@codesourcery.com>
13172 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13173 (define_insn "trap_if_false"): Add exit after trap.
13175 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13177 PR rtl-optimization/85638
13178 * bb-reorder.c: Include common/common-target.h.
13179 (create_forwarder_block): New function extracted from...
13180 (fix_up_crossing_landing_pad): ...here. Rename into...
13181 (dw2_fix_up_crossing_landing_pad): ...this.
13182 (sjlj_fix_up_crossing_landing_pad): New function.
13183 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13184 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13185 from both partitions and exit the loop after one iteration.
13187 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13190 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13192 (Basic PowerPC Built-in Functions): The new name of the
13193 subsection previously known as "PowerPC Built-in Functions".
13194 (Basic PowerPC Built-in Functions Available on all Configurations):
13196 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13198 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13200 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13202 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13205 2018-05-08 Jim Wilson <jimw@sifive.com>
13207 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13208 (LD_EMUL_SUFFIX): New.
13209 (LINK_SPEC): Use it.
13211 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13213 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13215 (Basic PowerPC Built-in Functions): The new name of the
13216 subsection previously known as "PowerPC Built-in Functions".
13217 (Basic PowerPC Built-in Functions Available on all Configurations):
13219 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13221 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13223 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13225 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13228 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13231 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13232 after cmpelim optimization.
13234 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13236 * config.gcc: Support "goldmont".
13237 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13238 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13239 PROCESSOR_GOLDMONT.
13240 * config/i386/i386.c (m_GOLDMONT): Define.
13241 (processor_target_table): Add "goldmont".
13242 (PTA_GOLDMONT): Define.
13243 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13244 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13245 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13246 (fold_builtin_cpu): Add "goldmont".
13247 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13248 (ix86_option_override_internal): Add "goldmont".
13249 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13250 (processor_type): Add PROCESSOR_GOLDMONT.
13251 * config/i386/i386.md: Add CPU "glm".
13252 * config/i386/glm.md: New file.
13253 * config/i386/x86-tune.def: Add m_GOLDMONT.
13254 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13256 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13259 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13261 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13262 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13263 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13266 * config/i386/i386.c (ix86_fold_builtin): Handle
13267 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13270 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13271 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13273 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13276 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13277 (check_arch): Likewise.
13278 (check_fpu): Return the result rather than printing it.
13279 (end arch): Fix operator precedence.
13280 (end cpu): Likewise.
13281 (END): Print the result from check_fpu.
13283 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13284 Alan Hayward <alan.hayward@arm.com>
13285 David Sherwood <david.sherwood@arm.com>
13287 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13288 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13289 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13290 (*fcmuo<mode>_and): New patterns.
13292 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13294 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13295 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13296 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13297 (cmp_op, sve_imm_con): New code attributes.
13298 (SVE_COND_INT_CMP, imm_con): Delete.
13299 (cmp_op): Remove above unspecs from int attribute.
13300 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13302 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13303 comparison-specific unspecs.
13304 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13305 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13306 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13307 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13308 (*vec_fcm<cmp_op><mode>): Rename to...
13309 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13310 (*vec_fcmuo<mode>): Rename to...
13311 (*fcmuo<mode>): ...this and adjust likewise.
13312 (*pred_fcm<cmp_op><mode>): New pattern.
13313 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13314 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13316 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13318 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13319 (aarch64_emit_sve_predicated_cond): New function.
13320 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13321 (aarch64_emit_unspec_cond_or): Replace with...
13322 (aarch64_emit_sve_or_conds): ...this new function. Use
13323 aarch64_emit_sve_ptrue_op for the individual comparisons and
13324 aarch64_emit_binop to OR them together.
13325 (aarch64_emit_inverted_unspec_cond): Replace with...
13326 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13327 aarch64_emit_sve_ptrue_op for the comparison and
13328 aarch64_emit_unop to invert the result.
13329 (aarch64_expand_sve_vec_cmp_float): Update after the above
13330 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13332 2018-05-07 Nathan Sidwell <nathan@acm.org>
13334 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13335 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13336 (Backwards Compatibility): Likewise.
13338 2018-05-07 Luis Machado <luis.machado@linaro.org>
13342 2018-05-07 Luis Machado <luis.machado@linaro.org>
13344 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13345 <prefetch_dynamic_strides>: New const bool field.
13346 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13347 prefetch_dynamic_strides.
13348 (exynosm1_prefetch_tune): Likewise.
13349 (thunderxt88_prefetch_tune): Likewise.
13350 (thunderx_prefetch_tune): Likewise.
13351 (thunderx2t99_prefetch_tune): Likewise.
13352 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13354 (aarch64_override_options_internal): Update to set
13355 PARAM_PREFETCH_DYNAMIC_STRIDES.
13356 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13357 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13358 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13359 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13360 prefetch-dynamic-strides setting.
13362 2018-05-07 Luis Machado <luis.machado@linaro.org>
13364 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13365 <minimum_stride>: New const int field.
13366 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13367 minimum_stride field.
13368 (exynosm1_prefetch_tune): Likewise.
13369 (thunderxt88_prefetch_tune): Likewise.
13370 (thunderx_prefetch_tune): Likewise.
13371 (thunderx2t99_prefetch_tune): Likewise.
13372 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13373 (aarch64_override_options_internal): Update to set
13374 PARAM_PREFETCH_MINIMUM_STRIDE.
13375 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13376 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13377 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13378 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13379 stride is constant and is below the minimum stride threshold.
13381 2018-05-07 Luis Machado <luis.machado@linaro.org>
13383 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13386 2018-05-07 Luis Machado <luis.machado@linaro.org>
13388 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13389 <prefetch_dynamic_strides>: New const bool field.
13390 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13391 prefetch_dynamic_strides.
13392 (exynosm1_prefetch_tune): Likewise.
13393 (thunderxt88_prefetch_tune): Likewise.
13394 (thunderx_prefetch_tune): Likewise.
13395 (thunderx2t99_prefetch_tune): Likewise.
13396 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13398 (aarch64_override_options_internal): Update to set
13399 PARAM_PREFETCH_DYNAMIC_STRIDES.
13400 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13401 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13402 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13403 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13404 prefetch-dynamic-strides setting.
13406 2018-05-07 Luis Machado <luis.machado@linaro.org>
13408 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13409 <minimum_stride>: New const int field.
13410 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13411 minimum_stride field.
13412 (exynosm1_prefetch_tune): Likewise.
13413 (thunderxt88_prefetch_tune): Likewise.
13414 (thunderx_prefetch_tune): Likewise.
13415 (thunderx2t99_prefetch_tune): Likewise.
13416 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13417 (aarch64_override_options_internal): Update to set
13418 PARAM_PREFETCH_MINIMUM_STRIDE.
13419 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13420 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13421 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13422 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13423 stride is constant and is below the minimum stride threshold.
13425 2018-05-06 Jakub Jelinek <jakub@redhat.com>
13428 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
13429 the type is addressable. Don't force op into register if it has
13432 2018-05-05 Roland McGrath <mcgrathr@google.com>
13435 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
13436 any section for which we don't know a specific type it should have,
13437 regardless of name. Previously this was done only for the exact
13438 names ".init_array", ".fini_array", and ".preinit_array".
13439 (default_elf_asm_named_section): Add comment about
13440 relationship with default_section_type_flags and SECTION_NOTYPE.
13441 (get_section): Don't consider it a type conflict if one side has
13442 SECTION_NOTYPE and the other doesn't, as long as neither has the
13443 SECTION_BSS et al used in the default_section_type_flags logic.
13445 2018-05-05 Tom de Vries <tom@codesourcery.com>
13448 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
13449 (workaround_barsyncs): New function.
13450 (nvptx_reorg): Use workaround_barsyncs.
13451 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13452 (define_expand "nvptx_membar_cta"): New define_expand.
13453 (define_insn "*nvptx_membar_cta"): New insn.
13455 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
13457 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
13458 To improve optimization opportunities.
13459 * builtin-types.def: The new needed builtin types for the above.
13461 2018-05-04 Richard Biener <rguenther@suse.de>
13463 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
13464 * gimple-ssa-store-merging.c
13465 (imm_store_chain_info::output_merged_store): Remove redundant create,
13466 release split_store vector contents on failure.
13467 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
13468 scalar stmt vector on cache hit.
13470 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
13472 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
13474 * config.gcc (powerpc-xilinx-eabi*): Remove.
13475 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
13477 (fusion_addis_mem_combo_load): Ditto.
13478 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
13480 (rs6000_cpu_cpp_builtins): Ditto.
13481 * config/rs6000/rs6000-linux.c
13482 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
13483 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
13484 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
13486 (rs6000_setup_reg_addr_masks): Ditto.
13487 (rs6000_init_hard_regno_mode_ok): Ditto.
13488 (rs6000_option_override_internal): Ditto.
13489 (legitimate_lo_sum_address_p): Ditto.
13490 (rs6000_legitimize_address): Ditto.
13491 (rs6000_legitimize_reload_address): Ditto.
13492 (rs6000_legitimate_address_p): Ditto.
13493 (abi_v4_pass_in_fpr): Ditto.
13494 (setup_incoming_varargs): Ditto.
13495 (rs6000_gimplify_va_arg): Ditto.
13496 (rs6000_split_multireg_move): Ditto.
13497 (rs6000_savres_strategy): Ditto.
13498 (rs6000_emit_prologue_components): Ditto.
13499 (rs6000_emit_epilogue_components): Ditto.
13500 (rs6000_emit_prologue): Ditto.
13501 (rs6000_emit_epilogue): Ditto.
13502 (rs6000_elf_file_end): Ditto.
13503 (rs6000_function_value): Ditto.
13504 (rs6000_libcall_value): Ditto.
13505 * config/rs6000/rs6000.h: Ditto.
13506 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
13507 (TARGET_MINMAX): ... this. New.
13508 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
13509 * config/rs6000/rs6000.md: Remove Xilinx FP support.
13510 (*movsi_internal1_single): Delete.
13511 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
13512 mfpu=, mxilinx-fpu): Delete.
13513 * config/rs6000/singlefp.h: Delete.
13514 * config/rs6000/sysv4.h: Remove Xilinx FP support.
13515 * config/rs6000/t-rs6000: Ditto.
13516 * config/rs6000/t-xilinx: Delete.
13517 * config/rs6000/titan.md: Adjust for fp_type removal.
13518 * config/rs6000/vsx.md: Remove Xilinx FP support.
13519 (VStype_simple): Delete.
13520 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
13521 * config/rs6000/xfpu.h: Delete.
13522 * config/rs6000/xfpu.md: Delete.
13523 * config/rs6000/xilinx.h: Delete.
13524 * config/rs6000/xilinx.opt: Delete.
13525 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
13526 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
13528 2018-05-04 Tom de Vries <tom@codesourcery.com>
13531 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
13534 2018-05-04 Richard Biener <rguenther@suse.de>
13536 PR middle-end/85627
13537 * tree-complex.c (update_complex_assignment): We are always in SSA form.
13538 (expand_complex_div_wide): Likewise.
13539 (expand_complex_operations_1): Likewise.
13540 (expand_complex_libcall): Preserve EH info of the original stmt.
13541 (tree_lower_complex): Handle removed blocks.
13542 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
13543 on complex multiplication and division libcall builtins.
13545 2018-05-04 Richard Biener <rguenther@suse.de>
13547 PR middle-end/85574
13548 * fold-const.c (negate_expr_p): Restrict negation of operand
13549 zero of a division to when we know that can happen without
13551 (fold_negate_expr_1): Likewise.
13553 2018-05-04 Jakub Jelinek <jakub@redhat.com>
13556 * real.h (real_nextafter): Declare.
13557 * real.c (real_nextafter): New function.
13558 * fold-const-call.c (fold_const_nextafter): New function.
13559 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
13560 CASE_CFN_NEXTTOWARD.
13561 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
13562 even when arg1_mode is different from arg0_mode.
13564 2018-05-03 Nathan Sidwell <nathan@acm.org>
13566 * doc/extend.texi (Deprecated Features): Remove
13567 -ffriend-injection.
13568 (Backwards Compatibility): Likewise.
13569 * doc/invoke.texi (C++ Language Options): Likewise.
13570 (C++ Dialect Options): Likewise.
13572 2018-05-03 Jakub Jelinek <jakub@redhat.com>
13575 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
13576 _mm512_mask_mullox_epi64): New intrinsics.
13578 2018-05-03 Tom de Vries <tom@codesourcery.com>
13581 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13582 dump files): Add offload-tree.
13584 2018-05-03 Richard Biener <rguenther@suse.de>
13586 PR tree-optimization/85615
13587 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
13588 to loops not nested in BBs loop father to avoid creating multi-entry
13591 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13593 PR tree-optimization/70291
13594 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
13595 arguments. Change return type to tree. Emit libcall as a new
13596 statement rather than replacing existing one when inplace_p is true.
13597 (expand_complex_multiplication_components): New function.
13598 (expand_complex_multiplication): Expand floating-point complex
13599 multiplication using the above.
13600 (expand_complex_division): Rename inner_type parameter to type.
13601 Update expand_complex_libcall call-site.
13602 (expand_complex_operations_1): Update expand_complex_multiplication
13603 and expand_complex_division call-sites.
13605 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13608 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13609 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13610 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13611 the highest significant bit of the shift count mask is clear. In
13612 check whether and[sq]i3 is needed verify that all significant bits
13613 of the shift count other than the highest are set.
13615 2018-05-02 Tom de Vries <tom@codesourcery.com>
13618 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13619 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13620 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13621 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13622 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13623 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13624 __builtin_goacc_parlevel_size.
13626 2018-05-02 Richard Biener <rguenther@suse.de>
13628 PR tree-optimization/85597
13629 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13630 do not use split vect_get_vec_defs call but call vect_get_slp_defs
13633 2018-05-02 Tom de Vries <tom@codesourcery.com>
13636 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13637 dump files): Add ltrans-tree.
13639 2018-05-02 Tom de Vries <tom@codesourcery.com>
13642 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13643 dump files): Add wpa-ipa.
13645 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
13647 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
13648 powerpc*-*-linux*paired* target.
13649 * config/rs6000/750cl.h: Delete.
13650 * config/rs6000/paired.h: Delete.
13651 * config/rs6000/paired.md: Delete.
13652 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13654 * config/rs6000/rs6000-builtin.def: Remove paired float support.
13655 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13656 comment. Remove paired float support.
13657 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13658 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13660 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13661 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13663 * config/rs6000/rs6000.c: Remove paired float support.
13664 (paired_expand_vector_init, paired_expand_vector_move,
13665 paired_emit_vector_compare, paired_emit_vector_cond_expr,
13666 (paired_expand_lv_builtin, paired_expand_stv_builtin,
13667 paired_expand_builtin, paired_expand_predicate_builtin,
13668 paired_init_builtins): Delete.
13669 * config/rs6000/rs6000.h: Remove paired float support.
13670 * config/rs6000/rs6000.md: Remove paired float support.
13671 (move_from_CR_ov_bit): Delete.
13672 * config/rs6000/rs6000.opt (mpaired): Delete.
13673 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13674 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13676 2018-05-02 Richard Biener <rguenther@suse.de>
13678 PR middle-end/85567
13679 * gimplify.c (gimplify_save_expr): When in SSA form allow
13680 SAVE_EXPRs to compute to SSA vars.
13682 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13685 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13686 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13687 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13688 clobber operands[2], instead use a new pseudo. Formatting fixes.
13690 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
13692 PR tree-optimization/85586
13693 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13694 exit early for statements in the same group if the accesses are
13697 2018-05-02 Tom de Vries <tom@codesourcery.com>
13700 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13703 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
13705 PR tree-optimization/85143
13706 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13708 2018-05-01 Tom de Vries <tom@codesourcery.com>
13711 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13712 not found" error message.
13714 2018-05-01 Tom de Vries <tom@codesourcery.com>
13717 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13718 * vec.c (test_ordered_remove_if): New function.
13719 (vec_c_tests): Call test_ordered_remove_if.
13720 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13721 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13722 * tree-vect-patterns.c (vect_pattern_recog_1): Use
13723 VEC_ORDERED_REMOVE_IF.
13725 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13727 PR tree-optimization/82665
13728 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13729 pointer subtraction where arguments come from a memchr call.
13731 2018-05-01 Jakub Jelinek <jakub@redhat.com>
13733 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
13734 --push-state --as-needed and --pop-state instead of --as-needed and
13735 --no-as-needed if ld supports it.
13736 * configure: Regenerated.
13739 * doc/install.texi2html: Replace _002d with - and _002a with * in
13740 generated html files using sed.
13742 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13745 * gcc-rich-location.c (blank_line_before_p): New function.
13746 (use_new_line): New function.
13747 (gcc_rich_location::add_fixit_insert_formatted): New function.
13748 * gcc-rich-location.h
13749 (gcc_rich_location::add_fixit_insert_formatted): New function.
13751 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13753 * selftest.c (assert_streq): Rename "expected" and "actual" to
13754 "val1" and "val2". Extend NULL-handling to cover both inputs
13755 symmetrically, while still requiring both to be non-NULL for a pass.
13756 * selftest.h (assert_streq): Rename "expected" and "actual" to
13758 (ASSERT_EQ): Likewise.
13759 (ASSERT_EQ_AT): Likewise.
13760 (ASSERT_KNOWN_EQ): Likewise.
13761 (ASSERT_KNOWN_EQ_AT): Likewise.
13762 (ASSERT_NE): Likewise.
13763 (ASSERT_MAYBE_NE): Likewise.
13764 (ASSERT_MAYBE_NE_AT): Likewise.
13765 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
13766 the assertion to pass.
13767 (ASSERT_STREQ_AT): Likewise.
13769 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13771 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
13772 interaction with -pie.
13774 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13776 * selftest.h: Fix alphabetization of per-source-file selftest
13779 2018-04-30 Jason Merrill <jason@redhat.com>
13781 PR c++/61982 - dead stores to destroyed objects.
13782 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
13785 2018-04-30 Jason Merrill <jason@redhat.com>
13787 * tree.c (build_clobber): New.
13788 * tree.h: Declare it.
13789 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13791 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13793 * diagnostic-show-locus.c (layout::layout): Update for
13794 location_get_source_line returning a char_span.
13795 (struct char_span): Move to input.h.
13796 (struct correction): Update for fields in char_span becoming
13798 (struct source_line): Update for location_get_source_line
13799 returning a char_span.
13800 (layout::print_line): Likewise.
13801 * edit-context.c (edited_file::print_content): Likewise.
13802 (edited_file::print_diff_hunk): Likewise.
13803 (edited_file::print_run_of_changed_lines): Likewise.
13804 (edited_file::get_num_lines): Likewise.
13805 (edited_line::edited_line): Likewise.
13806 * final.c (asm_show_source): Likewise.
13807 * input.c (location_get_source_line): Convert return type
13808 from const char * to char_span, losing the final "line_len"
13810 (dump_location_info): Update for the above.
13811 (get_substring_ranges_for_loc): Likewise. Use a char_span
13812 when handling the literal within the line.
13813 (test_reading_source_line): Update for location_get_source_line
13814 returning a char_span.
13815 * input.h (class char_span): Move here from
13816 diagnostic-show-locus.c, converting from a struct to a class.
13817 Make data members private.
13818 (char_span::operator bool): New.
13819 (char_span::length): New.
13820 (char_span::get_buffer): New.
13821 (char_span::operator[]): New.
13822 (char_span::subspan): Make const.
13823 (char_span::xstrdup): New.
13824 (location_get_source_line): Convert return type from const char *
13825 to char_span, losing the final "line_size" param.
13827 2018-04-30 Jan Hubicka <jh@suse.cz>
13829 * lto-wrapper.c (ltrans_priorities): New static var.
13830 (cmp_priority): New.
13831 (run_gcc): Read priorities and if doing parallel build order
13832 the Makefile by them.
13834 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13836 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
13838 2018-04-30 Richard Biener <rguenther@suse.de>
13840 * tree-cfg.c (verify_address): Remove base argument, add
13841 flag whether to check TREE_ADDRESSABLE and do that.
13842 (verify_expr): Remove.
13843 (verify_types_in_gimple_reference): Add pieces from verify_expr.
13844 (verify_gimple_assign_single): Likewise.
13845 (verify_gimple_switch): Likewise.
13846 (verify_expr_location_1): Dereference tp once. Add (disabled)
13847 piece from verify_expr.
13848 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
13850 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13852 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
13854 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13856 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
13857 (small_data_pattern): Likewise.
13858 (arc_rewrite_small_data): Likewise.
13859 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
13860 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
13861 (get_symbol_alignment): New function.
13862 (legitimate_small_data_address_p): Likewise.
13863 (legitimate_scaled_address): Update, call
13864 legitimate_small_data_address_p.
13865 (output_sdata): New static variable.
13866 (arc_print_operand): Update how we handle small data operands.
13867 (arc_print_operand_address): Likewise.
13868 (arc_legitimate_address_p): Update, use
13869 legitimate_small_data_address_p.
13870 (arc_rewrite_small_data_p): Remove.
13871 (arc_rewrite_small_data_1): Likewise.
13872 (arc_rewrite_small_data): Likewise.
13873 (small_data_pattern): Likewise.
13874 (compact_sda_memory_operand): Update to use
13875 legitimate_small_data_address_p and get_symbol_alignment.
13876 (prepare_move_operands): Don't rewite sdata pattern.
13877 (prepare_extend_operands): Remove.
13878 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
13880 (zero_extendqisi2): Likewise.
13881 (zero_extendhisi2): Likewise.
13882 (extendqihi2): Likewise.
13883 (extendqisi2): Likewise.
13884 (extendhisi2): Likewise.
13885 (addsi3): Likewise.
13886 (subsi3): Likewise.
13887 (andsi3): Likewise.
13888 * config/arc/constraints.md (Usd): Change it to memory constraint.
13890 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13892 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
13893 as source of std instructions.
13894 * config/arc/arc.md (movsi_insn): Update pattern predicate to
13895 allow 6-bit constants as source for store instructions.
13896 (movdi_insn): Update instruction pattern to allow 6-bit constants
13897 as source for store instructions.
13899 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13901 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
13903 2018-04-30 Nathan Sidwell <nathan@acm.org>
13904 Sandra Loosemore <sandra@codesourcery.com>
13906 * dumpfile.c (dump_open): Allow '-' for stdout.
13907 * doc/invoke.texi (Developer Options): Document dump filename
13908 determination early. Document stdin/stdout selection.
13910 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
13912 Microblaze Target: PIC data text relative
13914 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
13915 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
13917 * config/microblaze/microblaze.h (microblaze_constant_address_p):
13918 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
13919 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
13920 New addressing mode for data-text relative position indepenedent code.
13921 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
13922 'ADDRESS_SYMBOLIC_TXT_REL'.
13923 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
13924 (microblaze_legitimate_pic_operand): Exclude function calls from
13925 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
13926 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
13928 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13929 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13930 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
13931 for 'address + offset'.
13932 (microblaze_expand_prologue): Add new function prologue call for
13934 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
13935 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
13936 table in case of TARGET_PIC_DATA_TEXT_REL.
13937 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
13938 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
13939 Add new macros 'UNSPEC_TEXT',
13940 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
13941 + exclude function calls from 'UNSPEC_PLT' in case of data text
13943 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13944 new target hook for generating address diff vector tables in case of
13946 * doc/tm.texi : Regenerate.
13947 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
13948 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
13949 of addr diff vector generation.
13950 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13951 target hook definition.
13952 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
13953 Add default function for generate_pic_addr_diff_vec -> flag_pic.
13954 * doc/invoke.texi (Add new pic option): Add new microblaze pic
13955 option for data text relative.
13957 2018-04-30 Richard Biener <rguenther@suse.de>
13959 * tree-chrec.h (evolution_function_is_constant_p): Remove
13961 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
13963 2018-04-30 Richard Biener <rguenther@suse.de>
13966 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
13968 2018-04-30 Richard Biener <rguenther@suse.de>
13970 PR tree-optimization/28364
13971 PR tree-optimization/85275
13972 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
13973 copying first exit test.
13975 2018-04-28 Mark Wielaard <mark@klomp.org>
13977 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
13978 dwarf_version >= 5.
13979 (dwarf_AT): Handle DW_AT_addr_base.
13980 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
13982 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
13985 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
13986 (*ashl<dwi>3_doubleword_mask_1): Ditto.
13987 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
13988 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
13990 2018-04-28 Richard Biener <rguenther@suse.de>
13992 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
13993 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
13994 to reflect use. Only add interesting stmts.
13996 2018-04-27 Martin Jambor <mjambor@suse.cz>
13999 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14000 the jump function allows for passing through aggregate values.
14002 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14004 * input.h (in_system_header_at): Convert from macro to inline
14006 (from_macro_expansion_at): Likewise.
14007 (from_macro_definition_at): Likewise.
14009 2018-04-27 Jeff Law <law@redhat.com>
14011 * config.gcc: Mark tile* targets as deprecated/obsolete.
14013 2018-04-27 Richard Biener <rguenther@suse.de>
14015 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14018 2018-04-27 Richard Biener <rguenther@suse.de>
14020 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14022 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14024 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14025 with Yd constraint. Set "preferred_for_speed" attribute from
14026 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14027 with Yd constraint.
14028 (*movdi_internal): Ditto.
14029 (movti_interunit splitters): Remove
14030 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14031 (movdi_interunit splitters): Ditto.
14032 * config/i386/constraints.md (Ye): Remove.
14033 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14035 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14038 * config/aarch64/constraints.md (Usg): Limit to 31.
14039 (Usj): Limit to 63.
14041 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14043 PR tree-optimization/85529
14044 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14045 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14046 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14047 zero extension or masking of the MSB bit.
14048 (optimize_range_tests): Add FIRST_BB argument, pass it through
14049 to optimize_range_tests_var_bound.
14050 (maybe_optimize_range_tests, reassociate_bb): Adjust
14051 optimize_range_tests callers.
14053 2018-04-26 Richard Biener <rguenther@suse.de>
14054 Jakub Jelinek <jakub@redhat.com>
14056 * cgraph.h (symbol_table): Just declare debug method here.
14057 * symtab.c (symbol_table::debug): Define.
14059 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14061 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14063 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14065 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14066 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14067 (*movdi_internal): Substitute Yi and Yj constraint with x
14068 and Ym and Yn constraint with y constraint. Update "isa"
14069 attribute and set "preferred_for_speed" attribute from
14070 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14071 (*movsi_internal): Ditto.
14072 (*movdf_internal): Ditto.
14073 (*movsf_internal): Ditto.
14074 (*zero_extendsidi2): Ditto.
14075 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14076 (sse2_loadld): Ditto.
14077 (*vec_extract<ssevecmodelower>_0): Ditto.
14078 (*vec_extractv4si_0_zext_sse4): Ditto.
14079 (vec_concatv2di): Ditto.
14080 (*vec_dup<mode>): Ditto.
14081 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14082 * config/i386/constraints.md (Yi): Remove.
14087 2018-04-26 Nathan Sidwell <nathan@acm.org>
14089 * dumpfile.c (dump_open): New.
14090 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14091 (dump_finish): Detect stdio/stderr by value not name.
14093 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14095 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14097 2018-04-26 Tom de Vries <tom@codesourcery.com>
14100 * config/nvptx/nvptx.c (verify_neutering_jumps)
14101 (verify_neutering_labels): New function
14102 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14104 2018-04-26 Tom de Vries <tom@codesourcery.com>
14107 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14108 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14111 2018-04-26 Richard Biener <rguenther@suse.de>
14112 Tom de Vries <tom@codesourcery.com>
14115 * lto-streamer-out.c (output_function): Fixup loops if required to match
14116 discovery done in the reader.
14118 2018-04-26 Richard Biener <rguenther@suse.de>
14120 PR tree-optimization/85116
14121 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14122 have a loop exit from the single latch predecessor. Remove
14123 case of header with just condition.
14124 (ch_base::copy_headers): Exclude infinite loops from any
14126 (pass_ch::execute): Record exits.
14128 2018-04-26 Richard Biener <rguenther@suse.de>
14130 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14131 prologue cost vector and pass it to vect_get_load_cost.
14132 (vect_get_peeling_costs_all_drs): Likewise.
14133 (vect_peeling_hash_get_lowest_cost): Likewise.
14134 (vect_enhance_data_refs_alignment): Likewise.
14136 2018-04-26 Richard Biener <rguenther@suse.de>
14138 PR middle-end/85450
14139 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14140 checking of integer<->pointer conversions.
14141 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14142 sign-/zero-extending pointer types.
14143 (expand_omp_for_static_chunk): Likewise.
14145 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14146 Jean Lee <xiaoyur347@gmail.com>
14148 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14149 (TARGET_ASAN_SHADOW_OFFSET): Define.
14150 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14151 true for -fsanitize=address.
14153 2018-04-25 Mark Wielaard <mark@klomp.org>
14155 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14158 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14160 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14161 than "alu", remove explicit "memory" and "imm_disp" attributes.
14162 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14164 PR middle-end/85414
14165 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14166 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14167 gen_lowpart_no_emit.
14169 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14172 * config/i386/i386.c (ix86_expand_builtin): Change memory
14173 operand to XI, extend p0 to Pmode.
14174 * config/i386/i386.md: Change unspec volatile and operand
14175 1 mode to XI, change operand 0 mode to P.
14177 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14179 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14180 GET_MODE_MASK before any checking.
14181 (nds32_can_use_bset_p): Likewise.
14182 (nds32_can_use_btgl_p): Likewise.
14184 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14186 * config/nds32/nds32-doubleword.md: New define_split pattern for
14187 illegal register number.
14189 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14191 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14193 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14195 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14197 2018-04-25 Richard Biener <rguenther@suse.de>
14199 * lto-streamer.h (LTO_major_version): Bump to 8.
14201 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14203 * BASE-VER: Set to 9.0.0.
14205 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14207 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14208 in __abskf2 and __powikf2.
14210 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14213 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14214 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14215 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14216 Use the above on operand 2. Reindent.
14217 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14219 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14222 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14224 * config/i386/i386.opt (mcet): Removed.
14225 * doc/install.texi: Remove -mcet documentation.
14226 * doc/invoke.texi: Likewise.
14228 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14231 * doc/install.texi: Remove -mcet from bootstrap-cet.
14233 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14236 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14237 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14241 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14242 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14243 containing a CONST_VECTOR.
14245 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14247 * doc/install.texi: Update newlib dependency for nvptx.
14249 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14252 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14253 instead of INTVAL when shifting x left.
14255 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14257 PR tree-optimization/85478
14258 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14259 vect_grouped_store_supported for single element vectors.
14261 2018-04-24 Richard Biener <rguenther@suse.de>
14264 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14265 load cost increase to the case of non-constant step.
14267 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14270 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14271 destination if any_malformed_asm.
14273 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14275 PR middle-end/85496
14276 * expr.c (store_field): In the bitfield case, if the value comes from
14277 a function call and is returned in registers by means of a PARALLEL,
14278 do not change the mode of the temporary unless BLKmode and VOIDmode.
14280 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14282 PR rtl-optimization/85423
14283 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14284 dependencies to debug insns when the previous insn is non-debug.
14286 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14288 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14289 enums into a single definition.
14290 (fls): Fix predicates and printing.
14293 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14295 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14296 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14297 and short u6 immediate.
14298 (check_if_valid_sleep_operand): Remove.
14299 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14301 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14303 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14304 flag_always_save_lp condition.
14305 * config/nds32/nds32.opt (malways-save-lp): New option.
14307 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14309 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14310 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14311 * config/nds32/nds32.h
14312 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14313 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14315 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14317 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14318 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14320 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14321 Chung-Ju Wu <jasonwucj@gmail.com>
14323 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14325 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14326 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14328 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14330 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14332 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14334 * config/nds32/nds32-protos.h (nds32_data_alignment,
14335 nds32_local_alignment): Declare.
14336 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14337 nds32_local_alignment): New functions.
14338 (TARGET_CONSTANT_ALIGNMENT): Define.
14339 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14341 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14343 * config/nds32/nds32.c
14344 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14345 (TARGET_MODES_TIEABLE_P): Likewise.
14347 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14349 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14350 level Ofast and Og.
14352 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14353 Chung-Ju Wu <jasonwucj@gmail.com>
14355 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14356 for unaligned access.
14357 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14359 * config/nds32/nds32-intrinsic.md: Likewise.
14360 * config/nds32/nds32_intrinsic.h: Likewise.
14361 * config/nds32/nds32.h (nds32_builtins): Likewise.
14362 * config/nds32/nds32.opt (munaligned-access): New option.
14363 * config/nds32/nds32.c (nds32_asm_file_start): Display
14364 flag_unaligned_access status.
14366 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14368 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14369 -mno-relax is present.
14370 * config/riscv/linux.h (LINK_SPEC): Ditto.
14372 2018-04-20 Martin Sebor <msebor@redhat.com>
14375 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14377 (gimple_fold_builtin_stxcpy_chk): Same.
14378 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14380 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14383 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14384 __powikf2 when long double is IEEE 128-bit.
14386 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14388 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14389 step to make sure stack always aligned.
14391 2018-04-20 Carl Love <cel@us.ibm.com>
14394 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14395 size check for arg0.
14397 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14398 Tom de Vries <tom@codesourcery.com>
14401 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14402 Emit insns for calls too.
14403 (nvptx_find_par): Always look for worker-level predecessor insn.
14404 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14406 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14407 (nvptx_process_pars): Propagate frames for calls.
14409 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14412 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14414 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14415 (ix86_handle_option): Don't handle OPT_mibt.
14416 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14418 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14420 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14421 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14422 (ix86_target_macros): Define __CET__ with flag_cf_protection
14423 for -fcf-protection.
14424 * config/i386/i386.c (isa2_opts): Remove -mibt.
14425 * config/i386/i386.h (TARGET_IBT): Removed.
14426 (TARGET_IBT_P): Likewise.
14427 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14428 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
14429 * config/i386/i386.opt (mcet): Update help message.
14430 (mshstk): Likewise.
14432 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
14433 -mcet as an alias for -mshstk.
14435 2018-04-20 Richard Biener <rguenther@suse.de>
14437 PR middle-end/85475
14438 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
14439 complexity by forcing a single use of the multiply operand.
14441 2018-04-20 Martin Jambor <mjambor@suse.cz>
14444 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
14445 recursion dependency to only apply to non-clones.
14447 2018-04-20 Martin Jambor <mjambor@suse.cz>
14450 * ipa-cp.c (create_specialized_node): Check that clones of
14451 self-recursive edges exist during IPA-CP.
14453 2018-04-19 Toon Moene <toon@moene.org>
14455 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
14458 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14460 PR tree-optimization/85467
14461 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
14462 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
14463 VECTOR_CST element to type.
14465 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14468 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
14469 * config/i386/i386.md (builtin_setjmp_setup): Removed.
14470 (builtin_longjmp): Likewise.
14471 (save_stack_nonlocal): New pattern.
14472 (restore_stack_nonlocal): Likewise.
14474 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14477 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14478 Replace ASM_OUTPUT_LABEL with fprintf.
14480 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14483 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14484 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
14485 * config/i386/i386-c.c (ix86_target_macros_internal): Also
14486 define __IBT__ and __SHSTK__ for -fcf-protection.
14487 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
14489 (ix86_trampoline_init): Likewise.
14490 (x86_output_mi_thunk): Likewise.
14491 (ix86_notrack_prefixed_insn_p): Likewise.
14492 (ix86_option_override_internal): Don't disallow -fcf-protection.
14493 * config/i386/i386.md (rdssp<mode>): Also enable for
14495 (incssp<mode>): Likewise.
14496 (nop_endbr): Likewise.
14497 * config/i386/i386.opt (mcet): Change help message to built-in
14500 (mshstk): Likewise.
14501 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
14502 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
14503 enable CET built-in functions.
14505 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
14507 * common/config/i386/i386-common.c
14508 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
14509 OPTION_MASK_ISA_MOVDIRI_UNSET,
14510 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
14511 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
14512 * config.gcc (movdirintrin.h): New header.
14513 * config/i386/cpuid.h (bit_MOVDIRI,
14514 bit_MOVDIR64B): New bits.
14515 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
14517 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
14518 (VOID, PVOID, PCVOID)): New function types.
14519 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
14520 __builtin_ia32_directstoreu_u64,
14521 __builtin_ia32_movdir64b): New builtins.
14522 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
14523 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
14525 (ix86_valid_target_attribute_inner_p): Ditto.
14526 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
14527 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
14528 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
14529 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
14530 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
14531 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
14532 (movdiri<mode>, movdir64b_<mode>): New.
14533 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
14534 * config/i386/immintrin.h: Include movdirintrin.h.
14535 * config/i386/movdirintrin.h: New file.
14536 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
14538 2018-04-19 Richard Biener <rguenther@suse.de>
14540 PR middle-end/85455
14541 * cfg.c (clear_bb_flags): When loop state says we have
14542 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
14544 2018-04-19 Richard Biener <rguenther@suse.de>
14546 PR tree-optimization/84737
14547 * tree-vect-data-refs.c (vect_copy_ref_info): New function
14548 copying restrict info.
14549 (vect_setup_realignment): Use it.
14550 * tree-vectorizer.h (vect_copy_ref_info): Declare.
14551 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
14552 the first DR to all generated stores.
14553 (vectorizable_load): Likewise for loads.
14555 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14557 PR tree-optimization/85446
14558 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
14559 the integral and pointer types to have the same precision.
14561 * doc/install.texi: Document --disable-cet being the default and
14564 2018-04-18 Martin Liska <mliska@suse.cz>
14566 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
14569 2018-04-18 Martin Liska <mliska@suse.cz>
14572 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14575 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14576 arguments if they are comparable.
14578 2018-04-18 Martin Liska <mliska@suse.cz>
14581 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14584 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14587 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
14590 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14591 ENDBR after calling __morestack.
14593 2018-04-18 David Malcolm <dmalcolm@redhat.com>
14596 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
14597 by using gcc_base_ver to generate a gcc_driver_version, and use
14598 it when generating GCC_DRIVER_NAME.
14599 * configure: Regenerate.
14601 2018-04-18 Jakub Jelinek <jakub@redhat.com>
14604 * config.gcc: Obsolete powerpc*-*-*spe*.
14606 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14609 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14610 (stabstr_D): Change type of unum from unsigned int to
14611 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
14614 2018-04-17 Jim Wilson <jimw@sifive.com>
14617 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14618 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14619 Set arg_pointer_offset after using pretend_args_size.
14621 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14623 PR rtl-optimization/85431
14624 * dse.c (record_store): Ignore zero width stores.
14627 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
14628 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14629 __builtin_stack_restore rather than after it.
14630 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14631 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14632 argument instead of virtual_dynamic_stack_rtx.
14634 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
14636 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14638 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14639 Add note to error message to explain internal mapping of overloaded
14640 built-in function name to non-overloaded built-in function name.
14641 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14644 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14647 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14648 where the inputs overlap with the output.
14650 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14652 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14653 (=v, v) alternative and explicit "memory" attribute.
14654 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
14655 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14657 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14658 "sselog1" type instead of "sselog".
14659 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14660 "sselog". Remove explicit "memory" attribute.
14661 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14662 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14664 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14665 "sselog1" type instead of "sselog". Remove explicit "memory"
14667 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14668 use "sselog1" type instead of "sselog". Remove explicit "memory"
14670 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14671 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14673 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14674 "sselog1" type instead of "sselog". Remove explicit "memory"
14676 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14677 use "sselog1" type instead of "sselog". Remove explicit "memory"
14681 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14683 PR middle-end/85414
14684 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14687 2018-04-17 Martin Jambor <mjambor@suse.cz>
14690 * ipa-cp.c (create_specialized_node): Call
14691 expand_all_artificial_thunks if necessary.
14693 2018-04-17 Martin Liska <mliska@suse.cz>
14696 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14697 in message, remote space in between '_G' and '('.
14699 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14702 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14703 avx512f_vmcmp<mode>3<round_saeonly_name>,
14704 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14705 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14706 avx512f_rndscale<mode><round_saeonly_name>,
14707 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14708 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14709 Use %<iptr>2 instead of %2 for -masm=intel.
14710 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14711 avx512f_vcvttss2usi<round_saeonly_name>,
14712 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14714 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14715 avx512f_vcvttsd2usi<round_saeonly_name>,
14716 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14717 Use %q1 instead of %1 for -masm=intel.
14718 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14719 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14720 of %3 for -masm=intel.
14721 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14723 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14725 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14727 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14728 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14730 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14732 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
14733 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
14734 %g1 and one with %0 and %1.
14735 (avx512er_vmrcp28<mode><round_saeonly_name>,
14736 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
14737 %1 for -masm=intel.
14738 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
14739 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
14740 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
14741 of %0 and %{%4%} for -masm=intel.
14742 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
14743 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
14744 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
14745 order of %0 and %{%5%}%{z%} for -masm=intel.
14747 2018-04-17 Jan Hubicka <jh@suse.cz>
14750 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
14752 2018-04-17 Martin Liska <mliska@suse.cz>
14755 * multiple_target.c (create_dispatcher_calls): Set apostrophes
14756 for target_clone error message. Make default implementation
14757 clone to be a local declaration.
14758 (separate_attrs): Add new argument and check for an empty
14760 (expand_target_clones): Handle it.
14761 (ipa_target_clone): Make redirection just for target_clones
14764 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
14765 Tom de Vries <tom@codesourcery.com>
14767 PR middle-end/84955
14768 * omp-expand.c (expand_oacc_for): Add dummy false branch for
14769 tiled basic blocks without omp continue statements.
14771 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
14774 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
14775 vec_extract expression as having side effects to make sure it gets
14778 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
14781 * config/i386/i386.c (get_builtin_code_for_version): Check
14784 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
14787 * config.gcc: Support "skylake".
14788 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14790 * config/i386/i386.c (m_SKYLAKE): Define.
14791 (processor_target_table): Add "skylake".
14792 (ix86_option_override_internal): Add "skylake".
14793 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
14794 PROCESSOR_CANNONLAKE.
14795 (get_builtin_code_for_version): Fix priority for
14796 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
14797 PROCESSOR_SKYLAKE-AVX512.
14798 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
14799 (processor_type): Add PROCESSOR_SKYLAKE.
14801 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
14802 Jason Merrill <jason@redhat.com>
14805 * convert.c (convert_to_integer_1): Use direct recursion for
14806 enumeral types and types with a precision less than the number
14807 of bits in their mode.
14809 2018-04-16 Julia Koval <julia.koval@intel.com>
14812 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
14813 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
14815 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
14818 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
14819 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
14820 and -mno-direct-move.
14822 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
14825 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
14826 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
14827 Ensure negative shifts result in {0}.
14829 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
14831 PR rtl-optimization/79916
14832 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
14833 regs (if any) to define how to gnerate SD moves when LRA is in
14836 2018-04-13 Jakub Jelinek <jakub@redhat.com>
14838 PR rtl-optimization/85393
14839 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
14840 * except.c (expand_dw2_landing_pad_for_region): Make static.
14841 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
14842 a label and unconditional jump to old_bb, rather than
14843 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
14846 PR rtl-optimization/85376
14847 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
14848 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
14849 instead of a specific value.
14851 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14852 Bin Cheng <bin.cheng@arm.com>
14854 PR tree-optimization/82965
14855 PR tree-optimization/83991
14856 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
14857 by_profile_only parameter.
14858 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
14859 information if the loop was predicted to iterate too many times.
14860 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
14862 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14865 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
14868 2018-04-13 Martin Liska <mliska@suse.cz>
14869 Jakub Jelinek <jakub@redhat.com>
14871 PR middle-end/81657
14872 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
14873 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
14874 * builtins.c (expand_builtin_memory_copy_args): Use
14875 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
14876 handle dest_addr == pc_rtx.
14878 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
14881 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
14882 asked to not generate direct moves.
14883 (fix_trunc<mode>si2_stfiwx): Similar.
14884 (fix_trunc<mode>si2_internal): Similar.
14886 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14889 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
14890 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
14891 lookup if dest in some wider mode is known to be const0_rtx and
14892 if so, record permanent equivalence for it to be ZERO_EXTEND of
14893 the narrower mode destination.
14895 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14897 * lto-streamer-out.c (output_function): Revert 259346.
14898 * omp-expand.c (expand_oacc_for): Likewise.
14900 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
14902 PR rtl-optimization/85354
14903 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
14904 * sel-sched.c (sel_global_init): ... here.
14906 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
14909 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
14910 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14911 mode for PE-COFF targets.
14912 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
14913 (i386_pe_asm_lto_end): Likewise.
14914 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
14915 (TARGET_ASM_LTO_END): Likewise.
14916 * config/i386/winnt.c (saved_debug_info_level): New static variable.
14917 (i386_pe_asm_lto_start): New function.
14918 (i386_pe_asm_lto_end): Likewise.
14920 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14921 Richard Biener <rguenther@suse.de>
14923 PR middle-end/84955
14924 * lto-streamer-out.c (output_function): Fix CFG loop state before
14926 * omp-expand.c (expand_oacc_for): Handle calls to internal
14927 functions like regular functions.
14929 2018-04-12 Richard Biener <rguenther@suse.de>
14932 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
14933 for the early LTO debug to properly generate references to it
14934 during DIE emission. Do not re-use that for the skeleton for
14936 (dwarf2out_early_finish): Likewise.
14938 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14941 * config/i386/sse.md
14942 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
14943 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
14944 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
14945 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
14946 and output is a reg, avoid creating invalid lowpart subreg, but
14947 instead split into a 512-bit move. Don't split if not AVX512VL,
14948 input is xmm16+ reg and output is a mem.
14949 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
14950 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
14951 xmm16+ reg and output is a mem.
14953 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14955 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
14956 also for flag_dwarf2_cfi_asm.
14958 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14960 PR rtl-optimization/85342
14961 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
14962 a bool scalar var inside of the loop instead. Don't try to update
14963 recog_data.operand after failed apply_change_group.
14965 2018-04-12 Tom de Vries <tom@codesourcery.com>
14968 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
14969 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
14970 array with flexible array member as array without given dimension.
14971 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
14972 argument for undefined param to true.
14974 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
14977 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14978 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
14979 from PowerPC section.
14980 * config/rs6000/sysv4.opt (mcall-): Improve help text.
14981 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
14982 help text that is too long.
14983 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
14984 help text that is too long.
14985 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
14986 help text that is too long.
14988 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
14990 * config/alpha/alpha.md (stack_probe_internal): Rename
14991 from "probe_stack". Update all callers.
14993 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14995 PR rtl-optimization/84566
14996 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
14997 sched_macro_fuse_insns.
14999 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15002 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15003 (compute_block_dependences): ... from here.
15005 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15007 PR tree-optimization/85331
15008 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15009 from int to HOST_WIDE_INT.
15011 2018-04-11 Martin Jambor <mjambor@suse.cz>
15014 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15015 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15016 not the same as the source val.
15017 (cgraph_edge_brings_value_p): New parameter.
15018 (gather_edges_for_value): Pass destination value to
15019 cgraph_edge_brings_value_p.
15020 (perhaps_add_new_callers): Likewise.
15021 (get_info_about_necessary_edges): Likewise and exclude values brought
15022 only by self-recursive edges.
15023 (create_specialized_node): Redirect only clones of self-calling edges.
15024 (+self_recursive_pass_through_p): New function.
15025 (find_more_scalar_values_for_callers_subset): Use it.
15026 (find_aggregate_values_for_callers_subset): Likewise.
15027 (known_aggs_to_agg_replacement_list): Removed.
15028 (decide_whether_version_node): Re-calculate known constants for all
15029 remaining context clones.
15031 2018-04-11 Richard Biener <rguenther@suse.de>
15034 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15035 from early DWARF output.
15036 (dwarf2out_early_finish): Output line info unconditionally into
15037 early DWARF and add reference to it.
15039 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15042 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15043 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15044 other than V2DFmode using iptr mode attribute.
15045 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15047 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15049 PR rtl-optimization/84659
15050 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15052 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15055 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15057 (output_loc_list): Pass address of a dummy size variable even in the
15058 locview handling loop.
15059 (index_location_lists): Add comment on why skip_loc_list_entry can't
15062 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15065 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15068 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15071 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15072 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15073 and -mstring-compare-inline-limit.
15075 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15078 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15079 for stack clash protection in a register whenever we need it to be in
15082 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15084 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15085 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15087 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15090 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15092 (mlong-double-): Ditto.
15093 * config/rs6000/sysv4.opt (msdata=): Ditto.
15094 (mtls-size=): Ditto.
15096 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15098 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15099 erroneous entries for
15100 "vector int vec_ldl (int, long int *)", and
15101 "vector unsigned int vec_ldl (int, unsigned long int *)".
15102 Add comments and entries for
15103 "vector bool char vec_ldl (int, bool char *)",
15104 "vector bool short vec_ldl (int, bool short *)",
15105 "vector bool int vec_ldl (int, bool int *)",
15106 "vector bool long long vec_ldl (int, bool long long *)",
15107 "vector pixel vec_ldl (int, pixel *)",
15108 "vector long long vec_ldl (int, long long *)",
15109 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15110 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15111 type tree bool_long_long_type_node and correct definition of
15112 bool_V2DI_type_node to make reference to this new type tree.
15113 (rs6000_mangle_type): Replace erroneous reference to
15114 bool_long_type_node with bool_long_long_type_node.
15115 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15116 comments to emphasize sign distinctions for char and int types and
15117 replace RS6000_BTI_bool_long constant with
15118 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15119 use of RS6000_BTI_pixel.
15120 (bool_long_type_node): Remove this macro definition.
15121 (bool_long_long_type_node): New macro definition
15123 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15125 PR rtl-optimization/85300
15126 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15127 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15128 simplify_unary_operation fails.
15130 2018-04-10 Martin Liska <mliska@suse.cz>
15132 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15133 cgraph_edge and ipa_ref.
15135 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15139 * config/i386/sse.md
15140 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15141 computation of the VEC_MERGE selector from mask.
15142 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15143 Fix decoding of the VEC_MERGE selector into mask.
15145 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15147 PR tree-optimization/85286
15148 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15150 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15152 * final.c (final_1): Set insn_last_address as well as
15153 insn_current_address.
15155 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15158 * explow.c (emit_stack_probe): Call validize_mem on memory location
15159 before passing it to gen_probe_stack. Create address operand and
15160 legitimize it for the probe_stack_address case.
15162 2018-04-09 Jan Hubicka <jh@suse.cz>
15165 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15166 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15167 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15169 (free_lang_data): Rebuild type inheritance graph.
15171 2018-04-09 Martin Sebor <msebor@redhat.com>
15173 * invoke.texi (-finline-small-functions): Mention other optimization
15175 (-findirect-inlining, -fpartial-inlining): Same.
15176 (-finline-functions-called-once): Same.
15177 (-freorder-blocks-and-partition): Same.
15179 2018-04-09 Jan Hubicka <jh@suse.cz>
15182 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15183 jumps; choose last target that matches the criteria (i.e.
15184 no partition changes for non-crossing jumps).
15185 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15186 support for redirecting crossing jumps to non-crossing.
15188 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15190 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15191 also for naked functions.
15193 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15195 * config/arc/arc.md (add_shift): New pattern.
15196 (add_shift2): Likewise.
15197 (sub_shift): Likewise.
15198 (sub_shift_cmp0_noout): Likewise.
15199 (compare_si_ashiftsi): Likewise.
15200 (xbfu_cmp0_noout): New combine pattern.
15201 (xbfu_cmp0"): Likewise.
15202 (movsi_set_cc_insn): Place the predicable variant first.
15203 (commutative_binary_cmp0_noout): Remove clobber.
15204 (commutative_binary_cmp0): New pattern.
15205 (noncommutative_binary_cmp0): Likewise.
15206 (noncommutative_binary_cmp0_noout): Likewise.
15207 (noncommutative_binary_comparison_result_used): Removed.
15208 (rsub_cmp0): New pattern.
15209 (rsub_cmp0_noout): Likewise.
15210 (extzvsi): Changed, keep only meaningful variants.
15211 (SQH, SEZ): New iterators.
15212 (SQH_postfix): New mode attribute.
15213 (SEZ_prefix): New code attribute.
15214 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15215 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15216 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15217 of numerical value.
15218 (noncommutative_operator): Check the availability of barrel
15221 2018-04-09 Richard Biener <rguenther@suse.de>
15223 PR tree-optimization/85284
15224 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15225 Only use the niter constraining form of simple_iv when the exit
15226 is always executed.
15228 2018-04-09 Tom de Vries <tom@codesourcery.com>
15231 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15232 (define_expand "*memory_barrier"): New define_expand.
15233 (define_insn "memory_barrier"): New insn.
15235 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15237 PR rtl-optimization/80463
15238 PR rtl-optimization/83972
15239 PR rtl-optimization/83480
15241 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15242 correct producer for the insn.
15243 (tidy_control_flow): Fixup seqnos in case of debug insns.
15245 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15247 PR rtl-optimization/83913
15249 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15250 different sched-times when merging exprs.
15252 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15254 PR rtl-optimization/83962
15256 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15257 tidy_fallthru_edge and tidy_control_flow.
15259 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15261 PR rtl-optimization/83530
15263 * sel-sched.c (force_next_insn): New global variable.
15264 (remove_insn_for_debug): When force_next_insn is true, also leave only
15265 next insn in the ready list.
15266 (sel_sched_region): When the region wasn't scheduled, make another pass
15267 over it with force_next_insn set to 1.
15269 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15271 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15273 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15274 for interrupt control.
15275 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15276 functions for interrupt control.
15277 * config/nds32/nds32-intrinsic.md: Likewise.
15278 * config/nds32/nds32_intrinsic.h: Likewise.
15279 * config/nds32/nds32.h (nds32_builtins): Likewise.
15281 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15283 * config/nds32/nds32.c (nds32_init_machine_status,
15284 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15285 strict_aligned_p field.
15286 (nds32_expand_to_rtl_hook): New function.
15287 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15288 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15290 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15291 Chung-Ju Wu <jasonwucj@gmail.com>
15293 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15294 * config/nds32/nds32-n7.md: New file.
15295 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15296 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15298 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15299 * config/nds32/nds32.md (pipeline_model): Add n7.
15300 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15301 * config/nds32/pipelines.md: Include n7 settings.
15303 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15304 Chung-Ju Wu <jasonwucj@gmail.com>
15306 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15307 * config/nds32/nds32-e8.md: New file.
15308 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15309 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15311 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15312 * config/nds32/nds32.md (pipeline_model): Add e8.
15313 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15314 * config/nds32/pipelines.md: Include e8 settings.
15316 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15317 Chung-Ju Wu <jasonwucj@gmail.com>
15319 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15320 * config/nds32/nds32-n8.md: New file.
15321 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15322 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15324 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15325 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15326 * config/nds32/nds32.md (pipeline_model): Add n8.
15327 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15328 * config/nds32/pipelines.md: Include n8 settings.
15330 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15331 Chung-Ju Wu <jasonwucj@gmail.com>
15333 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15334 * config/nds32/nds32-n9-2r1w.md: New file.
15335 * config/nds32/nds32-n9-3r2w.md: New file.
15336 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15337 nds32_register_ports): New or modify for cpu n9.
15338 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15340 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15341 * config/nds32/nds32-utils.c: New file.
15342 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15343 TARGET_MUL_SLOW): Define.
15344 * config/nds32/nds32.md (pipeline_model): New attribute.
15345 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15346 New options that support cpu n9.
15347 * config/nds32/pipelines.md: Include n9 settings.
15348 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15350 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15352 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15353 information if necessary.
15354 (output_cond_branch_compare_zero): Likewise.
15355 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15356 (nds32_target_alignment): Refine for alignment.
15357 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15358 (FUNCTION_BOUNDARY): Modify.
15359 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15361 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15363 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15365 * config/nds32/constants.md (unspec_volatile_element): Add values for
15366 TLB operation and data prefetch.
15367 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15368 functions for TLB operation and data prefetch.
15369 * config/nds32/nds32-intrinsic.md: Likewise.
15370 * config/nds32/nds32_intrinsic.h: Likewise.
15371 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15372 (nds32_print_operand): Likewise.
15373 * config/nds32/nds32.h (nds32_builtins): Likewise.
15375 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15376 Andrew Pinski <pinsika@gcc.gnu.org>
15378 PR middle-end/82976
15379 * match.pd: Use constant_boolean_node of correct type instead of
15380 boolean_true_node or boolean_false_node for simplifying
15381 pointer comparisons to zero.
15383 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15385 PR tree-optimization/80021
15386 * tree.c (verify_type_variant): Make error call in verify_variant_match
15387 translatable and remove final full stop.
15389 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15391 * config/nds32/constants.md (unspec_volatile_element): Add
15392 UNSPEC_VOLATILE_EH_RETURN.
15393 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15394 nds32_output_stack_pop): Support dwarf exception handling process.
15395 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15396 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15397 exception handling process.
15398 (nds32_compute_stack_frame): Likewise.
15399 (nds32_return_addr_rtx): Likewise.
15400 (nds32_initial_elimination_offset): Likewise.
15401 (nds32_expand_prologue): Likewise.
15402 (nds32_expand_epilogue): Likewise.
15403 (nds32_dynamic_chain_address): New function.
15404 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15405 exception handling.
15406 (DYNAMIC_CHAIN_ADDRESS): Define.
15407 (EH_RETURN_DATA_REGNO): Define.
15408 (EH_RETURN_STACKADJ_RTX): Define.
15409 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15410 patterns for dwarf exception handling.
15412 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15414 * config/nds32/nds32.h: Clean up obsolete macros.
15416 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15418 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15419 Add enum values for particular instructions.
15420 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15421 particular intrinsic functions.
15422 * config/nds32/nds32-intrinsic.md: Likewise.
15423 * config/nds32/nds32_intrinsic.h: Likewise.
15424 * config/nds32/nds32.h (nds32_builtins): Likewise.
15425 * config/nds32/nds32.md (type): Add pbsad and pbsada.
15426 (btst, ave): New patterns for particular instructions.
15428 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15430 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15431 Add enum values for atomic load/store and memory sync.
15432 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
15434 * config/nds32/nds32-intrinsic.md: Likewise.
15435 * config/nds32/nds32_intrinsic.h: Likewise.
15436 * config/nds32/nds32.h (nds32_builtins): Likewise.
15438 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15440 PR tree-optimization/85257
15441 * fold-const.c (native_encode_vector): If not all elts could fit
15442 and off is -1, return 0 rather than offset.
15443 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
15444 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
15445 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
15446 adjust buffer in native_interpret_expr call.
15448 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15450 * config/nds32/constants.md (unspec_volatile_element): Add cache
15451 control enum values.
15452 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15453 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15454 * config/nds32/nds32.c (nds32_cctl_names): New.
15455 (nds32_print_operand): Handle cache control register names.
15456 * config/nds32/nds32.h (nds32_builtins): New enum values.
15457 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
15459 * config/nds32/nds32.md (type): Add mmu.
15460 * config/nds32/pipelines.md (simple_insn): Add mmu.
15462 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15464 * config/nds32/nds32.md (type): Remove call.
15465 * config/nds32/pipelines.md (simple_insn): Likewise.
15467 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15469 * config/nds32/constants.md (unspec_volatile_element): Add
15470 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
15471 UNSPEC_VOLATILE_FMFCFG.
15472 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
15473 description for fmfcfg and fmfcsr.
15474 (bdesc_1arg): Add fmtcsr.
15475 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
15476 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
15477 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
15478 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
15479 unspec_fmfcfg): New patterns.
15480 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
15481 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
15482 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
15483 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
15484 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
15485 __nds32__fmfcfg): Define.
15487 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15489 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
15490 intrinsic register names.
15491 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
15492 intrinsic register enum values and macros.
15494 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15496 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
15497 for load/store addressing form.
15498 (nds32_print_operand_address): Likewise.
15500 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
15503 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
15504 based on LABEL_REF. Remove useless assertion.
15505 (pic_address_needs_scratch): Fix formatting.
15506 (sparc_legitimize_pic_address): Minor tweaks.
15507 (sparc_delegitimize_address): Adjust assertion accordingly.
15508 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
15509 into symbolic_operand.
15510 (movsi_high_pic_label_ref): Likewise.
15511 (movsi_lo_sum_pic_label_ref): Likewise.
15512 (movdi_pic_label_ref): Likewise.
15513 (movdi_high_pic_label_ref): Likewise.
15514 (movdi_lo_sum_pic_label_ref): Likewise.
15516 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
15518 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
15519 custom LIB_SPEC setup.
15521 2018-04-06 Ruslan Bukin <br@bsdpad.com>
15522 Kito Cheng <kito.cheng@gmail.com>
15524 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
15525 * config/riscv/freebsd.h: New.
15527 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15529 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
15530 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
15533 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15534 Kito Cheng <kito.cheng@gmail.com>
15536 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
15537 nds32_output_call, nds32_symbol_binds_local_p): New functions.
15538 * config/nds32/nds32-protos.h (nds32_output_call,
15539 nds32_output_return): Declare.
15540 * config/nds32/nds32.md: Refine all the call and return patterns.
15542 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15545 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
15546 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
15548 PR rtl-optimization/84872
15549 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
15550 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
15551 EDGE_CROSSING edge.
15553 2018-04-06 Tamar Christina <tamar.christina@arm.com>
15555 * expr.c (copy_blkmode_to_reg): Revert 254862.
15556 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
15558 2018-04-06 Richard Biener <rguenther@suse.de>
15560 PR middle-end/85244
15561 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
15562 after seeing a component reference with an adjacent field. Treat
15563 refs to arrays at struct end of external decls similar to
15564 refs to unconstrained commons.
15566 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15569 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
15570 look through SAVE_EXPRs with non-side-effects argument. Adjust
15572 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
15575 2018-04-06 Richard Biener <rguenther@suse.de>
15577 PR middle-end/85180
15578 * alias.c (find_base_term): New wrapper around find_base_term
15579 unwinding CSELIB_VAL_PTR changes.
15580 (find_base_term): Do not restore CSELIB_VAL_PTR during the
15583 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15585 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
15587 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
15588 constant definitions.
15589 ("nop"): lr 0,0 -> nopr r0
15590 ("nop_lr0", "nop_lr1"): New insn definitions.
15592 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15594 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
15595 NDS32_V3PUSH_AVAILABLE_P macro.
15597 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
15598 Chung-Ju Wu <jasonwucj@gmail.com>
15600 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15601 (nds32*-*-*): Add float and fpu_config into supported_defaults.
15602 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15603 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15604 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15605 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15606 * config/nds32/constraints.md: New constraints and checking for hard
15607 float configuration.
15608 * config/nds32/iterators.md: New mode iterator and attribute for hard
15609 float configuration.
15610 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15612 * config/nds32/nds32-fpu.md: New file.
15613 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15614 deal with hard float code generation.
15615 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15617 (abi_type, float_reg_number): New enum type.
15618 * config/nds32/nds32-predicates.c: New predicates for hard float.
15619 * config/nds32/nds32-protos.h: Declare functions for hard float.
15620 * config/nds32/nds32.c: Implementation for hard float configuration.
15621 * config/nds32/nds32.h: Definitions for hard float configuration.
15622 * config/nds32/nds32.md: Include hard float machine description and
15623 modify patterns for hard float configuration.
15624 * config/nds32/nds32.opt: New options for hard float configuration.
15625 * config/nds32/predicates.md: New predicates for hard float
15628 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15630 * common/config/nds32/nds32-common.c
15631 (nds32_option_optimization_table): Enable -mreleax-hint by default.
15633 2018-04-05 Jakub Jelinek <jakub@redhat.com>
15635 PR middle-end/85195
15636 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15637 CONSTRUCTOR_ELT (ctor, ...)->value.
15639 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
15642 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15644 2018-04-05 Tom de Vries <tom@codesourcery.com>
15647 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15650 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15651 Kito Cheng <kito.cheng@gmail.com>
15653 * config/nds32/constraints.md (U33): Fine-tune checking condition.
15654 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15655 * config/nds32/nds32.h (nds32_16bit_address_type): Add
15656 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15658 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15659 Kito Cheng <kito.cheng@gmail.com>
15661 * config/nds32/constraints.md (Ufe): New memory constraint.
15662 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15663 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15664 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15666 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15667 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15669 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15671 * config/nds32/nds32.md: Use optimize_size in the condition for
15672 alu-shift instructions.
15674 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15676 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15678 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15680 * config/nds32/nds32.md (negsi2): Refine pattern.
15682 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15683 Chung-Ju Wu <jasonwucj@gmail.com>
15685 * config/nds32/iterators.md (shift_rotate): New code iterator.
15686 (shift): New code attribute.
15687 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15688 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15689 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15690 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15691 bit-wise operations.
15692 (andsi3, *andsi3): Ditto.
15693 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15694 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15695 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15696 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15697 nds32_ior_operand, nds32_xor_operand): New predicates.
15699 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15701 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15702 (addsi3, subsi3): ... this.
15704 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15706 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15708 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15710 * config/nds32/nds32.md: Adjust indention.
15712 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15714 * config/nds32/nds32.md (feature): New attribute.
15716 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15718 * config/nds32/nds32.md (subtype): New attribute.
15720 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
15723 * config/arm/arm-builtins.c (arm_expand_builtin): Change
15724 expansion to perform a bitwise AND of the argument followed by a
15725 boolean negation of the result.
15727 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
15729 PR rtl-optimization/84878
15730 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15731 the basic block. Assert the use reference is not artificial and that
15732 it has an associated insn.
15734 2018-04-04 Michael Matz <matz@suse.de>
15736 * builtins.c (compute_objsize): Pass correct operand
15737 to array_at_struct_end_p.
15739 2018-04-04 Richard Biener <rguenther@suse.de>
15742 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
15743 from contexts for DINFO_LEVEL_TERSE and below.
15745 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15747 * config/nds32/nds32-doubleword.md (move_<mode>): Require
15748 resiter_operand condition.
15749 * config/nds32/nds32.md (*move<mode>): Ditto.
15751 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15752 Monk Chiang <sh.chiang04@gmail.com>
15754 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
15756 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15758 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
15760 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15761 Kito Cheng <kito.cheng@gmail.com>
15763 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
15764 nds32_cond_code_str, output_cond_branch,
15765 output_cond_branch_compare_zero, nds32_expand_cbranch,
15766 nds32_expand_cstore, nds32_expand_movcc,
15767 nds32_output_cbranchsi4_equality_zero,
15768 nds32_output_cbranchsi4_equality_reg,
15769 nds32_output_cbranchsi4_equality_reg_or_const_int,
15770 nds32_output_cbranchsi4_greater_less_zero: New functions.
15771 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
15772 nds32_expand_cstore, nds32_expand_movcc,
15773 nds32_output_cbranchsi4_equality_zero,
15774 nds32_output_cbranchsi4_equality_reg,
15775 nds32_output_cbranchsi4_equality_reg_or_const_int,
15776 nds32_output_cbranchsi4_greater_less_zero): Declare.
15777 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
15778 nds32_rimm11s_operand): New predicates.
15779 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
15780 * config/nds32/nds32.md: Rewrite all the branch and conditional move
15783 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15785 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
15786 * config/nds32/nds32.md: Ditto.
15787 * config/nds32/pipelines.md: Ditto.
15789 2018-04-04 Richard Biener <rguenther@suse.de>
15791 PR tree-optimization/85168
15792 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
15793 propagating abnormals.
15795 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15797 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
15799 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15800 Kito Cheng <kito.cheng@gmail.com>
15802 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
15803 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
15804 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
15805 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
15806 * config/nds32/nds32.md (sibcall_internal): New.
15807 (sibcall_register): Remove.
15808 (sibcall_immediate): Remove.
15809 (sibcall_value_internal): New.
15810 (sibcall_value_register): Remove.
15811 (sibcall_value_immediate): Remove.
15812 * config/nds32/predicates.md (nds32_general_register_operand): New.
15813 (nds32_call_address_operand): New.
15815 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15817 PR rtl-optimization/85167
15818 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
15819 bb_defs if *split_p, instead preinitialize it to NULL.
15821 PR tree-optimization/85156
15822 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
15823 evaluating the argument multiple times.
15825 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
15827 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
15829 (_mm_cvtpd_ps): Likewise.
15830 (_mm_cvttpd_epi32): Likewise.
15831 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
15832 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
15833 vector, pixel, and bool following altivec.h include.
15835 2018-04-03 Martin Sebor <msebor@redhat.com>
15837 * doc/extend.texi (Common Function Attributes): Clarify.
15838 (const attribute): Likewise.
15839 (pure attribute): Likewise.
15841 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15844 * config/i386/i386.c (ix86_expand_vector_set): Use
15845 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
15847 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
15849 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
15850 instructions when changing rounding bits to preserve precision bits
15851 in the x87 control word.
15853 2018-04-03 Martin Liska <mliska@suse.cz>
15855 PR tree-optimization/82491
15856 * rtl.h (strip_offset_and_add): Replace += suboffset with
15857 poly_uint64 () + suboffset.
15859 2018-03-29 Martin Liska <mliska@suse.cz>
15860 Martin Jambor <mjambor@suse.cz>
15863 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
15864 param_type is not an integral or pointer type.
15866 2018-04-03 Richard Biener <rguenther@suse.de>
15868 * sese.h (recompute_all_dominators): Remove.
15870 2018-04-02 Martin Sebor <msebor@redhat.com>
15872 * doc/invoke.texi (-Wrestrict): Fix typos.
15874 2018-04-02 Jim Wilson <jimw@sifive.com>
15876 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
15877 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
15878 (<optab>di3, <optab>si3_extend): Likewise.
15879 (<optab>si3_mask, <optab>si3_mask_1): New.
15880 (<optab>di3_mask, <optab>di3_mask_1): New.
15881 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
15882 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
15883 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
15885 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
15887 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
15890 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
15892 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
15893 (nds32_canonicalize_comparison): New function.
15895 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15896 Kito Cheng <kito.cheng@gmail.com>
15897 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15899 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
15900 * config/nds32/constants.md (unspec_volatile_element): Add
15901 UNSPEC_VOLATILE_RELAX_GROUP.
15902 * config/nds32/nds32-relax-opt.c: New file.
15903 * config/nds32/nds32-predicates.c
15904 (nds32_symbol_load_store_p): New function.
15905 * config/nds32/nds32-protos.h
15906 (nds32_symbol_load_store_p): Declare function.
15907 (make_pass_nds32_relax_opt): Declare new rtl pass function.
15908 * config/nds32/nds32.c
15909 (nds32_register_pass): New function to register pass.
15910 (nds32_register_passes): New function to register passes.
15911 * config/nds32/nds32.md (relax_group): New pattern.
15912 * config/nds32/nds32.opt (mrelax-hint): New option.
15913 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
15915 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
15917 * config/nds32/t-nds32: Modify files dependency.
15919 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15921 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
15922 (PROFILE_HOOK): Define its implementation.
15924 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15926 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
15927 type and 32-bit size.
15929 2018-04-01 Jakub Jelinek <jakub@redhat.com>
15931 PR middle-end/85090
15932 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
15933 (V_128_256): New mode iterator.
15934 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
15935 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
15936 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
15938 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
15939 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
15941 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
15944 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
15945 NaN inputs correctly.
15947 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
15950 * config/rs6000/vsx.md (??r): New mode attribute.
15951 (*vsx_mov<mode>_64bit): Use it.
15952 (*vsx_mov<mode>_32bit): Likewise.
15954 2018-03-30 Martin Sebor <msebor@redhat.com>
15956 PR tree-optimization/84818
15957 * builtins.c (check_access): Use warning_n.
15959 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15962 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
15964 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
15967 2018-03-30 Julia Koval <julia.koval@intel.com>
15970 * x86-tune.def (movx, partial_reg_dependency): Enable for
15973 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
15975 PR inline-asm/84985
15976 * lra-constraints.c (process_alt_operands): Move setting
15977 this_alternative_matches below.
15979 2018-03-29 Martin Liska <mliska@suse.cz>
15982 * doc/invoke.texi: Document how LTO works with debug info.
15983 Describe auto-load support of binutils. Mention 'x86-64'
15984 as valid option value of -march option.
15986 2018-03-29 Jakub Jelinek <jakub@redhat.com>
15988 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
15991 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
15992 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
15993 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
15996 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15999 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16000 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16001 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16002 for RS6000_BTM_POWERPC64.
16003 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16004 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16005 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16008 (DIVDEU): Likewise.
16010 2018-03-28 Carl Love <cel@us.ibm.com>
16013 2017-09-27 Carl Love <cel@us.ibm.com>
16015 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16016 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16017 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16020 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16022 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16023 instead of __vector bool.
16024 (_mm_max_pu8): Likewise.
16025 (_mm_min_pi16): Likewise.
16027 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16030 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16031 (DIVWEUO): Likewise.
16032 (DIVDEO): Likewise.
16033 (DIVDEUO): Likewise.
16034 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16035 DIVWEUO and DIVDEUO.
16036 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16037 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16038 (div_extend): Likewise.
16039 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16041 (__builtin_divweuo): Likewise.
16042 (__builtin_divdeo): Likewise.
16043 (__builtin_divdeuo): Likewise.
16045 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16048 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16049 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16051 PR tree-optimization/82004
16052 * gimple-match-head.c (optimize_pow_to_exp): New function.
16053 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16054 Don't fold to exp if optimize_pow_to_exp is false.
16056 2018-03-28 Martin Liska <mliska@suse.cz>
16059 * calls.c (initialize_argument_information): Fix trailing space.
16060 * common.opt: Fix typo and provide better explanation for
16061 -fsanitize-coverage option.
16062 * config/i386/i386.opt: Fix typo.
16064 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16065 Martin Liska <mliska@suse.cz>
16068 * gimplify.c (asan_poison_variable): Don't do the check for
16069 gimplify_omp_ctxp here.
16070 (gimplify_decl_expr): Do it here.
16071 (gimplify_target_expr): Likewise.
16073 2018-03-28 Martin Liska <mliska@suse.cz>
16076 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16077 chkp_type_bounds_count if MPX is not enabled.
16079 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16081 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16083 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16086 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16087 function to create the function decl for complex long double
16088 multiply and divide for -mabi=ieeelongdouble.
16089 (init_float128_ieee): Call it.
16091 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16094 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16095 -fcf-protection=branch -mibt.
16096 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16098 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16101 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16102 arm_word_relocations.
16104 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16107 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16108 extern array declarations.
16110 2018-03-27 Richard Biener <rguenther@suse.de>
16112 PR middle-end/84067
16113 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16114 explicit single_use checks.
16116 2018-03-27 Richard Biener <rguenther@suse.de>
16118 PR tree-optimization/85082
16119 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16122 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16124 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16125 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16126 Turn on fasynchronous-unwind-tables and funwind-tables.
16128 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16131 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16132 (*bmi_blsr_<mode>_ccz): Ditto.
16134 2018-03-26 Tom de Vries <tom@codesourcery.com>
16136 PR tree-optimization/85063
16137 * omp-general.c (offloading_function_p): New function. Factor out
16139 * omp-offload.c (pass_omp_target_link::gate): ... here.
16140 * omp-general.h (offloading_function_p): Declare.
16141 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16142 with attribute omp declare target for offloading functions.
16144 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16146 PR tree-optimization/84005
16147 * tree-data-ref.h (get_base_for_alignment): Declare.
16148 * tree-data-ref.c (get_base_for_alignment_1): New function.
16149 (get_base_for_alignment): Likewise.
16150 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16151 get_base_for_alignment to find a suitable base object, instead
16152 of always using drb->base_address.
16154 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16156 PR inline-asm/85022
16157 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16158 known size by default.
16160 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16162 PR inline-asm/85030
16163 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16164 and non BLKmode operands.
16166 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16169 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16170 Clean up attributes.
16172 2018-03-23 Richard Biener <rguenther@suse.de>
16175 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16176 we are going to emit early debug for LTO.
16178 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16180 PR inline-asm/85034
16181 * function.c (match_asm_constraints_1): Don't optimize if input
16182 doesn't satisfy general_operand predicate for output's mode.
16184 PR inline-asm/85022
16185 * alias.c (write_dependence_p): Don't require for x_canonicalized
16186 non-VOIDmode if x has VOIDmode.
16189 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16190 just don't try to optimize it rather than assert it never happens.
16192 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16194 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16195 macro expansions for definition of ST_INTERNAL_<mode> and
16196 LD_INTERNAL_<mode> builtins.
16197 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16199 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16201 (altivec_expand_st_builtin): Likewise.
16202 (altivec_expand_builtin): Remove calls to deleted functions.
16203 (rs6000_address_for_altivec): Delete this function.
16204 * config/rs6000/vector.md: Remove expands for
16205 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16207 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16210 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16211 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16212 re-computing once computed.
16213 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16214 (arm_init_machine_status): Initialize
16215 machine->static_chain_stack_bytes.
16217 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16220 * doc/extend.texi: Add four new prototypes for vec_ld.
16221 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16222 definitions for more logical presentation.
16223 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16224 entries for V1TI variants of __builtin_altivec_ld builtin.
16225 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16226 handling of V1TI variant of LVX icode pattern.
16227 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16228 (rs6000_gimple_fold_builtin): Likewise.
16229 (altivec_init_builtins): Add code to define
16230 __builtin_altivec_lvx_v1ti function.
16232 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16234 PR inline-asm/84941
16235 * function.c (match_asm_constraints_1): Don't do the optimization
16236 if input isn't a REG, SUBREG, MEM or constant.
16238 2018-03-22 Tom de Vries <tom@codesourcery.com>
16240 PR tree-optimization/84956
16241 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16242 bb_has_abnormal_pred.
16244 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16247 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16248 DECL_INITIAL (decl) to decl at the end.
16249 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16250 adjust the comment.
16252 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16254 * doc/extend.texi (__builtin_tgmath): Document when complex
16255 integer types are treated as _Complex _Float64.
16257 2018-03-21 Tom de Vries <tom@codesourcery.com>
16259 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16261 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16263 PR tree-optimization/84960
16264 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16265 if it is ENTRY block, move them into single succ of ENTRY in that case.
16267 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16269 PR tree-optimization/84811
16270 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16271 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16272 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16273 (known_subrange_p): Update accordingly. Cast each value involved
16274 in the size comparison, rather than casting the result of the
16277 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16279 PR tree-optimization/84982
16280 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16281 by flipping the least significant bit rather than all bits from
16282 bitpos to bitpos + bitsize - 1.
16284 2018-03-21 Nathan Sidwell <nathan@acm.org>
16286 * doc/extend.texi (Deprecated Features): Remove mention of
16287 long-deleted deprecations.
16289 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16292 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16293 * configure: Regenerate.
16295 2018-03-21 Tom de Vries <tom@codesourcery.com>
16297 PR tree-optimization/83126
16298 * tree-parloops.c (num_phis): New function.
16299 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16301 2018-03-21 Nathan Sidwell <nathan@acm.org>
16303 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16304 mention anon-struct/union members and trailing attributes.
16306 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16308 PR tree-optimization/84969
16309 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16310 builtin memset partitions if they set different rhs values.
16312 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16314 PR rtl-optimization/84989
16315 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16316 VEC_DUPLICATE with scalar result mode.
16318 2018-03-21 Martin Liska <mliska@suse.cz>
16321 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16322 not intended return statement.
16324 2018-03-21 Martin Liska <mliska@suse.cz>
16327 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16328 (chkp_find_bound_slots_1): Limit number of iterations.
16330 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16333 * Minor grammar fixes for x86 options.
16335 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16338 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16339 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16341 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16344 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16345 (altivec_lvx_<mode>_1op): Likewise.
16346 (altivec_stvx_<mode>_2op): Likewise.
16347 (altivec_stvx_<mode>_1op): Likewise.
16348 (altivec_lvx_<VM2:mode>): New define_expand.
16349 (altivec_stvx_<VM2:mode>): Likewise.
16350 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16351 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16352 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16353 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16354 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16355 (rs6000_gen_lvx): Likewise.
16356 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16357 (altivec_expand_stv_builtin): Likewise.
16358 (altivec_expand_builtin): Likewise.
16359 * config/rs6000/vector.md: Likewise.
16361 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16364 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16367 2018-03-20 Richard Biener <rguenther@suse.de>
16370 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16371 sign-conversions as zero, fall back to standard scalar_stmt
16374 2018-03-20 Martin Liska <mliska@suse.cz>
16377 * predict.c (rebuild_frequencies): Handle case when we have
16378 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16380 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16383 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16384 flag_section_anchors.
16385 * varasm.c (use_blocks_for_decl_p): Remove hack for
16386 dw2_force_const_mem.
16389 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16391 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16392 be created, use lowpart_subreg of operands[0] rather than operands[0]
16394 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16395 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16396 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16397 and n constraint instead of aarch64_shift_imm_di and Usd.
16398 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16399 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16401 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16404 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16405 to favor GPR over NEON registers.
16406 (<shift>di3_neon): Likewise.
16408 2018-03-20 Tom de Vries <tom@codesourcery.com>
16411 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16412 (nvptx_process_pars): Emit bar.sync asap and alap.
16414 2018-03-20 Tom de Vries <tom@codesourcery.com>
16417 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16418 seen_label if seen_label is already set.
16420 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16423 * config/i386/i386.c (fold_builtin_cpu): For features above 31
16424 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16425 Use 1U instead of 1. Formatting fixes.
16428 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
16429 instead of TREE_TYPE (s1) for the return value.
16431 2018-03-19 Jakub Jelinek <jakub@redhat.com>
16433 PR tree-optimization/84946
16434 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
16435 bitsize + bitsize in poly_uint64 rather than poly_int64.
16438 * dwarf2asm.c: Include fold-const.c.
16439 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
16440 of decl rather than decl itself.
16442 PR rtl-optimization/84643
16443 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
16445 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
16448 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
16449 calling assemble_variable.
16451 2018-03-19 Sudakshina Das <sudi.das@arm.com>
16454 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
16455 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
16457 2018-03-19 Jim Wilson <jimw@sifive.com>
16460 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
16461 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
16462 (riscv_first_stack_step): Likewise.
16463 (riscv_option_override): Use STACK_BOUNDARY instead of
16464 MIN_STACK_BOUNDARY.
16465 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
16466 MIN_STACK_BOUNDARY.
16467 (BIGGEST_ALIGNMENT): Set to 128.
16468 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
16469 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
16472 2018-03-19 Richard Biener <rguenther@suse.de>
16474 PR tree-optimization/84933
16475 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
16476 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
16478 2018-03-19 Richard Biener <rguenther@suse.de>
16480 PR tree-optimization/84859
16481 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
16482 (cond_if_else_store_replacement): Perform sinking operation on
16483 single-store BBs regardless of MAX_STORES_TO_SINK setting.
16484 Generalize what a BB with a single eligible store is.
16486 2018-03-19 Richard Biener <rguenther@suse.de>
16488 PR tree-optimization/84929
16489 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
16490 chrec_is_positive against non-chrec arg.
16492 2018-03-19 Tamar Christina <tamar.christina@arm.com>
16495 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
16497 2018-03-18 Martin Liska <mliska@suse.cz>
16499 PR rtl-optimization/84635
16500 * regrename.c (build_def_use): Use matches_mode only when
16503 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
16505 PR tree-optimization/84913
16506 * tree-vect-loop.c (vectorizable_reduction): Don't try to
16507 vectorize chains of COND_EXPRs.
16509 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16511 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
16513 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16515 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
16517 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16519 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16521 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16522 Kito Cheng <kito.cheng@gmail.com>
16524 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
16525 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
16526 (nds32_adjust_reg_alloc_order): New function.
16527 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
16529 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
16531 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
16532 nds32_print_operand, nds32_print_operand_address): Use
16533 HOST_WIDE_INT_PRINT_DEC instead.
16535 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16537 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
16539 2018-03-17 Jakub Jelinek <jakub@redhat.com>
16542 * config/i386/i386.c (initial_ix86_tune_features,
16543 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
16544 unsigned long long.
16545 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
16546 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
16547 rather than 1u << ix86_tune. Formatting fix.
16548 (ix86_option_override_internal): Change ix86_arch_mask from
16549 unsigned int to unsigned HOST_WIDE_INT, initialize to
16550 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
16551 (ix86_function_specific_restore): Likewise.
16553 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16556 * postreload.c (reload_combine_recognize_pattern): Perform
16557 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
16558 truncate_int_for_mode the result for the destination's mode.
16561 * hsa-gen.c (mem_type_for_type): Fix comment typo.
16562 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
16564 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16567 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
16570 * lra-assigns.c (lra_split_hard_reg_for): Don't use
16571 regno_allocno_class_array and sorted_pseudos.
16572 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
16573 insns where regno is used.
16575 2018-03-16 Martin Liska <mliska@suse.cz>
16578 * multiple_target.c (create_dispatcher_calls): Redirect
16579 reference in the symbol table.
16581 2018-03-16 Martin Liska <mliska@suse.cz>
16584 * multiple_target.c (create_dispatcher_calls): Redirect also
16587 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16591 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
16592 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
16593 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
16595 2018-03-16 Julia Koval <julia.koval@intel.com>
16597 * doc/invoke.texi (Skylake Server): Add CLWB.
16598 Cannonlake): Remove CLWB.
16600 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16602 PR tree-optimization/84841
16603 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16605 (FLOAT_ONE_CONST_TYPE): Define.
16606 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16607 (sort_by_operand_rank): Put entries with higher constant_type last
16608 rather than first to match comments.
16610 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
16612 * config/nios2/nios2.md (movsi_internal): Fix thinko in
16615 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16618 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16619 check and use address of target always.
16621 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
16624 * config/i386/i386.c (indirect_thunk_needed): Update comments.
16625 (indirect_thunk_bnd_needed): Likewise.
16626 (indirect_thunks_used): Likewise.
16627 (indirect_thunks_bnd_used): Likewise.
16628 (indirect_return_needed): New.
16629 (indirect_return_bnd_needed): Likewise.
16630 (output_indirect_thunk_function): Add a bool argument for
16632 (output_indirect_thunk_function): Don't generate alias for
16633 function return thunk.
16634 (ix86_code_end): Call output_indirect_thunk_function to generate
16635 function return thunks.
16636 (ix86_output_function_return): Set indirect_return_bnd_needed
16637 and indirect_return_needed instead of indirect_thunk_bnd_needed
16638 and indirect_thunk_needed.
16640 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
16642 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16643 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16644 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16646 2018-03-15 David Malcolm <dmalcolm@redhat.com>
16647 Paul Hua <paul.hua.gm@gmail.com>
16650 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16652 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
16654 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16655 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16656 resp. SFmode cases.
16658 2018-03-15 Tamar Christina <tamar.christina@arm.com>
16661 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16662 instead of GET_MODE_SIZE when comparing Units.
16664 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16667 * varasm.c (hash_section): Return an unchangeble hash value
16668 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16669 Return !aarch64_can_use_per_function_literal_pools_p ().
16671 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16674 * optabs.c (emit_conditional_move): Pass address of cmode's copy
16675 rather than address of cmode as last argument to prepare_cmp_insn.
16677 2018-03-15 Julia Koval <julia.koval@intel.com>
16679 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16680 F_AVX512VNNI, F_AVX512BITALG): New.
16682 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
16685 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
16686 insn for floating-point loads and stores.
16688 2018-03-14 Carl Love <cel@us.ibm.com>
16690 * config/rs6000/rs6000-c.c: Add macro definitions for
16691 ALTIVEC_BUILTIN_VEC_PERMXOR.
16692 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16693 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16694 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16695 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16697 * config/doc/extend.texi: Add prototypes for vec_permxor.
16699 2018-03-14 David Malcolm <dmalcolm@redhat.com>
16702 * diagnostic-show-locus.c (class layout_point): Convert m_line
16703 from int to linenum_type.
16704 (line_span::comparator): Use linenum "compare" function when
16705 comparing line numbers.
16706 (test_line_span): New function.
16707 (layout_range::contains_point): Convert param "row" from int to
16709 (layout_range::intersects_line_p): Likewise.
16710 (layout::will_show_line_p): Likewise.
16711 (layout::print_source_line): Likewise.
16712 (layout::should_print_annotation_line_p): Likewise.
16713 (layout::print_annotation_line): Likewise.
16714 (layout::print_leading_fixits): Likewise.
16715 (layout::annotation_line_showed_range_p): Likewise.
16716 (struct line_corrections): Likewise for field m_row.
16717 (line_corrections::line_corrections): Likewise for param "row".
16718 (layout::print_trailing_fixits): Likewise.
16719 (layout::get_state_at_point): Likewise.
16720 (layout::get_x_bound_for_row): Likewise.
16721 (layout::print_line): Likewise.
16722 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16723 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16724 * input.c (selftest::test_linenum_comparisons): New function.
16725 (selftest::input_c_tests): Call it.
16726 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16727 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16728 * selftest.h (ASSERT_GT): New macro.
16729 (ASSERT_GT_AT): New macro.
16730 (ASSERT_LT): New macro.
16731 (ASSERT_LT_AT): New macro.
16733 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
16735 PR rtl-optimization/84780
16736 * combine.c (distribute_links): Don't make a link based on pc_rtx.
16738 2018-03-14 Martin Liska <mliska@suse.cz>
16740 * tree.c (record_node_allocation_statistics): Use
16741 get_stats_node_kind.
16742 (get_stats_node_kind): New function extracted from
16743 record_node_allocation_statistics.
16744 (free_node): Use get_stats_node_kind.
16746 2018-03-14 Richard Biener <rguenther@suse.de>
16748 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
16749 that the value-set of ANTIC_IN doesn't grow.
16752 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16754 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16755 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16757 2018-03-14 Julia Koval <julia.koval@intel.com>
16759 * config.gcc (icelake-client, icelake-server): New.
16761 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
16762 (initial_ix86_arch_features): Ditto.
16763 (PTA_SKYLAKE): Add SGX.
16764 (PTA_ICELAKE): Remove.
16765 (PTA_ICELAKE_CLIENT): New.
16766 (PTA_ICELAKE_SERVER): New.
16767 (ix86_option_override_internal): Split up icelake on icelake client and
16769 (get_builtin_code_for_version): Ditto.
16770 (fold_builtin_cpu): Ditto.
16771 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
16772 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
16773 * config/i386/i386.h (processor_type): Ditto.
16774 * doc/invoke.texi: Ditto.
16776 2018-03-14 Jakub Jelinek <jakub@redhat.com>
16779 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
16780 INTEGER_CST offset, add it together with bitpos / 8 and
16781 sign extend based on POINTER_SIZE.
16785 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
16788 * config/i386/constraints.md (Yc): New register constraint.
16789 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
16790 Use Yc constraint for alternative 2 of operand 0. Remove
16791 preferred_for_speed attribute.
16793 2018-03-14 Richard Biener <rguenther@suse.de>
16795 PR tree-optimization/84830
16796 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
16797 with the old one to avoid oscillations.
16799 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
16802 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
16804 (assign_by_spills): Return a flag of reload assignment failure.
16805 Do not process the reload assignment failures. Do not spill other
16806 reload pseudos if they has the same reg class. Update n if
16808 (lra_assign): Add a return arg. Set up from the result of
16809 assign_by_spills call.
16810 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16811 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16812 usage_insns if it is not NULL.
16813 (spill_hard_reg_in_range): New function.
16814 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16815 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16816 function prototypes.
16817 (lra_assign): Change prototype.
16818 * lra.c (lra): Add code to deal with fails by splitting hard reg
16821 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
16823 * config/riscv/riscv.opt (mrelax): New option.
16824 * config/riscv/riscv.c (riscv_file_start): Emit ".option
16825 "norelax" when riscv_mrelax is disabled.
16826 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
16828 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16831 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
16832 reassociation for int modes.
16834 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16836 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
16837 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
16839 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
16840 * config/aarch64/aarch64-sve.md
16841 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
16842 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
16843 (*extend<mode><Vwide>2): Rename to...
16844 (aarch64_sve_extend<mode><Vwide>2): ...this.
16845 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
16846 renaming the old pattern to...
16847 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
16849 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
16850 define_expand, renaming the old pattern to...
16851 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
16852 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
16853 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
16854 account when deciding which SVE instruction the optab should use.
16855 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
16857 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16859 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
16860 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
16861 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
16862 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
16863 (tlsdesc_small_<mode>): Turn a define_expand and use
16864 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
16865 (tlsdesc_small_advsimd_<mode>): ...this.
16866 (tlsdesc_small_sve_<mode>): New pattern.
16868 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16870 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
16871 (UNSPEC_UMUL_HIGHPART): New constants.
16872 (MUL_HIGHPART): New int iteraor.
16873 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
16874 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
16876 (*<su>mul<mode>3_highpart): New define_insn.
16878 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16881 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16884 2018-03-13 Martin Liska <mliska@suse.cz>
16887 * (sem_item_optimizer::sem_item_optimizer): Initialize new
16889 (sem_item_optimizer::~sem_item_optimizer): Release it.
16890 (sem_item_optimizer::merge_classes): Register variable aliases.
16891 (sem_item_optimizer::fixup_pt_set): New function.
16892 (sem_item_optimizer::fixup_points_to_sets): Likewise.
16893 * ipa-icf.h: Declare new variables and functions.
16895 2018-03-13 Jakub Jelinek <jakub@redhat.com>
16897 PR middle-end/84834
16898 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
16899 integer_pow2p@2 and test integer_pow2p in condition.
16900 (A < 0 ? C : 0): Similarly for @1.
16902 PR middle-end/84831
16903 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
16904 characters starting at p contain '\0' character, don't look beyond
16908 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
16909 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
16912 * reg-stack.c (change_stack): Change update_end var from int to
16913 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
16914 also call set_block_for_insn on the newly added insns and rescan.
16917 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
16918 on the last operand.
16921 * tree.c (stabilize_reference_1): Return save_expr (e) for
16922 STATEMENT_LIST even if it doesn't have side-effects.
16924 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
16926 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
16928 2018-03-12 Renlin Li <renlin.li@arm.com>
16930 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
16931 aarch64_output_scalar_simd_mov_immediate.
16933 2018-03-12 Martin Sebor <msebor@redhat.com>
16935 PR tree-optimization/83456
16936 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
16937 for perfectly overlapping calls to memcpy.
16938 (gimple_fold_builtin_memory_chk): Same.
16939 (gimple_fold_builtin_strcpy): Handle no-warning.
16940 (gimple_fold_builtin_stxcpy_chk): Same.
16941 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
16943 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16945 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
16946 parameter. Use it for SFmode.
16947 (rs6000_function_arg_advance_1): Adjust.
16948 (rs6000_function_arg): Adjust.
16949 (rs6000_gimplify_va_arg): Pass false for that new parameter.
16951 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16953 PR rtl-optimization/84169
16954 PR rtl-optimization/84780
16955 * combine.c (can_combine_p): Check for a 2-insn combination whether
16956 the destination register is used between the two insns, too.
16958 2018-03-12 Richard Biener <rguenther@suse.de>
16960 PR tree-optimization/84803
16961 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
16962 for refs DR analysis didn't process.
16964 2018-03-12 Richard Biener <rguenther@suse.de>
16966 PR tree-optimization/84777
16967 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
16968 force-vectorize loops ignore whether we are optimizing for size.
16970 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
16972 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
16973 (TARGET_MD_ASM_ADJUST): Define.
16975 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
16976 Kito Cheng <kito.cheng@gmail.com>
16977 Chung-Ju Wu <jasonwucj@gmail.com>
16979 * config/nds32/nds32.c (nds32_compute_stack_frame,
16980 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
16981 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
16982 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
16983 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
16984 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
16985 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
16986 * config/nds32/nds32.md (prologue, epilogue): Use macro
16987 NDS32_V3PUSH_AVAILABLE_P to do checking.
16989 2018-03-11 Jakub Jelinek <jakub@redhat.com>
16992 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
16993 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
16994 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
16995 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
16996 addition of most attributes on !orig_type_die or the attribute not
16997 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
16999 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17000 Chung-Ju Wu <jasonwucj@gmail.com>
17002 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17003 __NDS32_VH__ macro.
17004 * config/nds32/nds32.opt (mvh): New option.
17006 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17007 Chung-Ju Wu <jasonwucj@gmail.com>
17009 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17011 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17012 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17015 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17016 Chung-Ju Wu <jasonwucj@gmail.com>
17018 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17020 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17021 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17023 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17024 Kito Cheng <kito.cheng@gmail.com>
17025 Chung-Ju Wu <jasonwucj@gmail.com>
17027 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17028 UNSPEC_FFMISM and UNSPEC_FLMISM.
17029 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17030 for ffb, ffmism and flmism.
17031 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17032 (unspec_ffmism): Ditto.
17033 (unspec_flmism): Ditto.
17034 (nds32_expand_builtin_impl): Check if string extension is available.
17035 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17036 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17038 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17041 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17044 * lra-assigns.c (assign_by_spills): Return a flag of reload
17045 assignment failure. Do not process the reload assignment
17046 failures. Do not spill other reload pseudos if they has the same
17048 (lra_assign): Add a return arg. Set up from the result of
17049 assign_by_spills call.
17050 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17051 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17052 usage_insns if it is not NULL.
17053 (spill_hard_reg_in_range): New function.
17054 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17055 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17056 function prototypes.
17057 (lra_assign): Change prototype.
17058 * lra.c (lra): Add code to deal with fails by splitting hard reg
17061 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17064 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17066 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17069 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17070 (dw_val_node): Add val_symbolic_view.
17071 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17072 (symview_upper_bound): New.
17073 (new_line_info_table): Initialize symviews_since_reset.
17074 (dwarf2out_source_line): Count symviews_since_reset and set
17075 symview_upper_bound.
17076 (dw_val_equal_p): Handle symview.
17077 (add_AT_symview): New.
17078 (print_dw_val): Handle symview.
17079 (attr_checksum, attr_checksum_ordered): Likewise.
17080 (same_dw_val_p, size_of_die): Likewise.
17081 (value_format, output_die): Likewise.
17082 (add_high_low_attributes): Use add_AT_symview for entry_view.
17083 (dwarf2out_finish): Reset symview_upper_bound, clear
17086 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17089 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17090 Add strict argument and use it.
17091 (rs6000_split_multireg_move): Update for new strict argument.
17092 (mem_operand_gpr): Disallow all non-offsettable addresses.
17093 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17095 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17098 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17099 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17100 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17103 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17104 decl, use remap_type if we want to use the type.
17106 2018-03-09 Martin Sebor <msebor@redhat.com>
17108 PR tree-optimization/84526
17109 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17111 (builtin_access::generic_overlap): Be prepared to handle non-array
17114 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17116 PR rtl-optimization/84682
17117 * lra-constraints.c (process_address_1): Check is_address flag
17118 for address constraints.
17119 (process_alt_operands): Likewise.
17120 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17121 preprocess_constraints.
17122 * recog.h (preprocess_constraints): Add oploc parameter.
17124 * recog.c (preprocess_constraints): Test address_operand for
17125 CT_ADDRESS constraints.
17127 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17130 * lra-assigns.c (assign_by_spills): Return a flag of reload
17131 assignment failure. Do not process the reload assignment
17132 failures. Do not spill other reload pseudos if they has the same
17134 (lra_assign): Add a return arg. Set up from the result of
17135 assign_by_spills call.
17136 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17137 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17138 usage_insns if it is not NULL.
17139 (spill_hard_reg_in_range): New function.
17140 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17141 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17142 function prototypes.
17143 (lra_assign): Change prototype.
17144 * lra.c (lra): Add code to deal with fails by splitting hard reg
17147 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17150 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17151 Accept complain bool parameter. Only emit errors if it is true.
17152 (arm_parse_cpu_option_name): Likewise.
17153 (arm_target_thumb_only): Adjust callers of the above.
17154 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17155 prototype to take a default true bool parameter.
17156 (arm_parse_arch_option_name): Likewise.
17158 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17159 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17163 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17164 * configure: Regenerate.
17165 * configure.ac ("linker --version-script option"): New.
17166 ("linker soname option"): New.
17168 2018-03-09 Richard Biener <rguenther@suse.de>
17170 PR tree-optimization/84775
17171 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17172 immediate uses of predicate stmts and mark them modified.
17175 PR tree-optimization/84178
17176 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17178 (version_loop_for_if_conversion): Delay update_ssa call.
17179 (tree_if_conversion): Delay update_ssa until after predicate
17182 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17185 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17186 when the function accesses prior frames.
17188 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17191 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17192 gen_llsym, otherwise call maybe_gen_llsym.
17194 PR inline-asm/84742
17195 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17196 has ',' character inside of it.
17198 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17201 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17202 as clobbering CC_REGNUM.
17204 2018-03-08 Richard Biener <rguenther@suse.de>
17206 PR middle-end/84552
17207 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17208 (follow_copies_to_constant): Do not follow SSA names registered
17211 2018-03-08 Richard Biener <rguenther@suse.de>
17213 PR tree-optimization/84178
17214 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17216 (version_loop_for_if_conversion): Delay update_ssa call.
17217 (tree_if_conversion): Delay update_ssa until after predicate
17220 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17222 PR tree-optimization/84178
17223 * tree-if-conv.c (release_bb_predicate): Remove the
17224 the assertion that the stmts have NULL use_ops.
17225 Discard the statements, asserting that they haven't
17226 yet been added to a BB.
17228 2018-03-08 Richard Biener <rguenther@suse.de>
17230 PR tree-optimization/84746
17231 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17232 (phi_translate): Pass in destination ANTIC_OUT set.
17233 (phi_translate_1): Likewise. For a simplified result lookup
17234 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17235 (phi_translate_set): Adjust.
17236 (do_pre_regular_insertion): Likewise.
17237 (do_pre_partial_partial_insertion): Likewise.
17239 2018-03-08 Martin Liska <mliska@suse.cz>
17241 PR gcov-profile/84735
17242 * doc/gcov.texi: Document usage of profile files.
17243 * gcov-io.h: Document changes in the format.
17245 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17249 * dwarf2out.c (struct dw_line_info_table): Update comments for
17251 (FORCE_RESET_NEXT_VIEW): New.
17252 (FORCE_RESETTING_VIEW_P): New.
17253 (RESETTING_VIEW_P): Check for -1 too.
17254 (ZERO_VIEW_P): Likewise.
17255 (new_line_info_table): Force-reset next view.
17256 (dwarf2out_begin_function): Likewise.
17257 (dwarf2out_source_line): Simplify zero_view_p initialization.
17258 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17259 view directly. Omit view when omitting .loc at line 0.
17261 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17263 PR tree-optimization/84740
17264 * tree-switch-conversion.c (process_switch): Call build_constructors
17265 only if info.phi_count is non-zero.
17267 PR tree-optimization/84739
17268 * tree-tailcall.c (find_tail_calls): Check call arguments against
17269 DECL_ARGUMENTS (current_function_decl) rather than
17270 DECL_ARGUMENTS (func) when checking for tail recursion.
17272 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17274 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17275 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17276 Volker Reichelt's entry and add entries for people that perform
17277 GCC fuzzy testing and report numerous bugs.
17279 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17282 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17283 readonly data in sdata, if that is disabled.
17284 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17285 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17286 -mreadonly-in-sdata option.
17288 2018-03-07 Martin Sebor <msebor@redhat.com>
17290 PR tree-optimization/84468
17291 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17292 basic block when looking for nul assignment.
17294 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17297 * except.h (output_function_exception_table): Adjust prototype.
17298 * except.c (output_function_exception_table): Remove FNNAME parameter
17299 and add SECTION parameter. Ouput one part of the table at a time.
17300 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17301 the first part of the exception table and emit unwind directives.
17302 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17303 (i386_pe_seh_cold_init): Likewise.
17304 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17305 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17306 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17307 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17308 * config/i386/winnt.c: Include except.h.
17309 (struct seh_frame_state): Add reg_offset, after_prologue and
17310 in_cold_section fields.
17311 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17312 (i386_pe_seh_cold_init): New function.
17313 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17314 to seh->in_cold_section.
17315 (seh_emit_push): Record the offset of the push.
17316 (seh_emit_save): Record the offet of the save.
17317 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17318 Test seh->after_prologue to disregard the epilogue.
17319 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17320 (i386_pe_end_cold_function): New function.
17322 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17325 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17326 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17329 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17331 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17332 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17334 PR middle-end/84723
17335 * multiple_target.c: Include tree-inline.h and intl.h.
17336 (expand_target_clones): Diagnose and fail if node->definition and
17337 !tree_versionable_function_p (node->decl).
17339 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17341 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17343 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17344 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17345 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17347 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17350 * combine.c (try_combine): Use reg_or_subregno instead of handling
17351 just paradoxical SUBREGs and REGs.
17353 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17355 * config/arc/arc.c (arc_finalize_pic): Remove function.
17356 (arc_must_save_register): We use single base PIC register, remove
17357 checks to save/restore the PIC register.
17358 (arc_expand_prologue): Likewise.
17359 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17361 (arc_verify_short): Likewise.
17362 (arc_attr_type): Likewise.
17363 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17364 (walk_stores): Likewise.
17365 (arc_address_cost): Make it static.
17366 (arc_verify_short): Likewise.
17367 (branch_dest): Likewise.
17368 (arc_attr_type): Likewise.
17369 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17370 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17371 (arc_final_prescan_insn): Remove inserting the nops due to
17372 hardware hazards. It is done in reorg step.
17373 (insn_length_variant_t): Remove.
17374 (insn_length_parameters_t): Likewise.
17375 (arc_insn_length_parameters): Likewise.
17376 (arc_get_insn_variants): Likewise.
17377 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17379 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17381 PR inline-asm/84683
17382 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17385 PR tree-optimization/84687
17386 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17388 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17390 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17392 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17393 Rename to ppc_speculation_barrier.
17394 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17395 __builtin_ppc_speculation_barrier.
17397 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17400 * combine.c (combine_simplify_rtx): Don't try to simplify if
17401 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17404 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17406 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17407 to 32 bytes when compiling for POWER9.
17409 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17412 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17413 regparm >= 3 with no arg reg available also for calls with
17414 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17417 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17419 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17421 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
17424 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17426 2018-03-05 Richard Biener <rguenther@suse.de>
17428 PR tree-optimization/84486
17429 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17430 When inserting a __builtin_assume_aligned call set the LHS
17431 SSA name alignment info accordingly.
17433 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
17435 PR tree-optimization/84114
17436 * config/aarch64/aarch64.c (aarch64_reassociation_width)
17437 Avoid reassociation of FLOAT_MODE addition.
17439 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
17441 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
17442 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
17443 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
17444 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
17445 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
17446 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
17447 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
17449 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
17450 __builtin_ia32_wbinvd): New builtins.
17451 (SPECIAL_ARGS2): New.
17452 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17453 (SPECIAL_ARGS2): New.
17454 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
17455 (ix86_valid_target_attribute_inner_p): Ditto.
17456 (ix86_init_mmx_sse_builtins): Add special_args2.
17457 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
17458 TARGET_WBNOINVD_P): New.
17459 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
17460 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
17461 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
17462 * config/i386/immintrin.h (_wbinvd): New intrinsic.
17463 * config/i386/pconfigintrin.h: New file.
17464 * config/i386/wbnoinvdintrin.h: Ditto.
17465 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
17467 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
17469 2018-03-05 Richard Biener <rguenther@suse.de>
17471 PR tree-optimization/84670
17472 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17474 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17475 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17476 (compute_antic_aux): Only assert the number of values in ANTIC_IN
17477 doesn't grow if all successors (recursively) were visited at least
17480 2018-03-05 Richard Biener <rguenther@suse.de>
17482 PR tree-optimization/84650
17483 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
17484 if executed in the loop pipeline.
17486 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
17488 * doc/configfiles.texi (Configuration Files): Move info about
17489 conditionalizing $target-protos.h to...
17490 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
17491 differs from $target-protos.h.
17493 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
17494 Chung-Ju Wu <jasonwucj@gmail.com>
17496 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
17497 * config/nds32/nds32-multiple.md (setmemsi): Define.
17498 * config/nds32/nds32-memory-manipulation.c
17499 (nds32_gen_dup_4_byte_to_word_value): New.
17500 (emit_setmem_word_loop): New.
17501 (emit_setmem_byte_loop): New.
17502 (nds32_expand_setmem_loop): New.
17503 (nds32_expand_setmem_loop_v3m): New.
17504 (nds32_expand_setmem_unroll): New.
17505 (nds32_expand_setmem): New.
17507 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17508 Chung-Ju Wu <jasonwucj@gmail.com>
17510 * config/nds32/nds32-memory-manipulation.c
17511 (nds32_emit_load_store): New.
17512 (nds32_emit_post_inc_load_store): New.
17513 (nds32_emit_mem_move): New.
17514 (nds32_emit_mem_move_block): New.
17515 (nds32_expand_movmemsi_loop_unknown_size): New.
17516 (nds32_expand_movmemsi_loop_known_size): New.
17517 (nds32_expand_movmemsi_loop): New.
17518 (nds32_expand_movmemsi_unroll): New.
17519 (nds32_expand_movmemqi): Rename ...
17520 (nds32_expand_movmemsi): ... to this.
17521 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
17522 (movmemsi): ... to this.
17523 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
17524 (nds32_expand_movmemsi): ... to this.
17526 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17527 Monk Chiang <sh.chiang04@gmail.com>
17528 Chung-Ju Wu <jasonwucj@gmail.com>
17530 * config/nds32/nds32-protos.h
17531 (nds32_expand_load_multiple): New arguments.
17532 (nds32_expand_store_multiple): Ditto.
17533 (nds32_valid_multiple_load_store): Rename ...
17534 (nds32_valid_multiple_load_store_p): ... to this.
17535 * config/nds32/nds32-memory-manipulation.c
17536 (nds32_expand_load_multiple): Refine implementation.
17537 (nds32_expand_store_multiple): Ditto.
17538 * config/nds32/nds32-multiple.md
17539 (load_multiple): Update nds32_expand_load_multiple interface.
17540 (store_multiple): Update nds32_expand_store_multiple interface.
17541 * config/nds32/nds32-predicates.c
17542 (nds32_valid_multiple_load_store): Rename ...
17543 (nds32_valid_multiple_load_store_p): ... to this and refine
17545 * config/nds32/predicates.md
17546 (nds32_load_multiple_and_update_address_operation): New predicate.
17547 (nds32_store_multiple_and_update_address_operation): New predicate.
17549 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17550 Chung-Ju Wu <jasonwucj@gmail.com>
17552 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
17553 (combo): New attribute.
17554 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
17556 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
17558 * config/nds32/nds32.opt: Change -mcmodel= default value.
17560 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
17561 Monk Chiang <sh.chiang04@gmail.com>
17562 Chung-Ju Wu <jasonwucj@gmail.com>
17564 * config/nds32/constants.md (unspec_element): New enum.
17565 * config/nds32/constraints.md (Umw): New constraint.
17566 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
17567 * config/nds32/nds32-intrinsic.md: Likewise.
17568 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
17569 (nds32_valid_smw_lwm_base_p): New.
17570 (nds32_output_smw_single_word): New.
17571 (nds32_output_lmw_single_word): New.
17572 (nds32_expand_unaligned_load): New.
17573 (nds32_expand_unaligned_store): New.
17574 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
17575 (nds32_output_smw_single_word): Declare.
17576 (nds32_output_lmw_single_word): Declare.
17577 (nds32_expand_unaligned_load): Declare.
17578 (nds32_expand_unaligned_store): Declare.
17579 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
17580 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
17581 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
17582 NDS32_BUILTIN_UASTORE_DW.
17583 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
17586 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
17587 Kito Cheng <kito.cheng@gmail.com>
17588 Chung-Ju Wu <jasonwucj@gmail.com>
17590 * config/nds32/nds32-intrinsic.c
17591 (nds32_expand_builtin_null_ftype_reg): Delete.
17592 (nds32_expand_builtin_reg_ftype_imm): Ditto.
17593 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
17594 (nds32_read_argument): New.
17595 (nds32_legitimize_target): Ditto.
17596 (nds32_legitimize_argument): Ditto.
17597 (nds32_check_constant_argument): Ditto.
17598 (nds32_expand_unop_builtin): Ditto.
17599 (nds32_expand_unopimm_builtin): Ditto.
17600 (nds32_expand_binop_builtin): Ditto.
17601 (nds32_builtin_decl_impl): Ditto.
17602 (builtin_description): Ditto.
17603 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17604 (nds32_init_builtins_impl): Ditto.
17605 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17606 (nds32_builtin_decl): New.
17607 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17608 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17610 2018-03-02 Jeff Law <law@redhat.com>
17612 * reorg.c (stop_search_p): Handle DEBUG_INSN.
17613 (redundant_insn, fill_simple_delay_slots): Likewise.
17614 (fill_slots_from_thread): Likewise.
17615 * resource.c (mark_referenced_resources): Likewise.
17616 (mark_set_resources, find_dead_or_set_registers): Likewise.
17618 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17620 * substring-locations.h (format_warning_va): Formatting fix for
17621 ATTRIBUTE_GCC_DIAG.
17622 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17624 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17625 * substring-locations.c: Include intl.h.
17626 (format_warning_va): Turned into small wrapper around
17627 format_warning_n_va, renamed to ...
17628 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17629 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17631 (format_warning_at_substring_n): New function.
17632 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17633 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
17634 format_warning_at_substring with just a shorter name instead of
17635 const function pointer.
17636 (fmtwarn_n): New function.
17637 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17638 appropriate, get rid of all the fmtstr temporaries, move conditionals
17639 with G_() wrapped string literals directly into fmtwarn arguments,
17640 cast dir.len to (int), formatting fixes.
17642 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
17644 * doc/invoke.texi: Remove "Cilk Plus" references.
17646 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17647 Richard Biener <rguenther@suse.de>
17650 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17651 for error or warning attributes if CALL_FROM_THUNK_P is set.
17654 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17657 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17658 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17661 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17662 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17664 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17665 instead of -1U in last predictors element's probability member.
17667 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17670 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17671 arguments if they are comparable.
17673 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17675 PR tree-optimization/84634
17676 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17677 masks and masked_loop_p with a single loop_masks, making sure it's
17678 null for bb vectorization.
17680 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17682 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17683 (vect_analyze_data_ref_access): Use loop->safe_len rather than
17684 loop->force_vectorize to check whether there is no alias.
17686 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17689 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17691 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17693 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17694 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17695 instead of a loop around prev_real_insn.
17696 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17699 PR inline-asm/84625
17700 * config/i386/i386.c (ix86_print_operand): Use conditional
17701 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17704 2018-03-02 Richard Biener <rguenther@suse.de>
17706 PR tree-optimization/84427
17707 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17708 (bitmap_set_subtract_values): Rewrite to handle multiple
17711 (prune_clobbered_mems): Likewise.
17712 (phi_translate): Take edge instead of pred/phiblock.
17713 (phi_translate_1): Likewise.
17714 (phi_translate_set): Likewise. Insert all translated
17715 exprs for a value into the set, keeping possibly multiple
17716 expressions per value.
17717 (compute_antic_aux): Adjust for phi_translate changes.
17718 When intersecting union the expressions and prune those
17719 not in the final value set, keeping possibly multiple
17720 expressions per value. Do not use value-insertion
17721 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17722 all expressions. Add verification that the value-sets
17723 only shrink during iteration.
17724 (compute_partial_antic_aux): Adjust for the phi_translate changes.
17725 (do_pre_regular_insertion): Likewise.
17726 (do_pre_partial_partial_insertion): Likewise.
17728 2018-03-02 Richard Biener <rguenther@suse.de>
17731 * config/darwin.c (saved_debug_info_level): New static global.
17732 (darwin_asm_lto_start): Disable debug info generation for LTO out.
17733 (darwin_asm_lto_end): Restore debug info generation settings.
17735 2018-03-01 Martin Liska <mliska@suse.cz>
17738 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
17739 volatile arguments.
17741 2018-03-01 Richard Biener <rguenther@suse.de>
17744 * dwarf2out.c (gen_variable_die): Properly handle late VLA
17745 type annotation with LTO when debug was disabled at compile-time.
17747 2018-03-01 Matthew Fortune <mfortune@gmail.com>
17749 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
17751 (mips_final_postscan_insn): Likewise.
17753 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
17755 PR rtl-optimization/84528
17756 * alias.c (init_alias_target): Add commentary.
17757 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
17758 a unique base value if the frame pointer is not eliminated
17759 to the stack pointer.
17761 2018-03-01 Tom de Vries <tom@codesourcery.com>
17763 PR rtl-optimization/83327
17764 * lra-int.h (hard_regs_spilled_into): Declare.
17765 * lra.c (hard_regs_spilled_into): Define.
17766 (init_reg_info): Init hard_regs_spilled_into.
17767 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
17768 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
17769 (process_bb_lives): Handle hard_regs_spilled_into.
17770 (lra_create_live_ranges_1): Before doing liveness propagation, clear
17771 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
17773 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
17775 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
17776 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
17777 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
17778 * config/rs6000/aix72.h: New file.
17780 2018-02-28 Jakub Jelinek <jakub@redhat.com>
17782 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
17783 instead of warning_at with conditional singular and plural messages
17787 * stor-layout.c (update_alignment_for_field): For
17788 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
17789 && !DECL_PACKED (field), do the alignment update, just use
17790 only desired_align instead of MAX (type_align, desired_align)
17792 (place_field): Don't do known_align < desired_align handling
17793 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
17794 is non-NULL, instead do it after rli->prev_field handling and
17795 only if not within a bitfield word. For DECL_PACKED (field)
17796 use type_align of BITS_PER_UNIT.
17798 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
17800 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
17801 superfluous parentheses and trailing spaces.
17803 2018-02-28 Richard Biener <rguenther@suse.de>
17805 PR tree-optimization/84584
17806 * graphite-scop-detection.c (scop_detection::add_scop): Discard
17807 SCoPs with fake exit edge.
17809 2018-02-28 Martin Liska <mliska@suse.cz>
17812 * timevar.c (timer::print): Fix format to properly print 100%
17815 2018-02-28 Richard Biener <rguenther@suse.de>
17817 PR middle-end/84607
17818 * genmatch.c (capture_info::walk_match): Do not mark
17819 captured expressions without operands as expr_p given
17820 they act more like predicates and should be subject to
17821 "lost tail" side-effect preserving.
17823 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
17825 PR rtl-optimization/81611
17826 * auto-inc-dec.c (attempt_change): Move dead note from
17827 mem_insn if it's the next use of regno
17828 (find_address): Take address use of reg holding
17829 non-incremented value. Add parm to limit search to the named
17831 (merge_in_block): Attempt to use a mem insn that is the next
17832 use of the original regno.
17834 2018-02-27 Martin Sebor <msebor@redhat.com>
17837 * doc/invoke.texi (-Wmissing-attributes): New option.
17838 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
17840 2018-02-27 Martin Sebor <msebor@redhat.com>
17842 PR translation/84207
17843 * diagnostic-core.h (warning_n, error_n, inform_n): Change
17844 n argument to unsigned HOST_WIDE_INT.
17845 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
17846 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
17847 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
17848 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
17850 2018-02-27 Richard Biener <rguenther@suse.de>
17852 PR tree-optimization/84512
17853 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
17854 Do not use the estimate returned from record_stmt_cost for
17855 the scalar iteration cost but sum properly using add_stmt_cost.
17857 2018-02-27 Richard Biener <rguenther@suse.de>
17859 PR tree-optimization/84466
17860 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17861 Adjust last change to less strictly validate use operands.
17863 2018-02-27 Martin Liska <mliska@suse.cz>
17865 PR gcov-profile/84548
17866 * gcov.c (process_file): Allow partial overlap and consider it
17867 also as group functions.
17868 (output_lines): Properly calculate range of lines for a group.
17870 2018-02-27 Martin Liska <mliska@suse.cz>
17872 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
17873 'ggc' suffixes. Change first column width.
17874 (timer::print): Fix formatting of the column.
17876 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
17878 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
17879 preserve inline entry blocks for the sake of debug inline
17880 entry point markers alone.
17881 (remove_unused_locals): Suggest in comments a better place to
17882 force the preservation of inline entry blocks that are
17883 otherwise unused, but do not preserve them.
17885 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17887 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
17889 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17892 * config/i386/constraints.md (Bs): Replace
17893 ix86_indirect_branch_register with
17894 TARGET_INDIRECT_BRANCH_REGISTER.
17896 * config/i386/i386.md (indirect_jump): Likewise.
17897 (tablejump): Likewise.
17898 (*sibcall_memory): Likewise.
17899 (*sibcall_value_memory): Likewise.
17900 Peepholes of indirect call and jump via memory: Likewise.
17901 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
17902 (*sibcall_value_GOT_32): Likewise.
17903 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17904 (GOT_memory_operand): Likewise.
17905 (call_insn_operand): Likewise.
17906 (sibcall_insn_operand): Likewise.
17907 (GOT32_symbol_operand): Likewise.
17908 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
17910 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17912 PR rtl-optimization/83496
17913 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
17914 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
17915 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
17916 redundant insn, if any.
17917 (relax_delay_slots): Likewise.
17918 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
17920 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
17922 PR tree-optimization/83965
17923 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
17924 that grouped statements are part of a reduction chain. Return
17925 true if the statement is not marked as a reduction itself but
17926 is part of a group.
17927 (vect_recog_dot_prod_pattern): Don't check whether the statement
17928 is part of a group here.
17929 (vect_recog_sad_pattern): Likewise.
17930 (vect_recog_widen_sum_pattern): Likewise.
17932 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17935 * final.c (rest_of_clean_state): Also look for calls inside sequences.
17937 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17940 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
17942 (ix86_output_indirect_function_return): New prototype.
17943 (ix86_split_simple_return_pop_internal): Likewise.
17944 * config/i386/i386.c (indirect_return_via_cx): New.
17945 (indirect_return_via_cx_bnd): Likewise.
17946 (indirect_thunk_name): Handle return va CX_REG.
17947 (output_indirect_thunk_function): Create alias for
17948 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
17949 (ix86_output_indirect_jmp): Remove the bool argument.
17950 (ix86_output_indirect_function_return): New function.
17951 (ix86_split_simple_return_pop_internal): Likewise.
17952 * config/i386/i386.md (*indirect_jump): Don't pass false
17953 to ix86_output_indirect_jmp.
17954 (*tablejump_1): Likewise.
17955 (simple_return_pop_internal): Change it to define_insn_and_split.
17956 Call ix86_split_simple_return_pop_internal to split it for
17957 -mfunction-return=.
17958 (simple_return_indirect_internal): Call
17959 ix86_output_indirect_function_return instead of
17960 ix86_output_indirect_jmp.
17962 2018-02-26 Jakub Jelinek <jakub@redhat.com>
17965 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
17966 memset and value initialization afterwards.
17968 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
17970 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
17972 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17975 * common/config/aarch64/aarch64-common.c
17976 (aarch_option_optimization_table[]): Switch
17977 off fomit-frame-pointer
17979 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17980 Chung-Ju Wu <jasonwucj@gmail.com>
17982 * config/nds32/nds32-multiple.md (load_multiple): Disallow
17984 (store_multiple): Ditto.
17986 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17988 * config.gcc: Add --with-cpu support for nds32 target.
17989 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
17990 * config/nds32/nds32.opt: Add -mcpu= option.
17992 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
17994 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
17995 isel=yes): Warn for these deprecated options.
17997 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
17999 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18000 ISA_2_5_MASKS_EMBEDDED.
18002 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18004 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18005 p->max as pointers rather than using iterative_hash_expr.
18007 2018-02-23 Carl Love <cel@us.ibm.com>
18009 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18010 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18012 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18013 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18014 P8V_BUILTIN_VEC_VUNSIGNED2.
18016 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18019 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18020 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18022 * lra-constraints.c (curr_insn_transform): Set up
18023 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18025 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18026 * lra-eliminations.c (spill_pseudos): Ditto.
18027 (process_insn_for_elimination): Ditto.
18028 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18029 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18030 LRA_NON_CLOBBERED_ALT.
18032 2018-02-22 Martin Sebor <msebor@redhat.com>
18034 PR tree-optimization/84480
18035 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18036 to maybe_diag_stxncpy_trunc. Call it.
18037 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18038 from gimple_fold_builtin_strcpy. Print inlining stack.
18039 (handle_builtin_stxncpy): Print inlining stack.
18040 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18042 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18045 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18046 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18047 and -fcheck-pointer-bounds are used together.
18048 (indirect_thunk_prefix): New enum.
18049 (indirect_thunk_need_prefix): New function.
18050 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18051 "_nt" instead of "_bnd" for NOTRACK prefix.
18052 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18053 (output_indirect_thunk_function): Likewise.
18055 (ix86_code_end): Update output_indirect_thunk_function calls.
18056 (ix86_output_indirect_branch_via_reg): Replace
18057 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18058 (ix86_output_indirect_branch_via_push): Likewise.
18059 (ix86_output_function_return): Likewise.
18060 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18061 incompatible with -fcf-protection=branch and
18062 -fcheck-pointer-bounds.
18064 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18067 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18068 Change gcc_assert call to output_operand_lossage.
18070 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18072 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18074 2018-02-22 DJ Delorie <dj@redhat.com>
18075 Sebastian Perta <sebastian.perta@renesas.com>
18076 Oleg Endo <olegendo@gcc.gnu.org>
18078 * config/rx/rx.c (rx_rtx_costs): New function.
18079 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18081 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18083 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18085 2018-02-22 Martin Liska <mliska@suse.cz>
18088 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18089 Add "native" as a possible value.
18091 2018-02-22 Martin Liska <mliska@suse.cz>
18094 * config/i386/i386.c (ix86_option_override_internal):
18095 Add "native" as a possible value for -march and -mtune.
18097 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18100 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18101 to all type variants.
18103 PR tree-optimization/84503
18104 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18105 width as info->bitpos + info->bitsize - start.
18106 (merged_store_group::merge_overlapping): Simplify width computation.
18107 (check_no_overlap): New function.
18108 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18109 start + width and last_order of the group, fail if check_no_overlap
18111 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18112 to group if check_no_overlap fails.
18114 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18116 * config/rs6000/altivec.md: Delete contraint arguments to
18117 define_expand, define_split, and define_peephole2, and in
18118 define_insn_and_split if always unused.
18119 * config/rs6000/darwin.md: Ditto.
18120 * config/rs6000/dfp.md: Ditto.
18121 * config/rs6000/rs6000.md: Ditto.
18122 * config/rs6000/sync.md: Ditto.
18123 * config/rs6000/vector.md: Ditto.
18124 * config/rs6000/vsx.md: Ditto.
18126 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18128 * config/rs6000/altivec.md: Write output control strings as braced
18129 blocks instead of double-quoted strings.
18130 * config/rs6000/darwin.md: Ditto.
18131 * config/rs6000/rs6000.md: Ditto.
18132 * config/rs6000/vector.md: Ditto.
18133 * config/rs6000/vsx.md: Ditto.
18135 2018-02-21 Jason Merrill <jason@redhat.com>
18137 PR c++/84314 - ICE with templates and fastcall attribute.
18138 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18140 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18142 * ipa-cp.c (determine_versionability): Fix comment typos.
18144 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18147 * ipa-cp.c (determine_versionability): Do not version functions caling
18150 2018-02-21 Martin Liska <mliska@suse.cz>
18153 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18154 Add "native" as a possible value.
18155 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18156 the macro when native cpu detection is available.
18158 2018-02-21 Martin Liska <mliska@suse.cz>
18161 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18162 Add "native" as a possible value.
18163 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18164 when native cpu detection is available.
18166 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18167 Martin Sebor <msebor@redhat.com>
18169 PR tree-optimization/84478
18170 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18172 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18173 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18174 support which is conservatively correct, for 2 only stay conservative
18175 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18176 argument to the 2 argument get_range_strlen, adjust 6 arg
18177 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18179 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18180 (gimple_fold_builtin_strlen): Pass true as last argument to
18183 2018-02-20 Martin Sebor <msebor@redhat.com>
18185 PR middle-end/84095
18186 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18187 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18188 (builtin_memref::builtin_memref): Factor out parts into
18189 set_base_and_offset and call it.
18191 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18193 PR middle-end/84406
18194 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18195 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18196 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18197 search at the associated MODE_INT.
18199 2018-02-20 Jeff Law <law@redhat.com>
18201 PR middle-end/82123
18202 PR tree-optimization/81592
18203 PR middle-end/79257
18204 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18205 for range data rather than using global data.
18206 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18207 range data rather than using global data.
18208 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18209 pass it to children as needed.
18210 (struct directive::fmtresult): Similarly.
18211 (struct directive::set_width): Similarly.
18212 (struct directive::set_precision): Similarly.
18213 (format_integer, format_directive, parse_directive): Similarly.
18214 (format_none): Accept unnamed vr_values parameter.
18215 (format_percent, format_floating, format_character): Similarly.
18216 (format_string, format_plain): Similarly.
18217 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18218 the EVRP range analyzer for range data rather than using global data.
18219 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18220 gimple-ssa-evrp-analyze.h
18221 (class sprintf_dom_walker): Add after_dom_children member function.
18222 Add evrp_range_analyzer member.
18223 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18224 range analyzer as needed.
18225 (sprintf_dom_walker::after_dom_children): New member function.
18226 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18228 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18229 (evrp_range_analyzer::pop_to_marker): Likewise.
18231 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18233 PR tree-optimization/84419
18234 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18235 with the required type if its current type is compatible but
18238 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18240 PR middle-end/82004
18241 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18242 after vectorization.
18244 2018-02-20 Martin Liska <mliska@suse.cz>
18247 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18248 possible values if we don't have a hint.
18250 2018-02-20 Martin Liska <mliska@suse.cz>
18254 * final.c (shorten_branches): Build align_tab array with one
18256 * opts.c (finish_options): Add alignment option limit check.
18257 (MAX_CODE_ALIGN): Likewise.
18258 (MAX_CODE_ALIGN_VALUE): Likewise.
18259 * doc/invoke.texi: Document maximum allowed option value for
18260 all -falign-* options.
18262 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18265 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18266 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18267 * var-tracking.c (emit_note_insn_var_location): Remove all references
18268 to NOTE_INSN_CALL_ARG_LOCATION.
18269 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18270 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18271 Use copy_rtx_if_shared.
18272 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18273 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18274 (dwarf2out_var_location): Remove handling of
18275 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18277 * final.c (final_scan_insn): Remove all references to
18278 NOTE_INSN_CALL_ARG_LOCATION.
18279 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18280 before dumping final insns.
18281 * except.c (emit_note_eh_region_end): Remove all references to
18282 NOTE_INSN_CALL_ARG_LOCATION.
18283 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18284 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18285 * config/arc/arc.c (hwloop_optimize): Likewise.
18286 * config/arm/arm.c (create_fix_barrier): Likewise.
18287 * config/s390/s390.c (s390_chunkify_start): Likewise.
18288 * config/sh/sh.c (find_barrier): Likewise.
18289 * config/i386/i386.c (rest_of_insert_endbranch,
18290 ix86_seh_fixup_eh_fallthru): Likewise.
18291 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18292 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18293 * config/frv/frv.c (frv_function_prologue): Likewise.
18294 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18296 (note_outside_basic_block_p): Remove all references to
18297 NOTE_INSN_CALL_ARG_LOCATION.
18298 * gengtype.c (adjust_field_rtx_def): Likewise.
18299 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18301 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18302 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18305 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18308 PR tree-optimization/84452
18309 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18310 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18313 2018-02-19 Martin Liska <mliska@suse.cz>
18316 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18318 2018-02-19 Martin Liska <mliska@suse.cz>
18319 Richard Sandiford <richard.sandiford@linaro.org>
18321 PR tree-optimization/82491
18322 * gimple-fold.c (get_base_constructor): Make earlier bail out
18325 2018-02-19 Carl Love <cel@us.ibm.com>
18327 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18328 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18330 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18331 P8V_BUILTIN_VEC_NEG.
18333 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18335 * config/rl78/rl78.md (movdf): New define expand.
18337 2018-02-19 Martin Liska <mliska@suse.cz>
18340 * doc/invoke.texi: Fix typo.
18341 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18343 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18345 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18346 handle rs6000_single_float and rs6000_double_float specially for
18349 2018-02-16 Jeff Law <law@redhat.com>
18351 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18352 (pop_regs): Use it.
18354 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18357 * ipa-inline.c (inline_small_functions): Fix a typo.
18359 2018-02-16 Nathan Sidwell <nathan@acm.org>
18361 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18363 2018-02-16 Carl Love <cel@us.ibm.com>
18365 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18366 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18367 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18368 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18369 expansion to P8V_BUILTIN_VEC_FLOAT2.
18371 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18373 PR rtl-optimization/70023
18374 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18375 src_regno into account.
18377 2018-02-16 Carl Love <cel@us.ibm.com>
18379 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18380 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18381 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18382 * config/rs6000/rs6000.c: Remove case statements for
18383 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18384 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18385 and P9V_BUILTIN_VEC_VINSERT4B.
18386 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18387 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18388 * config/rs6000/vsx.md:
18389 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18392 2018-02-16 Carl Love <cel@us.ibm.com>
18394 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18396 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18398 * config/rs6000/rs6000-c.c: Add the definitions for
18399 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18400 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18401 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18402 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18403 definition for insert4b and define insn *insert3b_internal.
18404 * doc/extend.texi: Add documentation for vec_extract4b.
18406 2018-02-16 Nathan Sidwell <nathan@acm.org>
18408 * doc/extend.texi (Backwards Compatibility): Mention friend
18409 injection. Note for-scope is deprecated.
18410 * doc/invoke.texi (-ffriend-injection): Deprecate.
18412 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18414 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18415 that moved to I2, also allow destinations that are a paradoxical
18416 subreg (instead of a normal reg).
18418 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
18421 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18424 2018-02-16 Richard Biener <rguenther@suse.de>
18426 PR tree-optimization/84037
18427 PR tree-optimization/84016
18429 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18430 Adjust vec_construct for the fact we need additional higher latency
18431 128bit inserts for AVX256 and AVX512 vector builds.
18432 (ix86_add_stmt_cost): Scale vector construction cost for
18435 2018-02-16 Richard Biener <rguenther@suse.de>
18437 PR tree-optimization/84417
18438 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
18439 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
18440 (non_rewritable_lvalue_p): Likewise, use poly-ints.
18442 2018-02-16 Martin Liska <mliska@suse.cz>
18445 * internal-fn.def (ASAN_CHECK): Set proper flags.
18446 (ASAN_MARK): Likewise.
18448 2018-02-16 Julia Koval <julia.koval@intel.com>
18450 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
18451 from PTA_CANNONLAKE.
18453 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18456 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
18457 Use ++iter rather than iter++ for std::list iterators.
18458 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
18459 defer deleting them until all nodes in the forest are processed. Do
18460 free even leaf nodes. Change to_process into auto_vec.
18463 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
18464 * vec.h (vec_default_construct): Use memset instead of placement new
18465 if BROKEN_VALUE_INITIALIZATION is defined.
18466 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
18467 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
18470 PR rtl-optimization/83723
18471 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
18472 * lra.c (lra_substitute_pseudo): Likewise. If true, use
18473 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
18475 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
18477 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
18479 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
18481 PR rtl-optimization/81443
18482 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
18483 from inner REGs to paradoxical SUBREGs.
18485 2018-02-16 Richard Biener <rguenther@suse.de>
18487 PR tree-optimization/84399
18488 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18489 For operands we can analyze at their definition make sure we can
18490 analyze them at each use as well.
18492 2018-02-16 Richard Biener <rguenther@suse.de>
18494 PR tree-optimization/84190
18495 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
18496 volatile accesses if the decl isn't volatile.
18498 2018-02-15 Jason Merrill <jason@redhat.com>
18500 PR c++/84314 - ICE with templates and fastcall attribute.
18501 * attribs.c (build_type_attribute_qual_variant): Don't clobber
18502 TYPE_CANONICAL on an existing type.
18504 2018-02-15 Jakub Jelinek <jakub@redhat.com>
18506 PR tree-optimization/84383
18507 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
18508 dstoff nor call operand_equal_p if dstbase is NULL.
18510 PR tree-optimization/84334
18511 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
18512 also a CONSTANT_CLASS_P, punt.
18514 2018-02-14 Jim Wilson <jimw@sifive.com>
18516 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
18517 first SMALL_OPERAND check. New local min_second_step. Move assert
18518 to where locals are set. Add TARGET_RVC support.
18519 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
18521 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
18523 * doc/invoke.texi: Correct -Wformat-overflow code sample.
18525 2018-02-14 Martin Sebor <msebor@redhat.com>
18527 PR tree-optimization/83698
18528 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
18529 arrays constrain the offset range to their bounds.
18530 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
18531 (builtin_access::overlap): Avoid setting the size of overlap if it's
18533 (maybe_diag_overlap): Also consider arrays when deciding what values
18534 of offsets to include in diagnostics.
18536 2018-02-14 Martin Sebor <msebor@redhat.com>
18539 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
18540 that correspond to the kind of a declaration.
18542 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
18545 * config/pa/pa.md: Load address of PIC label using the linkage table
18546 if the label is nonlocal.
18548 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
18550 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
18551 warning message if user requests -maltivec=be.
18552 * doc/invoke.texi: Document deprecation of -maltivec=be.
18554 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
18557 * config/rs6000/rs6000-c.c: Update definitions for
18558 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
18559 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
18561 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18564 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
18565 add _get_ssp intrinsics. Remove argument from
18566 __builtin_ia32_rdssp[d|q].
18567 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
18568 * config/i386/i386-builtin.def: Remove argument from
18569 __builtin_ia32_rdssp[d|q].
18570 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
18571 ix86_expand_special_args_builtin for _rdssp[d|q].
18572 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
18573 Clear register before usage.
18574 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
18575 Add documentation for new _get_ssp and _inc_ssp intrinsics.
18577 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
18579 PR tree-optimization/84357
18580 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
18581 operand 1 of an ARRAY_REF too.
18583 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
18586 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
18587 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18589 (set_of_reg): New struct.
18590 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
18591 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
18592 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18594 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
18595 Split into bitclr, bitset, bitinvert patterns if appropriate.
18596 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
18597 use rx_fuse_in_memory_bitop.
18598 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
18599 to named insn, correct maximum insn length.
18601 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
18604 * machmode.def: Define a complex mode for PARTIAL_INT.
18605 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18607 * doc/rtl.texi: Document CSPImode.
18608 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18610 (msp430_hard_regno_nregs_with_padding): Likewise.
18612 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
18615 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18617 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
18619 PR rtl-optimization/84169
18620 * combine.c (try_combine): New variable split_i2i3. Set it to true if
18621 we generated a parallel as new i3 and we split that to new i2 and i3
18622 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
18623 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
18624 those to i2, not i1. Partially rewrite this scan code.
18626 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18629 * stor-layout.c (place_field): For variable length fields, adjust
18630 offset_align afterwards not just based on the field's alignment,
18631 but also on the size.
18633 PR middle-end/84309
18634 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18635 of exps and logs in the use_exp2 case.
18637 2018-02-13 Jeff Law <law@redhat.com>
18639 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18640 entry for "vector".
18642 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18645 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
18649 * common.opt (gas-loc-support, gas-locview-support): New.
18650 (ginline-points, ginternal-reset-location-views): New.
18651 * doc/invoke.texi: Document them. Use @itemx where intended.
18652 (gvariable-location-views): Adjust.
18653 * target.def (reset_location_view): New.
18654 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18655 (TARGET_RESET_LOCATION_VIEW): New.
18656 * doc/tm.texi: Rebuilt.
18657 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18658 (dwarf2out_default_as_locview_support): New.
18659 (output_asm_line_debug_info): Use option variables.
18660 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18661 (output_loc_list): Likewise.
18662 (add_high_low_attributes): Check option variables.
18663 Don't output entry view attribute in strict mode.
18664 (gen_inlined_subroutine_die): Check option variables.
18665 (dwarf2out_inline_entry): Likewise.
18666 (init_sections_and_labels): Likewise.
18667 (dwarf2out_early_finish): Likewise.
18668 (maybe_reset_location_view): New, from...
18669 (dwarf2out_var_location): ... here. Call it.
18670 * debug.h (dwarf2out_default_as_loc_support): Declare.
18671 (dwarf2out_default_as_locview_support): Declare.
18672 * hooks.c (hook_int_rtx_insn_0): New.
18673 * hooks.h (hook_int_rtx_insn_0): Declare.
18674 * toplev.c (process_options): Take -gas-loc-support and
18675 -gas-locview-support from dwarf2out. Enable
18676 -gvariable-location-views by default only with locview
18677 assembler support. Enable -ginternal-reset-location-views by
18678 default only if the target defines the corresponding hook.
18679 Enable -ginline-points by default if location views are
18680 enabled; force it disabled if statement frontiers are
18682 * tree-inline.c (expand_call_inline): Check option variables.
18683 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18685 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
18687 PR tree-optimization/84321
18688 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18689 handling. Also check whether the anti-range contains any values
18690 that satisfy the mask; switch to a VR_RANGE if not.
18692 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
18695 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18697 2018-02-13 Martin Jambor <mjambor@suse.cz>
18700 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18701 of call statements, also set location of a load to a temporary.
18703 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
18705 * config/rl78/rl78.c (add_vector_labels): New function.
18706 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18707 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18708 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18709 which checks that no arguments are passed.
18710 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18711 * doc/extend.texi: Documentation for the new attribute.
18713 2018-02-13 Andreas Schwab <schwab@suse.de>
18715 * config/riscv/linux.h (CPP_SPEC): Define.
18717 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18720 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18721 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18722 OPTION_MASK_ISA_AES as first argument to def_builtin_const
18723 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18724 instead of OPTION_MASK_ISA_PCLMUL as first argument to
18725 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18726 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18727 temporarily for AES and PCLMUL builtins.
18729 PR tree-optimization/84339
18730 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18731 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
18734 PR middle-end/84309
18735 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
18736 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
18737 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
18739 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
18741 * omp-simd-clone.h: New file.
18742 * omp-simd-clone.c: Include omp-simd-clone.h.
18743 (expand_simd_clones): No longer static.
18744 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
18745 cgraph.h and omp-simd-clone.h.
18746 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
18747 (vect_recog_widen_shift_pattern): Formatting fix.
18748 (vect_pattern_recog_1): Don't check optab for calls.
18751 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
18752 operands[2] into a REG before using gen_lowpart on it.
18754 2018-02-12 Jeff Law <law@redhat.com>
18757 * config/sh/sh.c (find_barrier): Consider a sibling call
18760 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
18761 successfully back substituting a reg.
18763 2018-02-12 Richard Biener <rguenther@suse.de>
18765 PR tree-optimization/84037
18766 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
18767 parameter, move visited init to caller.
18768 (vect_slp_analyze_operations): Separate cost from validity
18769 check, initialize visited once for all instances.
18770 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
18772 * tree-vect-stmts.c (vect_model_simple_cost): Make early
18774 (vect_model_promotion_demotion_cost): Likewise.
18775 (vectorizable_bswap): Guard cost modeling with !slp_node
18776 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
18778 (vectorizable_call): Likewise.
18779 (vectorizable_conversion): Likewise.
18780 (vectorizable_assignment): Likewise.
18781 (vectorizable_shift): Likewise.
18782 (vectorizable_operation): Likewise.
18783 (vectorizable_store): Likewise.
18784 (vectorizable_load): Likewise.
18785 (vectorizable_condition): Likewise.
18786 (vectorizable_comparison): Likewise.
18788 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
18791 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
18792 (ASAN_MARK): Fix fnspec to account for return value, change pointer
18793 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
18795 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18797 PR middle-end/83665
18798 * params.def (inline-min-speedup): Increase from 8 to 15.
18799 (max-inline-insns-auto): Decrease from 40 to 30.
18800 * ipa-split.c (consider_split): Add some buffer for function to
18801 be considered inlining candidate.
18802 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
18805 2018-02-12 Richard Biener <rguenther@suse.de>
18807 PR tree-optimization/84037
18808 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
18809 matched stmts if we cannot swap the non-matched ones.
18811 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18813 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
18814 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
18815 _mm_maskz_scalef_round_ss): New intrinsics.
18816 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
18817 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
18818 __builtin_ia32_scalefss_round): Remove.
18819 (__builtin_ia32_scalefsd_mask_round,
18820 __builtin_ia32_scalefss_mask_round): New intrinsics.
18821 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
18822 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
18823 ((match_operand:VF_128 2 "<round_nimm_predicate>"
18824 "<round_constraint>")): Changed to ...
18825 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
18826 "<round_scalar_constraint>")): ... this.
18827 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
18828 %0, %1, %2<round_op3>}"): Changed to ...
18829 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
18830 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
18831 %2<round_scalar_mask_op3>}"): ... this.
18832 * config/i386/subst.md (round_scalar_nimm_predicate): New.
18834 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18836 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
18837 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
18838 (_mm_maskz_sqrt_round_ss): New intrinsics.
18839 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
18840 (__builtin_ia32_sqrtsd_mask_round)
18841 (__builtin_ia32_sqrtss_mask_round): New builtins.
18842 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
18843 (__builtin_ia32_sqrtss_round): Remove.
18844 (__builtin_ia32_sqrtsd_mask_round)
18845 (__builtin_ia32_sqrtss_mask_round): New builtins.
18846 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
18847 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
18848 ((match_operand:VF_128 1 "vector_operand"
18849 "xBm,<round_constraint>")): Changed to ...
18850 ((match_operand:VF_128 1 "vector_operand"
18851 "xBm,<round_scalar_constraint>")): ... this.
18852 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
18853 %0, %2, %<iptr>1<round_op3>}): Changed to ...
18854 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
18855 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
18856 %<iptr>1<round_scalar_mask_op3>}): ... this.
18857 ((set_attr "prefix" "<round_prefix>")): Changed to ...
18858 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
18860 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
18863 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
18864 Cast vec_cmpeq result to correct type.
18865 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
18866 Cast vec_cmpgt result to correct type.
18868 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
18870 * final.c (final_scan_insn_1): Renamed from...
18871 (final_scan_insn): ... this. New wrapper, to recover
18872 seen from the outermost call in recursive ones.
18873 * config/sparc/sparc.c (output_return): Drop seen from call.
18874 (output_sibcall): Likewise.
18875 * config/visium/visium.c (output_branch): Likewise.
18877 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
18879 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
18882 2018-02-10 Alan Modra <amodra@gmail.com>
18885 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
18886 Specify LR as an input.
18888 2018-02-10 Jakub Jelinek <jakub@redhat.com>
18891 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
18892 remove_member_access_dummy_vars): New functions.
18893 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
18894 lower_omp_1, execute_lower_omp): Use them.
18896 PR rtl-optimization/84308
18897 * shrink-wrap.c (spread_components): Release todo vector.
18899 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
18901 PR rtl-optimization/57193
18902 * ira-color.c (struct allocno_color_data): Add member
18903 conflict_allocno_hard_prefs.
18904 (update_conflict_allocno_hard_prefs): New.
18905 (bucket_allocno_compare_func): Add a preference based on
18906 conflict_allocno_hard_prefs.
18907 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
18908 (color_allocnos): Remove a dead code. Initiate
18909 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
18911 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18914 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
18915 constraint from =wa to wa. Avoid a subreg on the output operand,
18916 instead use a pseudo and subreg it in a move.
18917 (p9_xxbrd_<mode>): Changed to ...
18918 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
18919 (p9_xxbrd_v2df): New expander.
18920 (p9_xxbrw_<mode>): Changed to ...
18921 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
18922 (p9_xxbrw_v4sf): New expander.
18924 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18926 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
18928 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
18931 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
18932 multiply in 32-bit mode.
18933 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
18934 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
18937 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18939 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
18940 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
18941 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
18942 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
18944 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
18947 * dwarf2out.c (is_trivial_indirect_ref): New function.
18948 (dwarf2out_late_global_decl): Do not generate a location
18949 attribute for variables that have a non-trivial DECL_VALUE_EXPR
18950 and that are not defined in the current unit.
18952 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18954 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
18955 instead of a libcall for UNORDERED.
18957 2018-02-09 Tamar Christina <tamar.christina@arm.com>
18960 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
18961 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
18963 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18966 * config/s390/s390.c (s390_set_current_function): Invoke
18967 s390_indirect_branch_settings also if fndecl didn't change.
18969 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18971 * config/rs6000/rs6000.md (blockage): Set length to zero.
18973 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18975 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
18977 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18980 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
18981 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
18985 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
18986 PARALLEL incoming that failed vt_get_decl_and_offset check.
18988 PR middle-end/84237
18989 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
18990 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
18992 (get_variable_section): For decls in named .bss* sections pass true as
18993 second argument to bss_initializer_p.
18995 2018-02-09 Marek Polacek <polacek@redhat.com>
18996 Jakub Jelinek <jakub@redhat.com>
18999 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19000 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19001 Sync some changes from cxx_fold_indirect_ref.
19003 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19005 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19007 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19008 (BLOCK_INLINE_ENTRY_LABEL): New.
19009 (dwarf2out_var_location): Disregard inline entry markers.
19010 (inline_entry_data): New struct.
19011 (inline_entry_data_hasher): New hashtable type.
19012 (inline_entry_data_hasher::hash): New.
19013 (inline_entry_data_hasher::equal): New.
19014 (inline_entry_data_table): New variable.
19015 (add_high_low_attributes): Add DW_AT_entry_pc and
19016 DW_AT_GNU_entry_view attributes if a pending entry is found
19017 in inline_entry_data_table. Add old entry_pc attribute only
19018 if debug nonbinding markers are disabled.
19019 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19020 markers are enabled.
19021 (block_within_block_p, dwarf2out_inline_entry): New.
19022 (dwarf2out_finish): Check that no entries remained in
19023 inline_entry_data_table.
19024 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19025 (final_scan_insn, notice_source_line): Likewise.
19026 (rest_of_clean_state): Skip inline entry markers.
19027 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19029 * gimple.c (gimple_build_debug_inline_entry): New.
19030 * gimple.h (enum gimple_debug_subcode): Add
19031 GIMPLE_DEBUG_INLINE_ENTRY.
19032 (gimple_build_debug_inline_entry): Declare.
19033 (gimple_debug_inline_entry_p): New.
19034 (gimple_debug_nonbind_marker_p): Adjust.
19035 * insn-notes.def (INLINE_ENTRY): New.
19036 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19037 inline entry marker notes.
19038 (print_insn): Likewise.
19039 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19040 (INSN_DEBUG_MARKER_KIND): Likewise.
19041 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19042 * tree-inline.c (expand_call_inline): Build and insert
19043 debug_inline_entry stmt.
19044 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19045 inline entry blocks early, if nonbind markers are enabled.
19046 (dump_scope_block): Dump fragment info.
19047 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19048 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19049 (gimple_build_debug_inline_entry): New.
19050 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19051 Enable/disable inline entry points too.
19052 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19053 (DEBUG_INSN): Describe inline entry markers.
19055 * common.opt (gvariable-location-views): New.
19056 (gvariable-location-views=incompat5): New.
19057 * config.in: Rebuilt.
19058 * configure: Rebuilt.
19059 * configure.ac: Test assembler for view support.
19060 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19061 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19062 * dwarf2out.c (var_loc_view): New typedef.
19063 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19064 (dwarf2out_locviews_in_attribute): New.
19065 (dwarf2out_locviews_in_loclist): New.
19066 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19067 (enum dw_line_info_opcode): Add LI_adv_address.
19068 (struct dw_line_info_table): Add view.
19069 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19070 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19071 (zero_view_p): New variable.
19072 (ZERO_VIEW_P): New macro.
19073 (output_asm_line_debug_info): New.
19074 (struct var_loc_node): Add view.
19075 (add_AT_view_list, AT_loc_list): New.
19076 (add_var_loc_to_decl): Add view param. Test it against last.
19077 (new_loc_list): Add view params. Record them.
19078 (AT_loc_list_ptr): Handle loc and view lists.
19079 (view_list_to_loc_list_val_node): New.
19080 (print_dw_val): Handle dw_val_class_view_list.
19081 (size_of_die): Likewise.
19082 (value_format): Likewise.
19083 (loc_list_has_views): New.
19084 (gen_llsym): Set vl_symbol too.
19085 (maybe_gen_llsym, skip_loc_list_entry): New.
19086 (dwarf2out_maybe_output_loclist_view_pair): New.
19087 (output_loc_list): Output view list or entries too.
19088 (output_view_list_offset): New.
19089 (output_die): Handle dw_val_class_view_list.
19090 (output_dwarf_version): New.
19091 (output_compilation_unit_header): Use it.
19092 (output_skeleton_debug_sections): Likewise.
19093 (output_rnglists, output_line_info): Likewise.
19094 (output_pubnames, output_aranges): Update version comments.
19095 (output_one_line_info_table): Output view numbers in asm comments.
19096 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19097 Call maybe_gen_llsym.
19098 (loc_list_from_tree_1): Adjust.
19099 (add_AT_location_description): Create view list attribute if
19100 needed, check it's absent otherwise.
19101 (convert_cfa_to_fb_loc_list): Adjust.
19102 (maybe_emit_file): Call output_asm_line_debug_info for test.
19103 (dwarf2out_var_location): Reset views as needed. Precompute
19104 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19105 attribute. Set view.
19106 (new_line_info_table): Reset next view.
19107 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19108 (dwarf2out_source_line): Likewise. Output view resets and labels to
19109 the assembler, or select appropriate line info opcodes.
19110 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19111 (optimize_string_length): Catch it. Adjust.
19112 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19113 dw_val_class_view_list, and remove it if no longer needed.
19114 (hash_loc_list): Hash view numbers.
19115 (loc_list_hasher::equal): Compare them.
19116 (optimize_location_lists): Check whether a view list symbol is
19117 needed, and whether the locview attribute is present, and
19118 whether they match. Remove the locview attribute if no longer
19120 (index_location_lists): Call skip_loc_list_entry for test.
19121 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19122 Use output_dwarf_version.
19123 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19124 (struct dw_val_node): Add val_view_list.
19125 * final.c (SEEN_NEXT_VIEW): New.
19126 (set_next_view_needed): New.
19127 (clear_next_view_needed): New.
19128 (maybe_output_next_view): New.
19129 (final_start_function): Rename to...
19130 (final_start_function_1): ... this. Take pointer to FIRST,
19131 add SEEN parameter. Emit param bindings in the initial view.
19132 (final_start_function): Reintroduce SEEN-less interface.
19133 (final): Rename to...
19134 (final_1): ... this. Take SEEN parameter. Output final pending
19135 next view at the end.
19136 (final): Reintroduce seen-less interface.
19137 (final_scan_insn): Output pending next view before switching
19138 sections or ending a block. Mark the next view as needed when
19139 outputting variable locations. Notify debug backend of section
19140 changes, and of location view changes.
19141 (rest_of_handle_final): Adjust.
19142 * toplev.c (process_options): Autodetect value for debug variable
19143 location views option. Warn on incompat5 without -gdwarf-5.
19144 * doc/invoke.texi (gvariable-location-views): New.
19145 (gvariable-location-views=incompat5): New.
19146 (gno-variable-location-views): New.
19148 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19150 PR tree-optimization/84136
19151 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19152 that the result of find_edge is non-NULL.
19154 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19157 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19158 storing integer register in SImode. Fix cost of 256 and 512
19159 byte aligned SSE register store.
19161 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19163 * config/i386/i386.c (ix86_multiplication_cost): Fix
19164 multiplication cost for TARGET_AVX512DQ.
19166 2018-02-08 Marek Polacek <polacek@redhat.com>
19168 PR tree-optimization/84238
19169 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19172 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19174 PR tree-optimization/84265
19175 * tree-vect-stmts.c (vectorizable_store): Don't treat
19176 VMAT_CONTIGUOUS accesses as grouped.
19177 (vectorizable_load): Likewise.
19179 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19181 PR tree-optimization/81635
19182 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19183 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19184 (test_round_for_mask): New functions.
19185 (wide_int_cc_tests): Call test_round_for_mask.
19186 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19187 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19188 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19189 range returned by get_range_info.
19191 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19194 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19195 * symtab.c: Include builtins.h
19196 (symtab_node::output_to_lto_symbol_table_p): Move here
19197 from lto-streamer-out.c:output_symbol_p.
19198 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19199 (output_symbol_p): Move all logic to symtab.c
19200 (produce_symtab): Update.
19202 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19204 * config/s390/s390-opts.h (enum indirect_branch): Define.
19205 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19206 (s390_indirect_branch_via_thunk)
19207 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19208 (enum s390_indirect_branch_type): Define.
19209 * config/s390/s390.c (struct s390_frame_layout, struct
19210 machine_function): Remove.
19211 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19212 (indirect_branch_table_label_no, indirect_branch_table_name):
19214 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19215 (enum s390_indirect_branch_option): Define.
19216 (s390_return_addr_from_memory): New function.
19217 (s390_handle_string_attribute): New function.
19218 (s390_attribute_table): Add new attribute handler.
19219 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19220 (s390_indirect_branch_via_thunk): New function.
19221 (s390_indirect_branch_via_inline_thunk): New function.
19222 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19223 sibling call optimization for non z10 compiles.
19224 (s390_emit_call): Force indirect branch target to be a single
19225 register. Add r1 clobber for non-z10 compiles.
19226 (s390_emit_epilogue): Emit return jump via return_use expander.
19227 (s390_reorg): Handle JUMP_INSNs as execute targets.
19228 (s390_option_override_internal): Perform validity checks for the
19229 new command line options.
19230 (s390_indirect_branch_attrvalue): New function.
19231 (s390_indirect_branch_settings): New function.
19232 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19233 (s390_output_indirect_thunk_function): New function.
19234 (s390_code_end): Implement target hook.
19235 (s390_case_values_threshold): Implement target hook.
19236 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19238 * config/s390/s390.h (struct s390_frame_layout)
19239 (struct machine_function): Move here from s390.c.
19240 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19241 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19242 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19243 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19244 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19245 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19246 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19247 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19248 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19249 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19250 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19251 (mnemonic attribute): Add values which aren't recognized
19253 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19254 pattern for branch conversion. Fix mnemonic attribute.
19255 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19256 indirect branch via thunk if requested.
19257 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19258 ("*indirect_jump"): Disable for branch conversion using out of
19260 ("indirect_jump_via_thunk<mode>_z10")
19261 ("indirect_jump_via_thunk<mode>")
19262 ("indirect_jump_via_inlinethunk<mode>_z10")
19263 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19264 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19265 ("casesi_jump_via_inlinethunk<mode>_z10")
19266 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19267 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19268 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19269 ("*indirect2_jump"): Disable for branch conversion.
19270 ("casesi_jump"): Turn into expander and expand patterns for branch
19272 ("return_use"): New expander.
19273 ("*return"): Emit return via thunk and rename it to ...
19274 ("*return<mode>"): ... this one.
19275 * config/s390/s390.opt: Add new options and and enum for the
19278 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19280 * lra-constraints.c (match_reload): Unconditionally use
19281 gen_lowpart_SUBREG, rather than selecting between that
19282 and equivalent gen_rtx_SUBREG code.
19284 2018-02-08 Richard Biener <rguenther@suse.de>
19286 PR tree-optimization/84233
19287 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19288 changed flag instead of boguously re-using phi_inserted.
19290 2018-02-08 Martin Jambor <mjambor@suse.cz>
19292 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19293 static local variables.
19295 2018-02-08 Richard Biener <rguenther@suse.de>
19297 PR tree-optimization/84278
19298 * tree-vect-stmts.c (vectorizable_store): When looking for
19299 smaller vector types to perform grouped strided loads/stores
19300 make sure the mode is supported by the target.
19301 (vectorizable_load): Likewise.
19303 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19305 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19306 Increase LDP/STP opportunities by adding adjacent callee-saves.
19308 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19310 PR rtl-optimization/84068
19311 PR rtl-optimization/83459
19312 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19314 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19316 PR tree-optimization/84224
19317 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19318 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19319 non-zero arguments.
19321 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19324 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19325 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19326 count, remove one USE.
19328 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19330 * doc/install.texi (Configuration): Document the
19331 --with-long-double-format={ibm,ieee} PowerPC configuration
19335 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19336 Convert from define_expand to be define_insn_and_split. Rework
19337 float/double/_Float128 conversions to QI/HI/SImode to work with
19338 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19339 conversions to QI/HImode types did a store and then a load to
19340 truncate the value. For conversions to VSX registers, don't split
19341 the insn, instead emit the code directly. Use the code iterator
19342 any_fix to combine signed and unsigned conversions.
19343 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19344 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19345 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19346 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19347 (fix_<mode>di2_hw): Likewise.
19348 (fixuns_<mode>di2_hw): Likewise.
19349 (fix_<mode>si2_hw): Likewise.
19350 (fixuns_<mode>si2_hw): Likewise.
19351 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19352 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19353 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19354 fix<uns>_trunc<SFDF:mode>si2_p8.
19355 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19356 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19357 (fix<uns>_<mode>_mem): Likewise.
19358 (fctiw<u>z_<mode>_mem): Likewise.
19359 (fix<uns>_<mode>_mem): Likewise.
19360 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19361 the register allocator from doing a direct move to the GPRs to do
19362 a store, and instead use the ISA 3.0 store byte/half-word from
19363 vector register instruction. For IEEE 128-bit floating point,
19364 also optimize stores of 32-bit ints.
19365 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19367 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19369 * genextract.c (push_pathstr_operand): New function to support
19371 (walk_rtx): Call push_pathstr_operand.
19372 (print_path): Support [a-zA-Z].
19374 2018-02-07 Richard Biener <rguenther@suse.de>
19376 PR tree-optimization/84037
19377 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19378 (cse_and_gimplify_to_preheader): Declare.
19379 (vect_get_place_in_interleaving_chain): Likewise.
19380 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19382 (_loop_vec_info::~_loop_vec_info): Delete it.
19383 (cse_and_gimplify_to_preheader): New function.
19384 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19385 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19386 (vectorizable_load): Likewise. For grouped stores always base
19387 the IV on the first element.
19388 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19389 condition before gimplifying.
19391 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19393 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19394 *DIV_EXPR and *MOD_EXPR.
19396 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19399 * config/i386/i386.c (ix86_option_override_internal): Mask out
19400 the CF_SET bit when checking -fcf-protection.
19402 2018-02-07 Tom de Vries <tom@codesourcery.com>
19405 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19408 2018-02-07 Richard Biener <rguenther@suse.de>
19410 PR tree-optimization/84204
19411 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19414 PR tree-optimization/84205
19415 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19416 special-case isl_ast_op_zdiv_r.
19418 PR tree-optimization/84223
19419 * graphite-scop-detection.c (gather_bbs::before_dom_children):
19420 Only add conditions from within the region.
19421 (gather_bbs::after_dom_children): Adjust.
19423 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
19426 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19427 * config/avr/avr.md: Only post-reload split REG-REG moves if
19428 either register is GENERAL_REG_P.
19430 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19432 PR tree-optimization/84235
19433 * tree-ssa-scopedtables.c
19434 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
19435 if the subtraction is performed in floating point type where NaNs are
19436 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
19437 build 1. Formatting fix.
19439 2018-02-06 Jakub Jelinek <jakub@redhat.com>
19442 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
19443 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
19444 and skip it regardless of bb boundaries. Use CALL_P macro,
19445 don't test INSN_P (insn) together with CALL_P or JUMP_P check
19446 unnecessarily, formatting fix.
19448 2018-02-06 Michael Collison <michael.collison@arm.com>
19450 * config/arm/thumb2.md:
19451 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19452 (*thumb_mov_notscc): Ditto.
19454 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
19457 * config/rs6000/rs6000.md (su code attribute): Use "u" for
19458 unsigned_fix, not "s".
19460 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19462 * configure.ac (gcc_fn_eh_frame_ro): New function.
19463 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
19464 correct .eh_frame permissions.
19465 * configure: Regenerate.
19467 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
19469 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
19470 irrelevant options.
19472 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19474 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19475 Display warning message for -mno-speculate-indirect-jumps.
19477 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
19479 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
19481 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
19483 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
19485 PR tree-optimization/84225
19486 * tree-eh.c (find_trapping_overflow): Only call
19487 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
19489 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19492 * config/i386/i386.c: Reimplement the check of possible options
19493 -mibt/-mshstk conbination. Change error messages.
19494 * doc/invoke.texi: Fix a typo: remove extra '='.
19496 2018-02-06 Marek Polacek <polacek@redhat.com>
19498 PR tree-optimization/84228
19499 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
19501 2018-02-06 Tamar Christina <tamar.christina@arm.com>
19504 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
19505 emitted arch directives.
19506 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
19507 __ARM_FEATURE_COPROC before changing architectures.
19509 2018-02-06 Richard Biener <rguenther@suse.de>
19511 * config/i386/i386.c (print_reg): Fix typo.
19512 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
19514 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
19516 * configure: Regenerate.
19518 2018-02-05 Martin Sebor <msebor@redhat.com>
19520 PR tree-optimization/83369
19521 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
19524 2018-02-05 Martin Liska <mliska@suse.cz>
19526 * doc/invoke.texi: Cherry-pick upstream r323995.
19528 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
19530 * ira.c (ira_init_register_move_cost): Adjust comment.
19532 2018-02-05 Martin Liska <mliska@suse.cz>
19534 PR gcov-profile/84137
19535 * doc/gcov.texi: Fix typo in documentation.
19537 2018-02-05 Martin Liska <mliska@suse.cz>
19539 PR gcov-profile/83879
19540 * doc/gcov.texi: Document necessity of --dynamic-list-data when
19541 using dlopen functionality.
19543 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
19545 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
19546 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
19547 _mm_maskz_range_ss, _mm_mask_range_round_ss,
19548 _mm_maskz_range_round_ss): New intrinsics.
19549 (__builtin_ia32_rangesd128_round)
19550 (__builtin_ia32_rangess128_round): Remove.
19551 (__builtin_ia32_rangesd128_mask_round,
19552 __builtin_ia32_rangess128_mask_round): New builtins.
19553 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
19554 __builtin_ia32_rangess128_round): Remove.
19555 (__builtin_ia32_rangesd128_mask_round,
19556 __builtin_ia32_rangess128_mask_round): New builtins.
19557 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
19558 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
19559 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
19560 "<round_saeonly_constraint>")): Changed to ...
19561 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
19562 "<round_saeonly_scalar_constraint>")): ... this.
19563 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
19564 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
19565 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
19566 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
19567 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
19569 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
19571 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
19573 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
19574 Remove all values except native, 8540 and 8548.
19576 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
19578 * config/i386/i386.c (ix86_output_function_return): Pass
19579 INVALID_REGNUM, instead of -1, as invalid register number to
19580 indirect_thunk_name and output_indirect_thunk.
19582 2018-02-02 Julia Koval <julia.koval@intel.com>
19584 * config.gcc: Add -march=icelake.
19585 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
19586 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
19587 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
19588 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
19589 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
19590 (processor_target_table): Add icelake.
19591 (ix86_option_override_internal): Handle new PTAs.
19592 (get_builtin_code_for_version): Handle icelake.
19593 (M_INTEL_COREI7_ICELAKE): New.
19594 (fold_builtin_cpu): Handle icelake.
19595 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
19596 * doc/invoke.texi: Add -march=icelake.
19598 2018-02-02 Julia Koval <julia.koval@intel.com>
19600 * config/i386/i386.c (ix86_option_override_internal): Change flags type
19601 to wide_int_bitmask.
19602 * wide-int-bitmask.h: New.
19604 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19607 * config/i386/i386.md: Replace Pmode with word_mode in
19608 builtin_setjmp_setup and builtin_longjmp to support x32.
19610 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
19614 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19616 (rs6000_supported_cpu_names): New static variable.
19617 (linux_cpu_translation_table): Likewise.
19618 (elf_platform) <cpu>: Define new static variable and use it.
19619 Translate kernel AT_PLATFORM name to canonical name if needed.
19620 Error if platform name is unknown.
19622 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
19625 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19627 2018-02-01 Jeff Law <law@redhat.com>
19630 * config/i386/i386.c (release_scratch_register_on_entry): Add new
19631 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
19632 the scratch if RELEASE_VIA_POP is false.
19633 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19634 If we have to save a temporary register, decrement SIZE appropriately.
19635 Pass new arguments to release_scratch_register_on_entry.
19636 (ix86_adjust_stack_and_probe): Likewise.
19637 (ix86_emit_probe_stack_range): Pass new arguments to
19638 release_scratch_register_on_entry.
19640 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
19642 PR rtl-optimization/84157
19643 * combine.c (change_zero_ext): Use REG_P predicate in
19644 front of HARD_REGISTER_P predicate.
19646 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
19648 * config/avr/avr.c (avr_option_override): Move disabling of
19649 -fdelete-null-pointer-checks to...
19650 * common/config/avr/avr-common.c (avr_option_optimization_table):
19653 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19655 PR tree-optimization/81635
19656 * tree-data-ref.c (split_constant_offset_1): For types that
19657 wrap on overflow, try to use range info to prove that wrapping
19660 2018-02-01 Renlin Li <renlin.li@arm.com>
19663 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19664 TAILCALL_ADDR_REGS.
19665 (aarch64_register_move_cost): Likewise.
19666 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19667 TAILCALL_ADDR_REGS.
19668 (REG_CLASS_NAMES): Likewise.
19669 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19670 TAILCALL_ADDR_REGS. Remove IP registers.
19671 * config/aarch64/aarch64.md (Ucs): Update register constraint.
19673 2018-02-01 Richard Biener <rguenther@suse.de>
19675 * domwalk.h (dom_walker::dom_walker): Add additional constructor
19676 for specifying RPO order and allow NULL for that.
19677 * domwalk.c (dom_walker::dom_walker): Likewise.
19678 (dom_walker::walk): Handle NULL RPO order.
19679 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19681 (rewrite_update_dom_walker): Likewise.
19682 (mark_def_dom_walker): Likewise.
19684 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19686 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19687 (aarch64_maybe_expand_sve_subreg_move): Declare.
19688 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19689 * config/aarch64/predicates.md (aarch64_any_register_operand): New
19691 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19692 that are semantically a reverse operation.
19693 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19694 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19695 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19697 (aarch64_can_change_mode_class): For big-endian, forbid changes
19698 between two SVE modes if they have different element sizes.
19700 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19702 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19703 the TImode handling for big-endian targets.
19705 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19707 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19708 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
19710 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19711 Remove BSWAP handing for big-endian targets and use the form of
19712 LD1RQ appropariate for the mode.
19714 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19716 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19717 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19718 duplicated element.
19720 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19723 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19724 check for operands that need to go through aarch64_sve_reload_be.
19726 2018-02-01 Jakub Jelinek <jakub@redhat.com>
19728 PR tree-optimization/81661
19729 PR tree-optimization/84117
19730 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19731 * tree-eh.c: Include gimplify.h.
19732 (find_trapping_overflow, replace_trapping_overflow,
19733 rewrite_to_non_trapping_overflow): New functions.
19734 * tree-vect-loop.c: Include tree-eh.h.
19735 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
19736 * tree-data-ref.c: Include tree-eh.h.
19737 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
19739 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
19741 PR rtl-optimization/84123
19742 * combine.c (change_zero_ext): Check if hard register satisfies
19743 can_change_dest_mode before calling gen_lowpart_SUBREG.
19745 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
19748 * ira.c (ira_init_register_move_cost): Remove assert.
19750 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19752 PR rtl-optimization/84071
19753 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
19754 * doc/tm.texi: Regenerate.
19756 2018-01-31 Richard Biener <rguenther@suse.de>
19758 PR tree-optimization/84132
19759 * tree-data-ref.c (analyze_miv_subscript): Properly
19760 check whether evolution_function_is_affine_multivariate_p
19761 before calling gcd_of_steps_may_divide_p.
19763 2018-01-31 Julia Koval <julia.koval@intel.com>
19766 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
19767 * config/i386/i386.md (rdpid_rex64) New.
19768 (rdpid): Make 32bit only.
19770 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
19773 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
19774 an IDENTIFIER_NODE for FUNCTION_TYPE's.
19776 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19779 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19781 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19783 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19785 PR rtl-optimization/84071
19786 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
19787 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
19789 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19791 * config/arc/arc.c (arc_handle_aux_attribute): New function.
19792 (arc_attribute_table): Add 'aux' attribute.
19793 (arc_in_small_data_p): Consider aux like variables.
19794 (arc_is_aux_reg_p): New function.
19795 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
19796 (arc_get_aux_arg): New function.
19797 (prepare_move_operands): Handle aux-register access.
19798 (arc_handle_aux_attribute): New function.
19799 * doc/extend.texi (ARC Variable attributes): Add subsection.
19801 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19803 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
19804 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
19805 (arc_attribute_table): Add 'uncached' attribute.
19806 (arc_print_operand): Print '.di' flag for uncached memory
19808 (arc_in_small_data_p): Do not consider for small data the uncached
19810 (arc_is_uncached_mem_p): New function.
19811 * config/arc/predicates.md (compact_store_memory_operand): Check
19812 for uncached memory accesses.
19813 (nonvol_nonimm_operand): Likewise.
19814 * doc/extend.texi (ARC Type Attribute): New subsection.
19816 2018-01-31 Jakub Jelinek <jakub@redhat.com>
19819 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
19820 falign-loops=): Add Optimization flag.
19822 2018-01-30 Jeff Law <law@redhat.com>
19825 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
19826 INT_REGISTERS_SAVED. Check it prior to calling
19827 get_scratch_register_on_entry.
19828 (ix86_adjust_stack_and_probe): Similarly.
19829 (ix86_emit_probe_stack_range): Similarly.
19830 (ix86_expand_prologue): Corresponding changes.
19832 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19835 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
19836 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
19838 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
19841 * lra-constraints.c (curr_insn_transform): Process AND in the
19844 2018-01-30 Jakub Jelinek <jakub@redhat.com>
19846 PR rtl-optimization/83986
19847 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
19848 dependence against last_pending_memory_flush in addition to
19849 pending_jump_insns.
19851 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
19853 PR tree-optimization/81611
19854 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
19857 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19860 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
19863 2018-01-30 Richard Biener <rguenther@suse.de>
19864 Jakub Jelinek <jakub@redhat.com>
19866 PR tree-optimization/84111
19867 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
19868 inner loops added during recursion, as they don't have up-to-date
19871 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19874 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
19875 (can_inline_edge_by_limits_p): ... here.
19876 (can_early_inline_edge_p, check_callers,
19877 update_caller_keys, update_callee_keys, recursive_inlining,
19878 add_new_edges_to_heap, speculation_useful_p,
19879 inline_small_functions,
19880 inline_small_functions, flatten_function,
19881 inline_to_all_callers_1): Update.
19883 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19885 * profile-count.c (profile_count::combine_with_ipa_count): Handle
19888 2018-01-30 Richard Biener <rguenther@suse.de>
19890 PR tree-optimization/83008
19891 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
19892 invariant and constant vector uses in stmts when they need
19893 more than one stmt.
19895 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19898 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
19899 * configure: Regenerate.
19901 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19903 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
19905 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
19906 Use gen_rtx_REG rather than gen_lowpart.
19908 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19910 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
19911 rather than 0 when creating partial subregs.
19913 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19915 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
19918 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
19921 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
19922 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
19923 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
19924 flags. This restores the settings used before the 2017-07-24.
19925 Turning off pre increment/decrement/modify allows IVOPTS to
19926 optimize DF/SF loops where the index is an int.
19928 2018-01-29 Richard Biener <rguenther@suse.de>
19929 Kelvin Nilsen <kelvin@gcc.gnu.org>
19932 * tree-vect-stmts.c (vectorizable_call): Don't call
19933 targetm.vectorize_builtin_md_vectorized_function if callee is
19936 2018-01-22 Carl Love <cel@us.ibm.com>
19938 * doc/extend.tex: Fix typo in second arg in
19939 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
19941 2018-01-29 Richard Biener <rguenther@suse.de>
19943 PR tree-optimization/84086
19944 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
19945 (flush_ssaname_freelist): When SSA names were released reset
19946 the SCEV hash table.
19948 2018-01-29 Richard Biener <rguenther@suse.de>
19950 PR tree-optimization/84057
19951 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
19952 removed paths when removing edges.
19954 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
19956 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
19957 -mfunction-return=@var{choice}.
19959 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19961 PR diagnostic/84034
19962 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
19963 Handle CR like TAB.
19964 (layout::print_source_line): Likewise.
19965 (test_get_line_width_without_trailing_whitespace): Add test cases.
19967 2018-01-27 Jakub Jelinek <jakub@redhat.com>
19969 PR middle-end/84040
19970 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
19973 2018-01-26 Jim Wilson <jimw@sifive.com>
19975 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
19977 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
19980 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19982 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
19983 and CMP + SUB-immediate -> SUBS.
19985 2018-01-26 Martin Sebor <msebor@redhat.com>
19987 PR tree-optimization/83896
19988 * tree-ssa-strlen.c (get_string_len): Rename...
19989 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
19990 Avoid assuming length is constant.
19991 (handle_char_store): Use HOST_WIDE_INT for string length.
19993 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
19996 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
19997 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
19999 2018-01-26 Richard Biener <rguenther@suse.de>
20001 PR rtl-optimization/84003
20002 * dse.c (record_store): Only record redundant stores when
20003 the earlier store aliases at least all accesses the later one does.
20005 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20007 PR rtl-optimization/83985
20008 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20009 REG_CFA_RESTORE insns.
20010 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20011 REG_CFA_RESTORE insns here.
20014 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20015 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20017 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20019 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20020 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20021 (arc_init): Likewise.
20022 (arc_override_options): Likewise.
20023 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20025 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20027 * config/arc/arc.h (TARGET_DBNZ): Define.
20028 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20029 properly set the tune attribute.
20030 (dbnz): Use TARGET_DBNZ guard.
20031 * config/arc/arc.opt (mtune): Add core3 option.
20033 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20035 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20036 recognize new pic like addresses.
20037 (arc_delegitimize_address): Clean up.
20039 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20041 * config/arc/arc-arches.def: Option mrf16 valid for all
20043 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20044 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20045 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20046 * config/arc/arc-tables.opt: Regenerate.
20047 * config/arc/arc.c (arc_conditional_register_usage): Handle
20048 reduced register file case.
20049 (arc_file_start): Set must have build attributes.
20050 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20051 mrf16 option value.
20052 * config/arc/arc.opt (mrf16): Add new option.
20053 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20054 * config/arc/genmultilib.awk: Handle new mrf16 option.
20055 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20056 * config/arc/t-multilib: Regenerate.
20057 * doc/invoke.texi (ARC Options): Document mrf16 option.
20059 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20061 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20062 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20063 (arc_attribute_table): Add 'secure_call' attribute.
20064 (arc_print_operand): Print secure call operand.
20065 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20067 (arc_is_secure_call_p): New function. * config/arc/arc.md
20068 (call_i): Add support for sjli instruction.
20069 (call_value_i): Likewise.
20070 * config/arc/constraints.md (Csc): New constraint.
20072 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20073 John Eric Martin <John.Martin@emmicro-us.com>
20075 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20076 * config/arc/arc.c (_arc_jli_section): New struct.
20077 (arc_jli_section): New type.
20078 (rc_jli_sections): New static variable.
20079 (arc_handle_jli_attribute): New function.
20080 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20081 (arc_file_end): New function.
20082 (TARGET_ASM_FILE_END): Define.
20083 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20084 (arc_add_jli_section): New function.
20085 (jli_call_scan): Likewise.
20086 (arc_reorg): Call jli_call_scan.
20087 (arc_output_addsi): Remove 'S' from printing asm operand.
20088 (arc_is_jli_call_p): New function.
20089 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20091 (movhi_insn): Likewise.
20092 (movsi_insn): Likewise.
20093 (movsi_set_cc_insn): Likewise.
20094 (loadqi_update): Likewise.
20095 (load_zeroextendqisi_update): Likewise.
20096 (load_signextendqisi_update): Likewise.
20097 (loadhi_update): Likewise.
20098 (load_zeroextendhisi_update): Likewise.
20099 (load_signextendhisi_update): Likewise.
20100 (loadsi_update): Likewise.
20101 (loadsf_update): Likewise.
20102 (movsicc_insn): Likewise.
20103 (bset_insn): Likewise.
20104 (bxor_insn): Likewise.
20105 (bclr_insn): Likewise.
20106 (bmsk_insn): Likewise.
20107 (bicsi3_insn): Likewise.
20108 (cmpsi_cc_c_insn): Likewise.
20109 (movsi_ne): Likewise.
20110 (movsi_cond_exec): Likewise.
20111 (clrsbsi2): Likewise.
20112 (norm_f): Likewise.
20122 (call_i): Remove 'S' asm letter, add jli instruction.
20123 (call_value_i): Likewise.
20124 * config/arc/arc.op (mjli-always): New option.
20125 * config/arc/constraints.md (Cji): New constraint.
20126 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20128 (subsf3_fpx): Likewise.
20129 (mulsf3_fpx): Likewise.
20130 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20132 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20133 function attrbutes.
20134 * doc/invoke.texi (ARC): Document mjli-always option.
20136 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20138 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20139 avoid addition with 0 and use incw and decw where possible.
20141 2018-01-26 Richard Biener <rguenther@suse.de>
20143 PR tree-optimization/81082
20144 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20145 association if it requires casting to unsigned.
20146 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20147 from fold_plusminus_mult_expr to catch important cases late when
20148 range info is available.
20150 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20152 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20153 * configure.ac (hidden_linkonce): New test.
20154 * configure: Regenerate.
20155 * config.in: Regenerate.
20157 2018-01-26 Julia Koval <julia.koval@intel.com>
20159 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20160 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20161 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20162 _mm_mask_bitshuffle_epi64_mask): Fix type.
20163 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20164 USI_FTYPE_V4DI_V4DI_USI): Remove.
20165 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20166 __builtin_ia32_vpshufbitqmb256_mask,
20167 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20168 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20169 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20171 2018-01-26 Alan Modra <amodra@gmail.com>
20174 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20175 UNSPEC_VBPERMQ. Sort other unspecs.
20177 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20179 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20181 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20183 PR middle-end/83055
20184 * predict.c (drop_profile): Do not push/pop cfun; update also
20186 (handle_missing_profiles): Fix logic looking for zero profiles.
20188 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20190 PR middle-end/83977
20191 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20192 on functions with #pragma omp declare simd or functions with simd
20194 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20195 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20196 Remove trailing \n from warning_at calls.
20198 2018-01-25 Tom de Vries <tom@codesourcery.com>
20201 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20202 for neutered workers.
20204 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20207 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20208 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20210 2018-01-24 Jeff Law <law@redhat.com>
20213 * i386.c (get_probe_interval): Move to earlier point.
20214 (ix86_compute_frame_layout): If -fstack-clash-protection and
20215 the frame is larger than the probe interval, then use pushes
20216 to save registers rather than reg->mem moves.
20217 (ix86_expand_prologue): Remove conditional for int_registers_saved
20220 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20223 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20224 min/max for never referenced object.
20226 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20228 PR middle-end/83977
20229 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20231 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20232 attributes from DECL_ATTRIBUTES (decl) without affecting
20233 DECL_ATTRIBUTES (current_function_decl).
20234 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20235 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20237 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20239 PR tree-optimization/83979
20240 * fold-const.c (fold_comparison): Use constant_boolean_node
20241 instead of boolean_{true,false}_node.
20243 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20245 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20248 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20250 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20251 Simplify the clause that sets the length attribute.
20252 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20253 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20254 clause that sets the length attribute.
20255 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20257 2018-01-24 Tom de Vries <tom@codesourcery.com>
20260 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20261 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20262 Add strict parameter.
20263 (prevent_branch_around_nothing): Insert dummy insn between branch to
20264 label and label with no ptx insn inbetween.
20265 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20267 2018-01-24 Tom de Vries <tom@codesourcery.com>
20270 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20271 for neutered threads in warp.
20272 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20274 2018-01-24 Richard Biener <rguenther@suse.de>
20276 PR tree-optimization/83176
20277 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20280 2018-01-24 Richard Biener <rguenther@suse.de>
20282 PR tree-optimization/82819
20283 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20284 code generating pluses that are no-ops in the target precision.
20286 2018-01-24 Richard Biener <rguenther@suse.de>
20288 PR middle-end/84000
20289 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20291 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20293 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20294 to merge probabilities.
20295 * predict.c (probably_never_executed): Also mark as cold functions
20296 with global 0 profile and guessed local profile.
20297 * profile-count.c (profile_probability::combine_with_count): New
20299 * profile-count.h (profile_probability::operator*,
20300 profile_probability::operator*=, profile_probability::operator/,
20301 profile_probability::operator/=): Reduce precision to adjusted
20302 and set value to guessed on contradictory divisions.
20303 (profile_probability::combine_with_freq): Remove.
20304 (profile_probability::combine_wiht_count): Declare.
20305 (profile_count::force_nonzero):: Set to adjusted.
20306 (profile_count::probability_in):: Set quality to adjusted.
20307 * tree-ssa-tail-merge.c (replace_block_by): Use
20308 combine_with_count.
20310 2018-01-23 Andrew Waterman <andrew@sifive.com>
20311 Jim Wilson <jimw@sifive.com>
20313 * config/riscv/riscv.c (riscv_stack_boundary): New.
20314 (riscv_option_override): Set riscv_stack_boundary. Handle
20315 riscv_preferred_stack_boundary_arg.
20316 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20317 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20318 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20319 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20320 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20321 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20323 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20326 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20327 of struct ix86_frame.
20328 (ix86_expand_epilogue): Likewise. Add a local variable for
20329 the reg_save_offset field in struct ix86_frame.
20331 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20333 PR tree-optimization/82604
20334 * tree-loop-distribution.c (enum partition_kind): New enum item
20335 PKIND_PARTIAL_MEMSET.
20336 (partition_builtin_p): Support above new enum item.
20337 (generate_code_for_partition): Ditto.
20338 (compute_access_range): Differentiate cases that equality can be
20339 proven at all loops, the innermost loops or no loops.
20340 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20341 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20342 (finalize_partitions, distribute_loop): Don't fuse partition of
20343 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20344 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20345 parloop is enabled.
20347 2018-01-23 Martin Liska <mliska@suse.cz>
20349 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20350 order to ignore the predictor.
20351 (PRED_POLYMORPHIC_CALL): Likewise.
20352 (PRED_RECURSIVE_CALL): Likewise.
20354 2018-01-23 Martin Liska <mliska@suse.cz>
20356 * tree-profile.c (tree_profiling): Print function header to
20357 aware reader which function we are working on.
20358 * value-prof.c (gimple_find_values_to_profile): Do not print
20359 not interesting value histograms.
20361 2018-01-23 Martin Liska <mliska@suse.cz>
20363 * profile-count.h (enum profile_quality): Add
20364 profile_uninitialized as the first value. Do not number values
20365 as they are zero based.
20366 (profile_count::verify): Update sanity check.
20367 (profile_probability::verify): Likewise.
20369 2018-01-23 Nathan Sidwell <nathan@acm.org>
20371 * doc/invoke.texi (ffor-scope): Deprecate.
20373 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20375 PR tree-optimization/83510
20376 * domwalk.c (set_all_edges_as_executable): New function.
20377 (dom_walker::dom_walker): Convert bool param
20378 "skip_unreachable_blocks" to enum reachability. Move setup of
20379 edge flags to set_all_edges_as_executable and only do it when
20380 reachability is REACHABLE_BLOCKS.
20381 * domwalk.h (enum dom_walker::reachability): New enum.
20382 (dom_walker::dom_walker): Convert bool param
20383 "skip_unreachable_blocks" to enum reachability.
20384 (set_all_edges_as_executable): New decl.
20385 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20386 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20388 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20389 but converting true to REACHABLE_BLOCKS.
20390 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20392 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20393 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20394 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20396 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20397 if check_all_array_refs will be called.
20399 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20401 * tree.c (selftest::test_location_wrappers): Add more test
20404 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20406 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20407 (selftest::test_bit_in_range): Likewise.
20409 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20412 * doc/sourcebuild.texi (vect_float): Say that the selector
20413 only describes the situation when -funsafe-math-optimizations is on.
20414 (vect_float_strict): Document.
20416 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20418 PR tree-optimization/83965
20419 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20420 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20421 instead of checking only for a reduction.
20422 (vect_recog_widen_sum_pattern): Likewise.
20424 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20426 * predict.c (probably_never_executed): Only use precise profile info.
20427 (compute_function_frequency): Skip after inlining hack since we now
20428 have quality checking.
20430 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20432 * profile-count.h (profile_probability::very_unlikely,
20433 profile_probability::unlikely, profile_probability::even): Set
20434 precision to guessed.
20436 2018-01-23 Richard Biener <rguenther@suse.de>
20438 PR tree-optimization/83963
20439 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
20440 Properly terminate dominator walk when crossing the exit edge not
20441 when visiting its source block.
20443 2018-01-23 Jakub Jelinek <jakub@redhat.com>
20446 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
20447 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
20449 2018-01-22 Jakub Jelinek <jakub@redhat.com>
20451 PR tree-optimization/83957
20452 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
20453 semicolon after for body surrounded by braces.
20455 PR tree-optimization/83081
20456 * profile-count.h (profile_probability::split): New method.
20457 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
20458 Use profile_probability::split.
20459 (do_compare_rtx_and_jump): Fix adjustment of probabilities
20460 when splitting a single conditional jump into 2.
20462 2018-01-22 David Malcolm <dmalcolm@redhat.com>
20464 PR tree-optimization/69452
20465 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
20468 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20470 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
20471 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
20472 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
20474 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20476 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
20478 * config/rl78/rl78.md (movdi): New define_expand.
20479 * config/rl78/rl78.c (rl78_split_movdi): New function.
20481 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
20484 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
20486 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
20487 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
20488 128-bit to produce an UNSPEC move to get the double word with the
20489 signbit and then a shift directly to do signbit.
20490 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
20491 implementation with a new version that just does either a direct
20492 move or a regular move. Move memory interface to separate insns.
20493 Move insns so they are next to the expander.
20494 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
20495 with signbit move. Split big and little endian case.
20496 (signbit<mode>2_dm_mem_le): Likewise.
20497 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
20498 (signbit<mode>2_dm2): Likewise.
20500 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20502 * config/rl78/rl78.md (anddi3): New define_expand.
20504 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20506 * config/rl78/rl78.md (umindi3): New define_expand.
20508 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20510 * config/rl78/rl78.md (smindi3): New define_expand.
20512 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20514 * config/rl78/rl78.md (smaxdi3): New define_expand.
20516 2018-01-22 Carl Love <cel@us.ibm.com>
20518 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
20519 LVX_V1TI): Add macro expansion.
20520 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
20521 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
20522 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
20523 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
20524 Change check to determine if the instruction is a byte reversing
20525 entry. Fix typo in comment.
20526 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
20527 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
20528 Add def_builtin calls for new builtins.
20529 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
20530 Add define_insn expansion.
20532 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20534 * config/rl78/rl78.md (umaxdi3): New define_expand.
20536 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20538 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
20539 for non-QImode registers.
20541 2018-01-22 Richard Biener <rguenther@suse.de>
20543 PR tree-optimization/83963
20544 * graphite-scop-detection.c (scop_detection::get_sese): Delay
20545 including the loop exit block.
20546 (scop_detection::merge_sese): Likewise.
20547 (scop_detection::add_scop): Do it here instead.
20549 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20551 * doc/sourcebuild.texi (arm_softfloat): Document.
20553 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
20556 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
20557 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
20558 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
20560 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20561 David Edelsohn <dje.gcc@gmail.com>
20564 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20565 Change "crset eq" to "crset 2".
20566 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20567 (*call_indirect_aix<mode>_nospec): Likewise.
20568 (*call_value_indirect_aix<mode>_nospec): Likewise.
20569 (*call_indirect_elfv2<mode>_nospec): Likewise.
20570 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
20571 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
20572 change assembly output from . to $.
20573 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20574 (indirect_jump<mode>_nospec): Change assembly output from . to $.
20575 (*tablejump<mode>_internal1_nospec): Likewise.
20577 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
20580 * config/sh/sh_optimize_sett_clrt.cc:
20581 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
20583 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
20585 PR tree-optimization/83940
20586 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
20587 offset_dt to vect_constant_def rather than vect_unknown_def_type.
20588 (vect_check_load_store_mask): Add a mask_dt_out parameter and
20589 use it to pass back the definition type.
20590 (vect_check_store_rhs): Likewise rhs_dt_out.
20591 (vect_build_gather_load_calls): Add a mask_dt argument and use
20592 it instead of a call to vect_is_simple_use.
20593 (vectorizable_store): Update calls to vect_check_load_store_mask
20594 and vect_check_store_rhs. Use the dt returned by the latter instead
20595 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
20596 instead of calls to vect_is_simple_use. Pass the scalar rather
20597 than the vector operand to vect_is_simple_use when handling
20598 second and subsequent copies of an rhs value.
20599 (vectorizable_load): Update calls to vect_check_load_store_mask
20600 and vect_build_gather_load_calls. Use the cached mask_dt and
20601 gs_info.offset_dt instead of calls to vect_is_simple_use.
20603 2018-01-20 Jakub Jelinek <jakub@redhat.com>
20605 PR middle-end/83945
20606 * tree-emutls.c: Include gimplify.h.
20607 (lower_emutls_2): New function.
20608 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20609 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20610 it before further processing.
20613 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20614 UINTVAL (trueop1) instead of INTVAL (op1).
20616 2018-01-19 Jakub Jelinek <jakub@redhat.com>
20620 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20621 INCOMING_FRAME_SP_OFFSET if not defined.
20622 (scan_trace): Add ENTRY argument. If true and
20623 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20624 emit a note to adjust the CFA offset.
20625 (create_cfi_notes): Adjust scan_trace callers.
20626 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20627 INCOMING_FRAME_SP_OFFSET in the CIE.
20628 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20629 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20631 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20632 * doc/tm.texi: Regenerated.
20634 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20636 PR rtl-optimization/83147
20637 * lra-constraints.c (remove_inheritance_pseudos): Use
20638 lra_substitute_pseudo_within_insn.
20640 2018-01-19 Tom de Vries <tom@codesourcery.com>
20641 Cesar Philippidis <cesar@codesourcery.com>
20644 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20646 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
20649 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20650 spaces for function labels.
20652 2018-01-19 Martin Liska <mliska@suse.cz>
20654 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20655 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20656 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20657 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20658 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20659 (PRED_CONST_RETURN): Change from 69 to 65.
20660 (PRED_NULL_RETURN): Change from 91 to 71.
20661 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20662 (PRED_LOOP_GUARD): Change from 66 to 73.
20664 2018-01-19 Martin Liska <mliska@suse.cz>
20666 * predict.c (predict_insn_def): Add new assert.
20667 (struct branch_predictor): Change type to signed integer.
20668 (test_prediction_value_range): Amend test to cover
20669 PROB_UNINITIALIZED.
20670 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20671 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20672 (PRED_LOOP_ITERATIONS_MAX): Likewise.
20673 (PRED_LOOP_IV_COMPARE): Likewise.
20674 * predict.h (PROB_UNINITIALIZED): Define new constant.
20676 2018-01-19 Martin Liska <mliska@suse.cz>
20678 * predict.c (dump_prediction): Add new format for
20679 analyze_brprob.py script which is enabled with -details
20681 * profile-count.h (precise_p): New function.
20683 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20685 PR tree-optimization/83922
20686 * tree-vect-loop.c (vect_verify_full_masking): Return false if
20687 there are no statements that need masking.
20688 (vect_active_double_reduction_p): New function.
20689 (vect_analyze_loop_operations): Use it when handling phis that
20690 are not in the loop header.
20692 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20694 PR tree-optimization/83914
20695 * tree-vect-loop.c (vectorizable_induction): Don't convert
20696 init_expr or apply the peeling adjustment for inductions
20697 that are nested within the vectorized loop.
20699 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20701 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20704 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20708 * function.h (gimplify_parameters): Add gimple_seq * argument.
20709 * function.c: Include gimple.h and options.h.
20710 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20711 for the added local temporaries if needed.
20712 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20713 if there are any parameter cleanups, wrap whole body into a
20714 try/finally with the cleanups.
20716 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
20719 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20720 Use GET_MODE_CLASS for scalar floating point.
20722 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20726 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20727 Fix call of call_cgraph_insertion_hooks.
20729 2018-01-18 Martin Sebor <msebor@redhat.com>
20731 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
20733 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20736 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
20739 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
20742 * common.opt: (-ffile-prefix-map): New option.
20743 * opts.c (common_handle_option): Defer it.
20744 * opts-global.c (handle_common_deferred_options): Handle it.
20745 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
20746 * file-prefix-map.h: New file.
20747 (remap_debug_filename, add_debug_prefix_map): ...here.
20748 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
20749 * final.c (debug_prefix_map, add_debug_prefix_map
20750 remap_debug_filename): Move to...
20751 * file-prefix-map.c: New file.
20752 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
20753 generalize, get rid of alloca(), use strrchr() instead of strchr().
20754 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
20755 Implement in terms of add_prefix_map().
20756 (remap_macro_filename, remap_debug_filename): Implement in term of
20758 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
20759 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
20760 * dbxout.c: Include file-prefix-map.h.
20761 * varasm.c: Likewise.
20762 * vmsdbgout.c: Likewise.
20763 * xcoffout.c: Likewise.
20764 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
20765 * doc/cppopts.texi (-fmacro-prefix-map): Document.
20766 * doc/invoke.texi (-ffile-prefix-map): Document.
20767 (-fdebug-prefix-map): Update description.
20769 2018-01-18 Martin Liska <mliska@suse.cz>
20771 * config/i386/i386.c (indirect_thunk_name): Document that also
20773 (output_indirect_thunk): Document why both instructions
20774 (pause and lfence) are generated.
20776 2018-01-18 Richard Biener <rguenther@suse.de>
20778 PR tree-optimization/83887
20779 * graphite-scop-detection.c
20780 (scop_detection::get_nearest_dom_with_single_entry): Remove.
20781 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
20782 (scop_detection::merge_sese): Re-implement with a flood-fill
20783 algorithm that properly finds a SESE region if it exists.
20785 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20788 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
20789 pointer_diff optimizations use view_convert instead of convert.
20791 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20793 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20794 Generate different code for -mno-speculate-indirect-jumps.
20795 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20796 (*call_indirect_aix<mode>): Disable for
20797 -mno-speculate-indirect-jumps.
20798 (*call_indirect_aix<mode>_nospec): New define_insn.
20799 (*call_value_indirect_aix<mode>): Disable for
20800 -mno-speculate-indirect-jumps.
20801 (*call_value_indirect_aix<mode>_nospec): New define_insn.
20802 (*sibcall_nonlocal_sysv<mode>): Generate different code for
20803 -mno-speculate-indirect-jumps.
20804 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20806 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
20808 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
20809 long double type, set the flags for noting the default long double
20810 type, even if we don't pass or return a long double type.
20812 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
20815 * ipa-inline.c (flatten_function): Do not overwrite final inlining
20818 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
20820 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
20821 support for merge[hl].
20822 (fold_mergehl_helper): New helper function.
20823 (tree-vector-builder.h): New #include for tree_vector_builder usage.
20824 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
20825 (altivec_vmrglw_direct): Add xxmrglw insn.
20827 2018-01-17 Andrew Waterman <andrew@sifive.com>
20829 * config/riscv/riscv.c (riscv_conditional_register_usage): If
20830 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
20832 2018-01-17 David Malcolm <dmalcolm@redhat.com>
20835 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
20836 call the lto_location_cache before reading the
20837 DECL_SOURCE_LOCATION of the types.
20839 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
20840 Richard Sandiford <richard.sandiford@linaro.org>
20842 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
20843 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
20844 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
20845 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
20846 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
20848 * config/aarch64/constraints.md (aarch64_movti_operand):
20850 * config/aarch64/predicates.md (Uti): Add new constraint.
20852 2018-01-17 Carl Love <cel@us.ibm.com>
20854 * config/rs6000/vsx.md (define_expand xl_len_r,
20855 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
20856 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
20858 (define_expand, define_insn): Move the shift left from the
20859 define_insn to the define_expand for lxvl and stxvl instructions.
20860 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
20861 and XL_LEN_R definitions to PURE.
20863 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
20865 * config/i386/i386.c (indirect_thunk_name): Declare regno
20866 as unsigned int. Compare regno with INVALID_REGNUM.
20867 (output_indirect_thunk): Ditto.
20868 (output_indirect_thunk_function): Ditto.
20869 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
20870 in the call to output_indirect_thunk_function.
20872 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
20874 PR middle-end/83884
20875 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
20876 rather than the size of inner_type to determine the stack slot size
20877 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
20879 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
20882 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
20885 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
20887 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
20888 endian Linux systems to optionally enable multilibs for selecting
20889 the long double type if the user configured an explicit type.
20890 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
20891 have no long double multilibs if not defined.
20892 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
20893 warn if the user used -mabi={ieee,ibm}longdouble and we built
20894 multilibs for long double.
20895 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
20896 appropriate multilib option.
20897 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
20899 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
20900 for building long double multilibs.
20901 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
20903 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
20905 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
20908 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
20910 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
20913 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
20916 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
20919 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20921 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
20922 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
20924 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
20926 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
20927 different rtl trees depending on TARGET_64BIT.
20928 (rs6000_gen_lvx): Likewise.
20930 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20932 * config/visium/visium.md (nop): Tweak comment.
20933 (hazard_nop): Likewise.
20935 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20937 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
20938 -mspeculate-indirect-jumps.
20939 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
20940 for -mno-speculate-indirect-jumps.
20941 (*call_indirect_elfv2<mode>_nospec): New define_insn.
20942 (*call_value_indirect_elfv2<mode>): Disable for
20943 -mno-speculate-indirect-jumps.
20944 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
20945 (indirect_jump): Emit different RTL for
20946 -mno-speculate-indirect-jumps.
20947 (*indirect_jump<mode>): Disable for
20948 -mno-speculate-indirect-jumps.
20949 (*indirect_jump<mode>_nospec): New define_insn.
20950 (tablejump): Emit different RTL for
20951 -mno-speculate-indirect-jumps.
20952 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
20953 (tablejumpsi_nospec): New define_expand.
20954 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
20955 (tablejumpdi_nospec): New define_expand.
20956 (*tablejump<mode>_internal1): Disable for
20957 -mno-speculate-indirect-jumps.
20958 (*tablejump<mode>_internal1_nospec): New define_insn.
20959 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
20962 2018-01-16 Artyom Skrobov tyomitch@gmail.com
20964 * caller-save.c (insert_save): Drop unnecessary parameter. All
20967 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20968 Richard Biener <rguenth@suse.de>
20971 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
20972 return early, inline manually is_gimple_sizepos. Make sure if we
20973 call gimplify_expr we don't end up with a gimple constant.
20974 * tree.c (variably_modified_type_p): Don't return true for
20975 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
20976 * gimplify.h (is_gimple_sizepos): Remove.
20978 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20980 PR tree-optimization/83857
20981 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
20982 vectorizable_live_operation for pure SLP statements.
20983 (vectorizable_live_operation): Handle PHIs.
20985 2018-01-16 Richard Biener <rguenther@suse.de>
20987 PR tree-optimization/83867
20988 * tree-vect-stmts.c (vect_transform_stmt): Precompute
20989 nested_in_vect_loop_p since the scalar stmt may get invalidated.
20991 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20994 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
20995 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
20996 If off is not INTEGER_CST, issue a may not be aligned warning
20997 rather than isn't aligned. Use isn%'t rather than isn't.
20998 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21000 <case MULT_EXPR>: Improve the case when bottom and one of the
21001 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21002 operand, in that case check if the other operand is multiple of
21003 bottom divided by the INTEGER_CST operand.
21005 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21008 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21009 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21010 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21011 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21012 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21013 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21014 (pa_function_arg_size): New function.
21016 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21018 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21019 in a separate statement.
21021 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21023 PR tree-optimization/83847
21024 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21025 group gathers and scatters.
21027 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21029 PR rtl-optimization/86620
21030 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21032 PR rtl-optimization/83213
21033 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21034 to last if both are JUMP_INSNs.
21036 PR tree-optimization/83843
21037 * gimple-ssa-store-merging.c
21038 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21039 store_immediate_info for bswap/nop orig_stores.
21041 2018-01-15 Andrew Waterman <andrew@sifive.com>
21043 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21045 <UDIV>: Increase cost if !TARGET_DIV.
21047 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21049 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21050 (define_attr "cr_logical_3op"): New.
21051 (cceq_ior_compare): Adjust.
21052 (cceq_ior_compare_complement): Adjust.
21053 (*cceq_rev_compare): Adjust.
21054 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21055 (is_cracked_insn): Adjust.
21056 (insn_must_be_first_in_group): Adjust.
21057 * config/rs6000/40x.md: Adjust.
21058 * config/rs6000/440.md: Adjust.
21059 * config/rs6000/476.md: Adjust.
21060 * config/rs6000/601.md: Adjust.
21061 * config/rs6000/603.md: Adjust.
21062 * config/rs6000/6xx.md: Adjust.
21063 * config/rs6000/7450.md: Adjust.
21064 * config/rs6000/7xx.md: Adjust.
21065 * config/rs6000/8540.md: Adjust.
21066 * config/rs6000/cell.md: Adjust.
21067 * config/rs6000/e300c2c3.md: Adjust.
21068 * config/rs6000/e500mc.md: Adjust.
21069 * config/rs6000/e500mc64.md: Adjust.
21070 * config/rs6000/e5500.md: Adjust.
21071 * config/rs6000/e6500.md: Adjust.
21072 * config/rs6000/mpc.md: Adjust.
21073 * config/rs6000/power4.md: Adjust.
21074 * config/rs6000/power5.md: Adjust.
21075 * config/rs6000/power6.md: Adjust.
21076 * config/rs6000/power7.md: Adjust.
21077 * config/rs6000/power8.md: Adjust.
21078 * config/rs6000/power9.md: Adjust.
21079 * config/rs6000/rs64.md: Adjust.
21080 * config/rs6000/titan.md: Adjust.
21082 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21084 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21085 ix86_indirect_branch_register logic.
21087 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21089 * config/i386/constraints.md (Bs): Update
21090 ix86_indirect_branch_register check. Don't check
21091 ix86_indirect_branch_register with GOT_memory_operand.
21093 * config/i386/predicates.md (GOT_memory_operand): Don't check
21094 ix86_indirect_branch_register here.
21095 (GOT32_symbol_operand): Likewise.
21097 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21099 * config/i386/predicates.md (constant_call_address_operand):
21100 Rewrite ix86_indirect_branch_register logic.
21101 (sibcall_insn_operand): Likewise.
21103 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21105 * config/i386/constraints.md (Bs): Replace
21106 ix86_indirect_branch_thunk_register with
21107 ix86_indirect_branch_register.
21109 * config/i386/i386.md (indirect_jump): Likewise.
21110 (tablejump): Likewise.
21111 (*sibcall_memory): Likewise.
21112 (*sibcall_value_memory): Likewise.
21113 Peepholes of indirect call and jump via memory: Likewise.
21114 * config/i386/i386.opt: Likewise.
21115 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21116 (GOT_memory_operand): Likewise.
21117 (call_insn_operand): Likewise.
21118 (sibcall_insn_operand): Likewise.
21119 (GOT32_symbol_operand): Likewise.
21121 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21123 PR middle-end/83837
21124 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21125 type rather than type addr's type points to.
21126 (expand_omp_atomic_mutex): Likewise.
21127 (expand_omp_atomic): Likewise.
21129 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21132 * config/i386/i386.c (output_indirect_thunk_function): Use
21133 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21134 for __x86_return_thunk.
21136 2018-01-15 Richard Biener <rguenther@suse.de>
21138 PR middle-end/83850
21139 * expmed.c (extract_bit_field_1): Fix typo.
21141 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21144 * config/arm/iterators.md (VF): New mode iterator.
21145 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21146 Remove integer-related logic from pattern.
21147 (neon_vabd<mode>_3): Likewise.
21149 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21151 PR middle-end/82694
21152 * common.opt (fstrict-overflow): No longer an alias.
21153 (fwrapv-pointer): New option.
21154 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21155 also for pointer types based on flag_wrapv_pointer.
21156 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21157 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21158 opts->x_flag_wrapv got set.
21159 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21160 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21161 POINTER_TYPE_OVERFLOW_UNDEFINED.
21162 * match.pd: Likewise in address comparison pattern.
21163 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21165 2018-01-15 Richard Biener <rguenther@suse.de>
21168 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21169 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21170 Reset type names to their identifier if their TYPE_DECL doesn't
21171 have linkage (and thus is used for ODR and devirt).
21172 (save_debug_info_for_decl): Remove.
21173 (save_debug_info_for_type): Likewise.
21174 (add_tree_to_fld_list): Adjust.
21175 * tree-pretty-print.c (dump_generic_node): Make dumping of
21176 type names more robust.
21178 2018-01-15 Richard Biener <rguenther@suse.de>
21180 * BASE-VER: Bump to 8.0.1.
21182 2018-01-14 Martin Sebor <msebor@redhat.com>
21185 * builtins.c (check_access): Avoid warning when the no-warning bit
21188 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21190 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21191 * ira-color (allocno_hard_regs_compare): Likewise.
21193 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21196 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21197 Use .pushsection/.popsection.
21199 2018-01-14 Martin Sebor <msebor@redhat.com>
21202 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21204 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21206 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21207 entry from extra_headers.
21208 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21209 extra_headers, make the list bitwise identical to the i?86-*-* one.
21211 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21213 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21214 -mcmodel=large with -mindirect-branch=thunk,
21215 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21216 -mfunction-return=thunk-extern.
21217 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21218 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21219 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21221 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21223 * config/i386/i386.c (print_reg): Print the name of the full
21224 integer register without '%'.
21225 (ix86_print_operand): Handle 'V'.
21226 * doc/extend.texi: Document 'V' modifier.
21228 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21230 * config/i386/constraints.md (Bs): Disallow memory operand for
21231 -mindirect-branch-register.
21233 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21234 (GOT_memory_operand): Likewise.
21235 (call_insn_operand): Likewise.
21236 (sibcall_insn_operand): Likewise.
21237 (GOT32_symbol_operand): Likewise.
21238 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21239 for -mindirect-branch-register.
21240 (tablejump): Likewise.
21241 (*sibcall_memory): Likewise.
21242 (*sibcall_value_memory): Likewise.
21243 Disallow peepholes of indirect call and jump via memory for
21244 -mindirect-branch-register.
21245 (*call_pop): Replace m with Bw.
21246 (*call_value_pop): Likewise.
21247 (*sibcall_pop_memory): Replace m with Bs.
21248 * config/i386/i386.opt (mindirect-branch-register): New option.
21249 * doc/invoke.texi: Document -mindirect-branch-register option.
21251 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21253 * config/i386/i386-protos.h (ix86_output_function_return): New.
21254 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21255 set function_return_type.
21256 (indirect_thunk_name): Add ret_p to indicate thunk for function
21258 (output_indirect_thunk_function): Pass false to
21259 indirect_thunk_name.
21260 (ix86_output_indirect_branch_via_reg): Likewise.
21261 (ix86_output_indirect_branch_via_push): Likewise.
21262 (output_indirect_thunk_function): Create alias for function
21263 return thunk if regno < 0.
21264 (ix86_output_function_return): New function.
21265 (ix86_handle_fndecl_attribute): Handle function_return.
21266 (ix86_attribute_table): Add function_return.
21267 * config/i386/i386.h (machine_function): Add
21268 function_return_type.
21269 * config/i386/i386.md (simple_return_internal): Use
21270 ix86_output_function_return.
21271 (simple_return_internal_long): Likewise.
21272 * config/i386/i386.opt (mfunction-return=): New option.
21273 (indirect_branch): Mention -mfunction-return=.
21274 * doc/extend.texi: Document function_return function attribute.
21275 * doc/invoke.texi: Document -mfunction-return= option.
21277 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21279 * config/i386/i386-opts.h (indirect_branch): New.
21280 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21281 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21282 with local indirect jump when converting indirect call and jump.
21283 (ix86_set_indirect_branch_type): New.
21284 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21285 (indirectlabelno): New.
21286 (indirect_thunk_needed): Likewise.
21287 (indirect_thunk_bnd_needed): Likewise.
21288 (indirect_thunks_used): Likewise.
21289 (indirect_thunks_bnd_used): Likewise.
21290 (INDIRECT_LABEL): Likewise.
21291 (indirect_thunk_name): Likewise.
21292 (output_indirect_thunk): Likewise.
21293 (output_indirect_thunk_function): Likewise.
21294 (ix86_output_indirect_branch_via_reg): Likewise.
21295 (ix86_output_indirect_branch_via_push): Likewise.
21296 (ix86_output_indirect_branch): Likewise.
21297 (ix86_output_indirect_jmp): Likewise.
21298 (ix86_code_end): Call output_indirect_thunk_function if needed.
21299 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21301 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21302 (ix86_attribute_table): Add indirect_branch.
21303 * config/i386/i386.h (machine_function): Add indirect_branch_type
21304 and has_local_indirect_jump.
21305 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21307 (tablejump): Likewise.
21308 (*indirect_jump): Use ix86_output_indirect_jmp.
21309 (*tablejump_1): Likewise.
21310 (simple_return_indirect_internal): Likewise.
21311 * config/i386/i386.opt (mindirect-branch=): New option.
21312 (indirect_branch): New.
21315 (thunk-inline): Likewise.
21316 (thunk-extern): Likewise.
21317 * doc/extend.texi: Document indirect_branch function attribute.
21318 * doc/invoke.texi: Document -mindirect-branch= option.
21320 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21323 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21325 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21327 * ipa-inline.c (want_inline_small_function_p): Return false if
21328 inlining has already failed with CIF_FINAL_ERROR.
21329 (update_caller_keys): Call want_inline_small_function_p before
21331 (update_callee_keys): Likewise.
21333 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21335 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21337 (rs6000_quadword_masked_address_p): Likewise.
21338 (quad_aligned_load_p): Likewise.
21339 (quad_aligned_store_p): Likewise.
21340 (const_load_sequence_p): Add comment to describe the outer-most loop.
21341 (mimic_memory_attributes_and_flags): New function.
21342 (rs6000_gen_stvx): Likewise.
21343 (replace_swapped_aligned_store): Likewise.
21344 (rs6000_gen_lvx): Likewise.
21345 (replace_swapped_aligned_load): Likewise.
21346 (replace_swapped_load_constant): Capitalize argument name in
21347 comment describing this function.
21348 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21349 and stores that access quad-word aligned addresses and replace
21350 with stvx or lvx instructions when appropriate.
21351 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21352 New function prototype.
21353 (rs6000_quadword_masked_address_p): Likewise.
21354 (rs6000_gen_lvx): Likewise.
21355 (rs6000_gen_stvx): Likewise.
21356 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21357 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21358 when memory address is aligned.
21359 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21360 this split to select lvx instruction when memory address is aligned.
21361 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21362 instruction when memory address is aligned.
21363 (*vsx_le_perm_load_v16qi): Likewise.
21364 (four unnamed splitters): Modify to select the stvx instruction
21365 when memory is aligned.
21367 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21369 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21370 which appears in the queue multiple times.
21372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21373 Alan Hayward <alan.hayward@arm.com>
21374 David Sherwood <david.sherwood@arm.com>
21376 * tree-vectorizer.h (vec_lower_bound): New structure.
21377 (_loop_vec_info): Add check_nonzero and lower_bounds.
21378 (LOOP_VINFO_CHECK_NONZERO): New macro.
21379 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21380 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21381 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21382 fields. Make seg_len the distance travelled, not including the
21384 (dr_direction_indicator): Declare.
21385 (dr_zero_step_indicator): Likewise.
21386 (dr_known_forward_stride_p): Likewise.
21387 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21389 (runtime_alias_check_p): Allow runtime alias checks with
21391 (operator ==): Compare access_size and align.
21392 (prune_runtime_alias_test_list): Rework for new distinction between
21393 the access_size and seg_len.
21394 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21396 (get_segment_min_max): New function.
21397 (create_intersect_range_checks): Use it.
21398 (dr_step_indicator): New function.
21399 (dr_direction_indicator): Likewise.
21400 (dr_zero_step_indicator): Likewise.
21401 (dr_known_forward_stride_p): Likewise.
21402 * tree-loop-distribution.c (data_ref_segment_size): Return
21403 DR_STEP * (niters - 1).
21404 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21406 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21407 (vect_preserves_scalar_order_p): New function, split out from...
21408 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21409 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21410 (vect_vfa_access_size): New function.
21411 (vect_vfa_align): Likewise.
21412 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21413 (dump_lower_bound): New function.
21414 (vect_check_lower_bound): Likewise.
21415 (vect_small_gap_p): Likewise.
21416 (vectorizable_with_step_bound_p): Likewise.
21417 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21418 depencies if the vectorization factor is 1. Convert the checks
21419 for nonzero steps into checks on the bounds of DR_STEP. Try using
21420 a bunds check for variable steps if the minimum required step is
21421 relatively small. Update calls to the dr_with_seg_len
21422 constructor and to vect_compile_time_alias.
21423 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21425 (vect_loop_versioning): Call it.
21426 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21428 (vect_estimate_min_profitable_iters): Account for any bounds checks.
21430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21431 Alan Hayward <alan.hayward@arm.com>
21432 David Sherwood <david.sherwood@arm.com>
21434 * doc/sourcebuild.texi (vect_scatter_store): Document.
21435 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
21437 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
21439 * genopinit.c (main): Add supports_vec_scatter_store and
21440 supports_vec_scatter_store_cached to target_optabs.
21441 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
21442 IFN_MASK_SCATTER_STORE.
21443 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
21445 * internal-fn.h (internal_store_fn_p): Declare.
21446 (internal_fn_stored_value_index): Likewise.
21447 * internal-fn.c (scatter_store_direct): New macro.
21448 (expand_scatter_store_optab_fn): New function.
21449 (direct_scatter_store_optab_supported_p): New macro.
21450 (internal_store_fn_p): New function.
21451 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21452 IFN_MASK_SCATTER_STORE.
21453 (internal_fn_mask_index): Likewise.
21454 (internal_fn_stored_value_index): New function.
21455 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
21456 for scatter stores.
21457 * optabs-query.h (supports_vec_scatter_store_p): Declare.
21458 * optabs-query.c (supports_vec_scatter_store_p): New function.
21459 * tree-vectorizer.h (vect_get_store_rhs): Declare.
21460 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
21461 true for scatter stores.
21462 (vect_gather_scatter_fn_p): Handle scatter stores too.
21463 (vect_check_gather_scatter): Consider using scatter stores if
21464 supports_vec_scatter_store_p.
21465 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
21466 scatter stores too.
21467 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21468 internal_fn_stored_value_index.
21469 (check_load_store_masking): Handle scatter stores too.
21470 (vect_get_store_rhs): Make public.
21471 (vectorizable_call): Use internal_store_fn_p.
21472 (vectorizable_store): Handle scatter store internal functions.
21473 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
21474 when deciding whether the end of the group has been reached.
21475 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
21476 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
21477 (mask_scatter_store<mode>): New insns.
21479 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21480 Alan Hayward <alan.hayward@arm.com>
21481 David Sherwood <david.sherwood@arm.com>
21483 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
21484 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
21485 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
21487 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
21488 Use vect_truncate_gather_scatter_offset if we can't treat the
21489 operation as a normal gather load or scatter store.
21490 (get_group_load_store_type): Take the gather_scatter_info
21491 as argument. Try using a gather load or scatter store for
21492 single-element groups.
21493 (get_load_store_type): Update calls to get_group_load_store_type
21494 and vect_use_strided_gather_scatters_p.
21496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21497 Alan Hayward <alan.hayward@arm.com>
21498 David Sherwood <david.sherwood@arm.com>
21500 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
21501 optional tree argument.
21502 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
21504 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
21505 but continue to use the current value as a fallback.
21506 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
21507 to compare the updates.
21508 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
21509 (get_load_store_type): Use it when handling a strided access.
21510 (vect_get_strided_load_store_ops): New function.
21511 (vect_get_data_ptr_increment): Likewise.
21512 (vectorizable_load): Handle strided gather loads. Always pass
21513 a step to vect_create_data_ref_ptr and bump_vector_ptr.
21515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21516 Alan Hayward <alan.hayward@arm.com>
21517 David Sherwood <david.sherwood@arm.com>
21519 * doc/md.texi (gather_load@var{m}): Document.
21520 (mask_gather_load@var{m}): Likewise.
21521 * genopinit.c (main): Add supports_vec_gather_load and
21522 supports_vec_gather_load_cached to target_optabs.
21523 * optabs-tree.c (init_tree_optimization_optabs): Use
21524 ggc_cleared_alloc to allocate target_optabs.
21525 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
21526 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
21528 * internal-fn.h (internal_load_fn_p): Declare.
21529 (internal_gather_scatter_fn_p): Likewise.
21530 (internal_fn_mask_index): Likewise.
21531 (internal_gather_scatter_fn_supported_p): Likewise.
21532 * internal-fn.c (gather_load_direct): New macro.
21533 (expand_gather_load_optab_fn): New function.
21534 (direct_gather_load_optab_supported_p): New macro.
21535 (direct_internal_fn_optab): New function.
21536 (internal_load_fn_p): Likewise.
21537 (internal_gather_scatter_fn_p): Likewise.
21538 (internal_fn_mask_index): Likewise.
21539 (internal_gather_scatter_fn_supported_p): Likewise.
21540 * optabs-query.c (supports_at_least_one_mode_p): New function.
21541 (supports_vec_gather_load_p): Likewise.
21542 * optabs-query.h (supports_vec_gather_load_p): Declare.
21543 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
21544 and memory_type field.
21545 (NUM_PATTERNS): Bump to 15.
21546 * tree-vect-data-refs.c: Include internal-fn.h.
21547 (vect_gather_scatter_fn_p): New function.
21548 (vect_describe_gather_scatter_call): Likewise.
21549 (vect_check_gather_scatter): Try using internal functions for
21550 gather loads. Recognize existing calls to a gather load function.
21551 (vect_analyze_data_refs): Consider using gather loads if
21552 supports_vec_gather_load_p.
21553 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
21554 (vect_get_gather_scatter_offset_type): Likewise.
21555 (vect_convert_mask_for_vectype): Likewise.
21556 (vect_add_conversion_to_patterm): Likewise.
21557 (vect_try_gather_scatter_pattern): Likewise.
21558 (vect_recog_gather_scatter_pattern): New pattern recognizer.
21559 (vect_vect_recog_func_ptrs): Add it.
21560 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21561 internal_fn_mask_index and internal_gather_scatter_fn_p.
21562 (check_load_store_masking): Take the gather_scatter_info as an
21563 argument and handle gather loads.
21564 (vect_get_gather_scatter_ops): New function.
21565 (vectorizable_call): Check internal_load_fn_p.
21566 (vectorizable_load): Likewise. Handle gather load internal
21568 (vectorizable_store): Update call to check_load_store_masking.
21569 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
21570 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
21571 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
21572 (aarch64_gather_scale_operand_d): New predicates.
21573 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
21574 (mask_gather_load<mode>): New insns.
21576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21577 Alan Hayward <alan.hayward@arm.com>
21578 David Sherwood <david.sherwood@arm.com>
21580 * optabs.def (fold_left_plus_optab): New optab.
21581 * doc/md.texi (fold_left_plus_@var{m}): Document.
21582 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
21583 * internal-fn.c (fold_left_direct): Define.
21584 (expand_fold_left_optab_fn): Likewise.
21585 (direct_fold_left_optab_supported_p): Likewise.
21586 * fold-const-call.c (fold_const_fold_left): New function.
21587 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
21588 * tree-parloops.c (valid_reduction_p): New function.
21589 (gather_scalar_reductions): Use it.
21590 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
21591 (vect_finish_replace_stmt): Declare.
21592 * tree-vect-loop.c (fold_left_reduction_fn): New function.
21593 (needs_fold_left_reduction_p): New function, split out from...
21594 (vect_is_simple_reduction): ...here. Accept reductions that
21595 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
21596 (vect_force_simple_reduction): Also store the reduction type in
21597 the assignment's STMT_VINFO_REDUC_TYPE.
21598 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
21599 (merge_with_identity): New function.
21600 (vect_expand_fold_left): Likewise.
21601 (vectorize_fold_left_reduction): Likewise.
21602 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
21603 scalar phi in place for it. Check for target support and reject
21604 cases that would reassociate the operation. Defer the transform
21605 phase to vectorize_fold_left_reduction.
21606 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21607 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21608 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21612 * tree-if-conv.c (predicate_mem_writes): Remove redundant
21613 call to ifc_temp_var.
21615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21616 Alan Hayward <alan.hayward@arm.com>
21617 David Sherwood <david.sherwood@arm.com>
21619 * target.def (legitimize_address_displacement): Take the original
21620 offset as a poly_int.
21621 * targhooks.h (default_legitimize_address_displacement): Update
21623 * targhooks.c (default_legitimize_address_displacement): Likewise.
21624 * doc/tm.texi: Regenerate.
21625 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21626 as an argument, moving assert of ad->disp == ad->disp_term to...
21627 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
21628 Try calling targetm.legitimize_address_displacement before expanding
21629 the address rather than afterwards, and adjust for the new interface.
21630 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21631 Match the new hook interface. Handle SVE addresses.
21632 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21633 new hook interface.
21635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21637 * Makefile.in (OBJS): Add early-remat.o.
21638 * target.def (select_early_remat_modes): New hook.
21639 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21640 * doc/tm.texi: Regenerate.
21641 * targhooks.h (default_select_early_remat_modes): Declare.
21642 * targhooks.c (default_select_early_remat_modes): New function.
21643 * timevar.def (TV_EARLY_REMAT): New timevar.
21644 * passes.def (pass_early_remat): New pass.
21645 * tree-pass.h (make_pass_early_remat): Declare.
21646 * early-remat.c: New file.
21647 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21649 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21652 Alan Hayward <alan.hayward@arm.com>
21653 David Sherwood <david.sherwood@arm.com>
21655 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21656 vfm1 with a bound_epilog parameter.
21657 (vect_do_peeling): Update calls accordingly, and move the prologue
21658 call earlier in the function. Treat the base bound_epilog as 0 for
21659 fully-masked loops and retain vf - 1 for other loops. Add 1 to
21660 this base when peeling for gaps.
21661 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21662 with fully-masked loops.
21663 (vect_estimate_min_profitable_iters): Handle the single peeled
21664 iteration in that case.
21666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21667 Alan Hayward <alan.hayward@arm.com>
21668 David Sherwood <david.sherwood@arm.com>
21670 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21671 single-element interleaving even if the size is not a power of 2.
21672 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21673 accesses for single-element interleaving if the group size is
21676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21677 Alan Hayward <alan.hayward@arm.com>
21678 David Sherwood <david.sherwood@arm.com>
21680 * doc/md.texi (fold_extract_last_@var{m}): Document.
21681 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21682 * optabs.def (fold_extract_last_optab): New optab.
21683 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21684 * internal-fn.c (fold_extract_direct): New macro.
21685 (expand_fold_extract_optab_fn): Likewise.
21686 (direct_fold_extract_optab_supported_p): Likewise.
21687 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21688 * tree-vect-loop.c (vect_model_reduction_cost): Handle
21689 EXTRACT_LAST_REDUCTION.
21690 (get_initial_def_for_reduction): Do not create an initial vector
21691 for EXTRACT_LAST_REDUCTION reductions.
21692 (vectorizable_reduction): Leave the scalar phi in place for
21693 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
21694 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
21695 epilogue code for EXTRACT_LAST_REDUCTION and defer the
21696 transform phase to vectorizable_condition.
21697 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21699 (vect_finish_stmt_generation): ...here.
21700 (vect_finish_replace_stmt): New function.
21701 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21702 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21704 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21707 Alan Hayward <alan.hayward@arm.com>
21708 David Sherwood <david.sherwood@arm.com>
21710 * doc/md.texi (extract_last_@var{m}): Document.
21711 * optabs.def (extract_last_optab): New optab.
21712 * internal-fn.def (EXTRACT_LAST): New internal function.
21713 * internal-fn.c (cond_unary_direct): New macro.
21714 (expand_cond_unary_optab_fn): Likewise.
21715 (direct_cond_unary_optab_supported_p): Likewise.
21716 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21717 loops using EXTRACT_LAST.
21718 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21719 (extract_last_<mode>): ...this optab.
21720 (vec_extract<mode><Vel>): Update accordingly.
21722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21723 Alan Hayward <alan.hayward@arm.com>
21724 David Sherwood <david.sherwood@arm.com>
21726 * target.def (empty_mask_is_expensive): New hook.
21727 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21728 * doc/tm.texi: Regenerate.
21729 * targhooks.h (default_empty_mask_is_expensive): Declare.
21730 * targhooks.c (default_empty_mask_is_expensive): New function.
21731 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
21732 if the target says that empty masks are expensive.
21733 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
21735 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
21737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21738 Alan Hayward <alan.hayward@arm.com>
21739 David Sherwood <david.sherwood@arm.com>
21741 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
21742 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
21743 (vect_use_loop_mask_for_alignment_p): New function.
21744 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
21745 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
21746 niters_skip argument. Make sure that the first niters_skip elements
21747 of the first iteration are inactive.
21748 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
21749 Update call to vect_set_loop_masks_directly.
21750 (get_misalign_in_elems): New function, split out from...
21751 (vect_gen_prolog_loop_niters): ...here.
21752 (vect_update_init_of_dr): Take a code argument that specifies whether
21753 the adjustment should be added or subtracted.
21754 (vect_update_init_of_drs): Likewise.
21755 (vect_prepare_for_masked_peels): New function.
21756 (vect_do_peeling): Skip prologue peeling if we're using a mask
21757 instead. Update call to vect_update_inits_of_drs.
21758 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21760 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
21761 alignment. Do not include the number of peeled iterations in
21762 the minimum threshold in that case.
21763 (vectorizable_induction): Adjust the start value down by
21764 LOOP_VINFO_MASK_SKIP_NITERS iterations.
21765 (vect_transform_loop): Call vect_prepare_for_masked_peels.
21766 Take the number of skipped iterations into account when calculating
21768 * tree-vect-stmts.c (vect_gen_while_not): New function.
21770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21771 Alan Hayward <alan.hayward@arm.com>
21772 David Sherwood <david.sherwood@arm.com>
21774 * doc/sourcebuild.texi (vect_fully_masked): Document.
21775 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
21776 default value to 0.
21777 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
21779 (vect_analyze_loop_2): ...here. Don't check the vectorization
21780 factor against the number of loop iterations if the loop is
21783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21784 Alan Hayward <alan.hayward@arm.com>
21785 David Sherwood <david.sherwood@arm.com>
21787 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
21788 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
21789 (dump_groups): Update accordingly.
21790 (iv_use::mem_type): New member variable.
21791 (address_p): New function.
21792 (record_use): Add a mem_type argument and initialize the new
21794 (record_group_use): Add a mem_type argument. Use address_p.
21795 Remove obsolete null checks of base_object. Update call to record_use.
21796 (find_interesting_uses_op): Update call to record_group_use.
21797 (find_interesting_uses_cond): Likewise.
21798 (find_interesting_uses_address): Likewise.
21799 (get_mem_type_for_internal_fn): New function.
21800 (find_address_like_use): Likewise.
21801 (find_interesting_uses_stmt): Try find_address_like_use before
21802 calling find_interesting_uses_op.
21803 (addr_offset_valid_p): Use the iv mem_type field as the type
21804 of the addressed memory.
21805 (add_autoinc_candidates): Likewise.
21806 (get_address_cost): Likewise.
21807 (split_small_address_groups_p): Use address_p.
21808 (split_address_groups): Likewise.
21809 (add_iv_candidate_for_use): Likewise.
21810 (autoinc_possible_for_pair): Likewise.
21811 (rewrite_groups): Likewise.
21812 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
21813 (determine_group_iv_cost): Update after split of USE_ADDRESS.
21814 (get_alias_ptr_type_for_ptr_address): New function.
21815 (rewrite_use_address): Rewrite address uses in calls that were
21816 identified by find_address_like_use.
21818 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21819 Alan Hayward <alan.hayward@arm.com>
21820 David Sherwood <david.sherwood@arm.com>
21822 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
21824 * gimple-expr.h (is_gimple_addressable: Likewise.
21825 * gimple-expr.c (is_gimple_address): Likewise.
21826 * internal-fn.c (expand_call_mem_ref): New function.
21827 (expand_mask_load_optab_fn): Use it.
21828 (expand_mask_store_optab_fn): Likewise.
21830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21831 Alan Hayward <alan.hayward@arm.com>
21832 David Sherwood <david.sherwood@arm.com>
21834 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
21835 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
21836 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
21837 (cond_umax@var{mode}): Document.
21838 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
21839 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
21840 (cond_umin_optab, cond_umax_optab): New optabs.
21841 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
21842 (COND_IOR, COND_XOR): New internal functions.
21843 * internal-fn.h (get_conditional_internal_fn): Declare.
21844 * internal-fn.c (cond_binary_direct): New macro.
21845 (expand_cond_binary_optab_fn): Likewise.
21846 (direct_cond_binary_optab_supported_p): Likewise.
21847 (get_conditional_internal_fn): New function.
21848 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
21849 Cope with reduction statements that are vectorized as calls rather
21851 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
21852 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
21853 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
21854 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21855 (UNSPEC_COND_EOR): New unspecs.
21856 (optab): Add mappings for them.
21857 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
21858 (sve_int_op, sve_fp_op): New int attributes.
21860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21861 Alan Hayward <alan.hayward@arm.com>
21862 David Sherwood <david.sherwood@arm.com>
21864 * optabs.def (while_ult_optab): New optab.
21865 * doc/md.texi (while_ult@var{m}@var{n}): Document.
21866 * internal-fn.def (WHILE_ULT): New internal function.
21867 * internal-fn.h (direct_internal_fn_supported_p): New override
21868 that takes two types as argument.
21869 * internal-fn.c (while_direct): New macro.
21870 (expand_while_optab_fn): New function.
21871 (convert_optab_supported_p): Likewise.
21872 (direct_while_optab_supported_p): New macro.
21873 * wide-int.h (wi::udiv_ceil): New function.
21874 * tree-vectorizer.h (rgroup_masks): New structure.
21875 (vec_loop_masks): New typedef.
21876 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
21877 and fully_masked_p.
21878 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
21879 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
21880 (vect_max_vf): New function.
21881 (slpeel_make_loop_iterate_ntimes): Delete.
21882 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
21883 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
21884 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
21885 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
21886 internal-fn.h, stor-layout.h and optabs-query.h.
21887 (vect_set_loop_mask): New function.
21888 (add_preheader_seq): Likewise.
21889 (add_header_seq): Likewise.
21890 (interleave_supported_p): Likewise.
21891 (vect_maybe_permute_loop_masks): Likewise.
21892 (vect_set_loop_masks_directly): Likewise.
21893 (vect_set_loop_condition_masked): Likewise.
21894 (vect_set_loop_condition_unmasked): New function, split out from
21895 slpeel_make_loop_iterate_ntimes.
21896 (slpeel_make_loop_iterate_ntimes): Rename to..
21897 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
21898 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
21899 (vect_do_peeling): Update call accordingly.
21900 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
21902 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21903 mask_compare_type, can_fully_mask_p and fully_masked_p.
21904 (release_vec_loop_masks): New function.
21905 (_loop_vec_info): Use it to free the loop masks.
21906 (can_produce_all_loop_masks_p): New function.
21907 (vect_get_max_nscalars_per_iter): Likewise.
21908 (vect_verify_full_masking): Likewise.
21909 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
21910 retries, and free the mask rgroups before retrying. Check loop-wide
21911 reasons for disallowing fully-masked loops. Make the final decision
21912 about whether use a fully-masked loop or not.
21913 (vect_estimate_min_profitable_iters): Do not assume that peeling
21914 for the number of iterations will be needed for fully-masked loops.
21915 (vectorizable_reduction): Disable fully-masked loops.
21916 (vectorizable_live_operation): Likewise.
21917 (vect_halve_mask_nunits): New function.
21918 (vect_double_mask_nunits): Likewise.
21919 (vect_record_loop_mask): Likewise.
21920 (vect_get_loop_mask): Likewise.
21921 (vect_transform_loop): Handle the case in which the final loop
21922 iteration might handle a partial vector. Call vect_set_loop_condition
21923 instead of slpeel_make_loop_iterate_ntimes.
21924 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
21925 (check_load_store_masking): New function.
21926 (prepare_load_store_mask): Likewise.
21927 (vectorizable_store): Handle fully-masked loops.
21928 (vectorizable_load): Likewise.
21929 (supportable_widening_operation): Use vect_halve_mask_nunits for
21931 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
21932 (vect_gen_while): New function.
21933 * config/aarch64/aarch64.md (umax<mode>3): New expander.
21934 (aarch64_uqdec<mode>): New insn.
21936 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21937 Alan Hayward <alan.hayward@arm.com>
21938 David Sherwood <david.sherwood@arm.com>
21940 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
21941 (reduc_xor_scal_optab): New optabs.
21942 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
21943 (reduc_xor_scal_@var{m}): Document.
21944 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
21945 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
21946 internal functions.
21947 * fold-const-call.c (fold_const_call): Handle them.
21948 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
21949 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
21950 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
21951 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
21952 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
21953 (UNSPEC_XORV): New unspecs.
21954 (optab): Add entries for them.
21955 (BITWISEV): New int iterator.
21956 (bit_reduc_op): New int attributes.
21958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21959 Alan Hayward <alan.hayward@arm.com>
21960 David Sherwood <david.sherwood@arm.com>
21962 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
21963 * internal-fn.def (VEC_SHL_INSERT): New internal function.
21964 * optabs.def (vec_shl_insert_optab): New optab.
21965 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
21966 (duplicate_and_interleave): Likewise.
21967 * tree-vect-loop.c: Include internal-fn.h.
21968 (neutral_op_for_slp_reduction): New function, split out from
21969 get_initial_defs_for_reduction.
21970 (get_initial_def_for_reduction): Handle option 2 for variable-length
21971 vectors by loading the neutral value into a vector and then shifting
21972 the initial value into element 0.
21973 (get_initial_defs_for_reduction): Replace the code argument with
21974 the neutral value calculated by neutral_op_for_slp_reduction.
21975 Use gimple_build_vector for constant-length vectors.
21976 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
21977 but the first group_size elements have a neutral value.
21978 Use duplicate_and_interleave otherwise.
21979 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
21980 Update call to get_initial_defs_for_reduction. Handle SLP
21981 reductions for variable-length vectors by creating one vector
21982 result for each scalar result, with the elements associated
21983 with other scalar results stubbed out with the neutral value.
21984 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
21985 Require IFN_VEC_SHL_INSERT for double reductions on
21986 variable-length vectors, or SLP reductions that have
21987 a neutral value. Require can_duplicate_and_interleave_p
21988 support for variable-length unchained SLP reductions if there
21989 is no neutral value, such as for MIN/MAX reductions. Also require
21990 the number of vector elements to be a multiple of the number of
21991 SLP statements when doing variable-length unchained SLP reductions.
21992 Update call to vect_create_epilog_for_reduction.
21993 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
21994 and remove initial values.
21995 (duplicate_and_interleave): Make public.
21996 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
21997 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
21999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22000 Alan Hayward <alan.hayward@arm.com>
22001 David Sherwood <david.sherwood@arm.com>
22003 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22004 (can_duplicate_and_interleave_p): New function.
22005 (vect_get_and_check_slp_defs): Take the vector of statements
22006 rather than just the current one. Remove excess parentheses.
22007 Restriction rejectinon of vect_constant_def and vect_external_def
22008 for variable-length vectors to boolean types, or types for which
22009 can_duplicate_and_interleave_p is false.
22010 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22011 (duplicate_and_interleave): New function.
22012 (vect_get_constant_vectors): Use gimple_build_vector for
22013 constant-length vectors and suitable variable-length constant
22014 vectors. Use duplicate_and_interleave for other variable-length
22015 vectors. Don't defer the update when inserting new statements.
22017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22018 Alan Hayward <alan.hayward@arm.com>
22019 David Sherwood <david.sherwood@arm.com>
22021 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22022 min_profitable_iters doesn't go negative.
22024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22025 Alan Hayward <alan.hayward@arm.com>
22026 David Sherwood <david.sherwood@arm.com>
22028 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22029 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22030 * optabs.def (vec_mask_load_lanes_optab): New optab.
22031 (vec_mask_store_lanes_optab): Likewise.
22032 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22033 (MASK_STORE_LANES): Likewise.
22034 * internal-fn.c (mask_load_lanes_direct): New macro.
22035 (mask_store_lanes_direct): Likewise.
22036 (expand_mask_load_optab_fn): Handle masked operations.
22037 (expand_mask_load_lanes_optab_fn): New macro.
22038 (expand_mask_store_optab_fn): Handle masked operations.
22039 (expand_mask_store_lanes_optab_fn): New macro.
22040 (direct_mask_load_lanes_optab_supported_p): Likewise.
22041 (direct_mask_store_lanes_optab_supported_p): Likewise.
22042 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22044 (vect_load_lanes_supported): Likewise.
22045 * tree-vect-data-refs.c (strip_conversion): New function.
22046 (can_group_stmts_p): Likewise.
22047 (vect_analyze_data_ref_accesses): Use it instead of checking
22048 for a pair of assignments.
22049 (vect_store_lanes_supported): Take a masked_p parameter.
22050 (vect_load_lanes_supported): Likewise.
22051 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22052 vect_store_lanes_supported and vect_load_lanes_supported.
22053 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22054 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22055 parameter. Don't allow gaps for masked accesses.
22056 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22057 and vect_load_lanes_supported.
22058 (get_load_store_type): Take a masked_p parameter and update
22059 call to get_group_load_store_type.
22060 (vectorizable_store): Update call to get_load_store_type.
22061 Handle IFN_MASK_STORE_LANES.
22062 (vectorizable_load): Update call to get_load_store_type.
22063 Handle IFN_MASK_LOAD_LANES.
22065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22066 Alan Hayward <alan.hayward@arm.com>
22067 David Sherwood <david.sherwood@arm.com>
22069 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22071 * config/aarch64/aarch64-protos.h
22072 (aarch64_sve_struct_memory_operand_p): Declare.
22073 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22074 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22075 (VPRED, vpred): Handle SVE structure modes.
22076 * config/aarch64/constraints.md (Utx): New constraint.
22077 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22078 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22079 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22080 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22081 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22082 structure modes. Split into pieces after RA.
22083 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22084 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22086 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22087 SVE structure modes.
22088 (aarch64_classify_address): Likewise.
22089 (sizetochar): Move earlier in file.
22090 (aarch64_print_operand): Handle SVE register lists.
22091 (aarch64_array_mode): New function.
22092 (aarch64_sve_struct_memory_operand_p): Likewise.
22093 (TARGET_ARRAY_MODE): Redefine.
22095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22096 Alan Hayward <alan.hayward@arm.com>
22097 David Sherwood <david.sherwood@arm.com>
22099 * target.def (array_mode): New target hook.
22100 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22101 * doc/tm.texi: Regenerate.
22102 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22103 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22104 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22105 targetm.array_mode.
22106 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22110 Alan Hayward <alan.hayward@arm.com>
22111 David Sherwood <david.sherwood@arm.com>
22113 * fold-const.c (fold_binary_loc): Check the argument types
22114 rather than the result type when testing for a vector operation.
22116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22118 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22119 * doc/tm.texi: Regenerate.
22121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22122 Alan Hayward <alan.hayward@arm.com>
22123 David Sherwood <david.sherwood@arm.com>
22125 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22126 (sve): Document new AArch64 extension.
22127 * doc/md.texi (w): Extend the description of the AArch64
22128 constraint to include SVE vectors.
22129 (Upl, Upa): Document new AArch64 predicate constraints.
22130 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22132 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22133 (msve-vector-bits=): New option.
22134 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22135 SVE when these are disabled.
22136 (sve): New extension.
22137 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22138 modes. Adjust their number of units based on aarch64_sve_vg.
22139 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22140 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22141 aarch64_addr_query_type.
22142 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22143 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22144 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22145 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22146 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22147 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22148 (aarch64_simd_imm_zero_p): Delete.
22149 (aarch64_check_zero_based_sve_index_immediate): Declare.
22150 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22151 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22152 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22153 (aarch64_sve_float_mul_immediate_p): Likewise.
22154 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22155 rather than an rtx.
22156 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22157 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22158 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22159 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22160 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22161 (aarch64_regmode_natural_size): Likewise.
22162 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22163 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22165 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22166 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22167 for VG and the SVE predicate registers.
22168 (V_ALIASES): Add a "z"-prefixed alias.
22169 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22170 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22171 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22172 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22173 (REG_CLASS_NAMES): Add entries for them.
22174 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22175 and the predicate registers.
22176 (aarch64_sve_vg): Declare.
22177 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22178 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22179 (REGMODE_NATURAL_SIZE): Define.
22180 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22182 * config/aarch64/aarch64.c: Include cfgrtl.h.
22183 (simd_immediate_info): Add a constructor for series vectors,
22184 and an associated step field.
22185 (aarch64_sve_vg): New variable.
22186 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22187 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22188 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22189 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22190 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22191 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22192 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22193 (aarch64_get_mask_mode): New functions.
22194 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22195 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22196 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22197 predicate modes and predicate registers. Explicitly restrict
22198 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22199 to store a vector mode if it is recognized by
22200 aarch64_classify_vector_mode.
22201 (aarch64_regmode_natural_size): New function.
22202 (aarch64_hard_regno_caller_save_mode): Return the original mode
22204 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22205 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22206 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22207 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22209 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22210 does not overlap dest if the function is frame-related. Handle
22212 (aarch64_split_add_offset): New function.
22213 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22214 them aarch64_add_offset.
22215 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22216 and update call to aarch64_sub_sp.
22217 (aarch64_add_cfa_expression): New function.
22218 (aarch64_expand_prologue): Pass extra temporary registers to the
22219 functions above. Handle the case in which we need to emit new
22220 DW_CFA_expressions for registers that were originally saved
22221 relative to the stack pointer, but now have to be expressed
22222 relative to the frame pointer.
22223 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22225 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22226 IP0 and IP1 values for SVE frames.
22227 (aarch64_expand_vec_series): New function.
22228 (aarch64_expand_sve_widened_duplicate): Likewise.
22229 (aarch64_expand_sve_const_vector): Likewise.
22230 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22231 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22232 into the register, rather than emitting a SET directly.
22233 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22234 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22235 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22236 (offset_9bit_signed_scaled_p): New functions.
22237 (aarch64_replicate_bitmask_imm): New function.
22238 (aarch64_bitmask_imm): Use it.
22239 (aarch64_cannot_force_const_mem): Reject expressions involving
22240 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22241 (aarch64_classify_index): Handle SVE indices, by requiring
22242 a plain register index with a scale that matches the element size.
22243 (aarch64_classify_address): Handle SVE addresses. Assert that
22244 the mode of the address is VOIDmode or an integer mode.
22245 Update call to aarch64_classify_symbol.
22246 (aarch64_classify_symbolic_expression): Update call to
22247 aarch64_classify_symbol.
22248 (aarch64_const_vec_all_in_range_p): New function.
22249 (aarch64_print_vector_float_operand): Likewise.
22250 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22251 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22252 and the FP immediates 1.0 and 0.5.
22253 (aarch64_print_address_internal): Handle SVE addresses.
22254 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22255 (aarch64_regno_regclass): Handle predicate registers.
22256 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22258 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22259 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22260 (aarch64_convert_sve_vector_bits): New function.
22261 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22262 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22263 rather than an rtx.
22264 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22265 Handle SVE vector and predicate modes. Accept VL-based constants
22266 that need only one temporary register, and VL offsets that require
22267 no temporary registers.
22268 (aarch64_conditional_register_usage): Mark the predicate registers
22269 as fixed if SVE isn't available.
22270 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22271 Return true for SVE vector and predicate modes.
22272 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22273 rather than an unsigned int. Handle SVE modes.
22274 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22276 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22278 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22279 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22280 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22281 (aarch64_sve_float_mul_immediate_p): New functions.
22282 (aarch64_sve_valid_immediate): New function.
22283 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22284 Explicitly reject structure modes. Check for INDEX constants.
22285 Handle PTRUE and PFALSE constants.
22286 (aarch64_check_zero_based_sve_index_immediate): New function.
22287 (aarch64_simd_imm_zero_p): Delete.
22288 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22289 vector modes. Accept constants in the range of CNT[BHWD].
22290 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22291 ask for an Advanced SIMD mode.
22292 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22293 (aarch64_simd_vector_alignment): Handle SVE predicates.
22294 (aarch64_vectorize_preferred_vector_alignment): New function.
22295 (aarch64_simd_vector_alignment_reachable): Use it instead of
22297 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22298 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22300 (MAX_VECT_LEN): Delete.
22301 (expand_vec_perm_d): Add a vec_flags field.
22302 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22303 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22304 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22306 (aarch64_evpc_rev): Rename to...
22307 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22308 (aarch64_evpc_rev_global): New function.
22309 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22310 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22312 (aarch64_evpc_sve_tbl): New function.
22313 (aarch64_expand_vec_perm_const_1): Update after rename of
22314 aarch64_evpc_rev. Handle SVE permutes too, trying
22315 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22316 than aarch64_evpc_tbl.
22317 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22318 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22319 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22320 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22321 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22322 (aarch64_expand_sve_vcond): New functions.
22323 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22324 of aarch64_vector_mode_p.
22325 (aarch64_dwarf_poly_indeterminate_value): New function.
22326 (aarch64_compute_pressure_classes): Likewise.
22327 (aarch64_can_change_mode_class): Likewise.
22328 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22329 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22330 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22331 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22332 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22333 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22334 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22335 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22337 (Dn, Dl, Dr): Accept const as well as const_vector.
22338 (Dz): Likewise. Compare against CONST0_RTX.
22339 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22340 of "vector" where appropriate.
22341 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22342 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22343 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22344 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22345 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22346 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22347 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22348 (v_int_equiv): Extend to SVE modes.
22349 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22351 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22352 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22353 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22354 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22355 (SVE_COND_FP_CMP): New int iterators.
22356 (perm_hilo): Handle the new unpack unspecs.
22357 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22359 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22360 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22361 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22362 (aarch64_equality_operator, aarch64_constant_vector_operand)
22363 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22364 (aarch64_sve_nonimmediate_operand): Likewise.
22365 (aarch64_sve_general_operand): Likewise.
22366 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22367 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22368 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22369 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22370 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22371 (aarch64_sve_float_arith_immediate): Likewise.
22372 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22373 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22374 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22375 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22376 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22377 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22378 (aarch64_sve_float_arith_operand): Likewise.
22379 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22380 (aarch64_sve_float_mul_operand): Likewise.
22381 (aarch64_sve_vec_perm_operand): Likewise.
22382 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22383 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22384 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22385 as well as const_vector.
22386 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22387 in file. Use CONST0_RTX and CONSTM1_RTX.
22388 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22389 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22390 Use aarch64_simd_imm_zero.
22391 * config/aarch64/aarch64-sve.md: New file.
22392 * config/aarch64/aarch64.md: Include it.
22393 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22394 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22395 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22396 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22397 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22398 (sve): New attribute.
22399 (enabled): Disable instructions with the sve attribute unless
22401 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22402 aarch64_expand_mov_immediate.
22403 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22404 CNT[BHSD] immediates.
22405 (movti): Split CONST_POLY_INT moves into two halves.
22406 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22407 Split additions that need a temporary here if the destination
22408 is the stack pointer.
22409 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22410 (*add<mode>3_poly_1): New instruction.
22411 (set_clobber_cc): New expander.
22413 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22415 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22416 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22417 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22418 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22419 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
22420 Change innermode from fixed_mode_size to machine_mode.
22421 (simplify_subreg): Update call accordingly. Handle a constant-sized
22422 subreg of a variable-length CONST_VECTOR.
22424 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22425 Alan Hayward <alan.hayward@arm.com>
22426 David Sherwood <david.sherwood@arm.com>
22428 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
22429 (add_offset_to_base): New function, split out from...
22430 (create_mem_ref): ...here. When handling a scale other than 1,
22431 check first whether the address is valid without the offset.
22432 Add it into the base if so, leaving the index and scale as-is.
22434 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22437 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
22438 fold_for_warn before checking if arg2 is INTEGER_CST.
22440 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
22442 * config/rs6000/predicates.md (load_multiple_operation): Delete.
22443 (store_multiple_operation): Delete.
22444 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
22445 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
22446 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
22447 guarded by TARGET_STRING.
22448 (rs6000_output_load_multiple): Delete.
22449 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
22450 OPTION_MASK_STRING / TARGET_STRING handling.
22451 (print_operand) <'N', 'O'>: Add comment that these are unused now.
22452 (const rs6000_opt_masks) <"string">: Change mask to 0.
22453 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22454 (MASK_STRING): Delete.
22455 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
22457 (load_multiple): Delete.
22464 (store_multiple): Delete.
22471 (movmemsi_8reg): Delete.
22472 (corresponding unnamed define_insn): Delete.
22473 (movmemsi_6reg): Delete.
22474 (corresponding unnamed define_insn): Delete.
22475 (movmemsi_4reg): Delete.
22476 (corresponding unnamed define_insn): Delete.
22477 (movmemsi_2reg): Delete.
22478 (corresponding unnamed define_insn): Delete.
22479 (movmemsi_1reg): Delete.
22480 (corresponding unnamed define_insn): Delete.
22481 * config/rs6000/rs6000.opt (mno-string): New.
22482 (mstring): Replace by deprecation warning stub.
22483 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
22485 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22487 * regrename.c (regrename_do_replace): If replacing the same
22488 reg multiple times, try to reuse last created gen_raw_REG.
22491 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
22492 main to workaround a bug in GDB.
22494 2018-01-12 Tom de Vries <tom@codesourcery.com>
22497 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
22499 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
22501 PR rtl-optimization/80481
22502 * ira-color.c (get_cap_member): New function.
22503 (allocnos_conflict_by_live_ranges_p): Use it.
22504 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
22505 (setup_slot_coalesced_allocno_live_ranges): Ditto.
22507 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
22510 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
22511 (*saddl_se_1): Ditto.
22512 (*ssubsi_1): Ditto.
22513 (*ssubl_se_1): Ditto.
22515 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22517 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
22518 rather than wi::to_widest for DR_INITs.
22519 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
22520 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
22521 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
22523 (vect_analyze_group_access_1): Note that here.
22525 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22527 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
22528 polynomial type sizes.
22530 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22532 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
22533 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
22534 (gimple_add_tmp_var): Likewise.
22536 2018-01-12 Martin Liska <mliska@suse.cz>
22538 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
22539 (gimple_alloc_sizes): Likewise.
22540 (dump_gimple_statistics): Use PRIu64 in printf format.
22541 * gimple.h: Change uint64_t to int.
22543 2018-01-12 Martin Liska <mliska@suse.cz>
22545 * tree-core.h: Use uint64_t instead of int.
22546 * tree.c (tree_node_counts): Likewise.
22547 (tree_node_sizes): Likewise.
22548 (dump_tree_statistics): Use PRIu64 in printf format.
22550 2018-01-12 Martin Liska <mliska@suse.cz>
22552 * Makefile.in: As qsort_chk is implemented in vec.c, add
22553 vec.o to linkage of gencfn-macros.
22554 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
22555 passing the info to record_node_allocation_statistics.
22556 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
22558 * ggc-common.c (struct ggc_usage): Add operator== and use
22559 it in operator< and compare function.
22560 * mem-stats.h (struct mem_usage): Likewise.
22561 * vec.c (struct vec_usage): Remove operator< and compare
22562 function. Can be simply inherited.
22564 2018-01-12 Martin Jambor <mjambor@suse.cz>
22567 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
22568 * tree-ssa-math-opts.c: Include domwalk.h.
22569 (convert_mult_to_fma_1): New function.
22570 (fma_transformation_info): New type.
22571 (fma_deferring_state): Likewise.
22572 (cancel_fma_deferring): New function.
22573 (result_of_phi): Likewise.
22574 (last_fma_candidate_feeds_initial_phi): Likewise.
22575 (convert_mult_to_fma): Added deferring logic, split actual
22576 transformation to convert_mult_to_fma_1.
22577 (math_opts_dom_walker): New type.
22578 (math_opts_dom_walker::after_dom_children): New method, body moved
22579 here from pass_optimize_widening_mul::execute, added deferring logic
22581 (pass_optimize_widening_mul::execute): Moved most of code to
22582 math_opts_dom_walker::after_dom_children.
22583 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
22584 * config/i386/i386.c (ix86_option_override_internal): Added
22585 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
22587 2018-01-12 Richard Biener <rguenther@suse.de>
22590 * dwarf2out.c (gen_variable_die): Do not reset old_die for
22591 inline instance vars.
22593 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
22596 * config/rx/rx.c (rx_is_restricted_memory_address):
22597 Handle SUBREG case.
22599 2018-01-12 Richard Biener <rguenther@suse.de>
22601 PR tree-optimization/80846
22602 * target.def (split_reduction): New target hook.
22603 * targhooks.c (default_split_reduction): New function.
22604 * targhooks.h (default_split_reduction): Declare.
22605 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22606 target requests first reduce vectors by combining low and high
22608 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22609 (get_vectype_for_scalar_type_and_size): Export.
22610 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22611 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22612 * doc/tm.texi: Regenerate.
22613 * config/i386/i386.c (ix86_split_reduction): Implement
22614 TARGET_VECTORIZE_SPLIT_REDUCTION.
22616 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22619 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22620 in PIC mode except for TARGET_VXWORKS_RTP.
22621 * config/sparc/sparc.c: Include cfgrtl.h.
22622 (TARGET_INIT_PIC_REG): Define.
22623 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22624 (sparc_pic_register_p): New predicate.
22625 (sparc_legitimate_address_p): Use it.
22626 (sparc_legitimize_pic_address): Likewise.
22627 (sparc_delegitimize_address): Likewise.
22628 (sparc_mode_dependent_address_p): Likewise.
22629 (gen_load_pcrel_sym): Remove 4th parameter.
22630 (load_got_register): Adjust call to above. Remove obsolete stuff.
22631 (sparc_expand_prologue): Do not call load_got_register here.
22632 (sparc_flat_expand_prologue): Likewise.
22633 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22634 (sparc_use_pseudo_pic_reg): New function.
22635 (sparc_init_pic_reg): Likewise.
22636 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22637 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22639 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
22641 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22642 Add item for branch_cost.
22644 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22646 PR rtl-optimization/83565
22647 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22648 not extend the result to a larger mode for rotate operations.
22649 (num_sign_bit_copies1): Likewise.
22651 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22654 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22656 Use values-Xc.o for -pedantic.
22657 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22659 2018-01-12 Martin Liska <mliska@suse.cz>
22662 * ipa-devirt.c (final_warning_record::grow_type_warnings):
22664 (possible_polymorphic_call_targets): Use it.
22665 (ipa_devirt): Likewise.
22667 2018-01-12 Martin Liska <mliska@suse.cz>
22669 * profile-count.h (enum profile_quality): Use 0 as invalid
22670 enum value of profile_quality.
22672 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
22674 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22675 -mext-string options.
22677 2018-01-12 Richard Biener <rguenther@suse.de>
22679 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22680 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22681 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22683 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22685 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
22687 * configure.ac (--with-long-double-format): Add support for the
22688 configuration option to change the default long double format on
22690 * config.gcc (powerpc*-linux*-*): Likewise.
22691 * configure: Regenerate.
22692 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22693 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22694 used without modification.
22696 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22698 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22699 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22700 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22701 MISC_BUILTIN_SPEC_BARRIER.
22702 (rs6000_init_builtins): Likewise.
22703 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22705 (speculation_barrier): New define_insn.
22706 * doc/extend.texi: Document __builtin_speculation_barrier.
22708 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22711 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22712 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22713 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22715 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
22716 integral modes instead of "ss" and "sd".
22717 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22718 vectors with 32-bit and 64-bit elements.
22719 (vecdupssescalarmodesuffix): New mode attribute.
22720 (vec_dup<mode>): Use it.
22722 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
22725 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22726 frame if argument is passed on stack.
22728 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22731 * ree.c (combine_reaching_defs): Optimize also
22732 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
22733 reg2=any_extend(exp); reg1=reg2;, formatting fix.
22735 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22737 PR middle-end/83189
22738 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
22740 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22742 PR middle-end/83718
22743 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
22744 after they are computed.
22746 2018-01-11 Bin Cheng <bin.cheng@arm.com>
22748 PR tree-optimization/83695
22749 * gimple-loop-linterchange.cc
22750 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
22751 reset cached scev information after interchange.
22752 (pass_linterchange::execute): Remove call to scev_reset_htab.
22754 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22756 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
22757 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
22758 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
22759 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
22760 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
22761 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
22762 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
22763 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
22764 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
22765 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
22766 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
22767 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
22768 (V_lane_reg): Likewise.
22769 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
22771 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
22772 (vfmal_lane_low<mode>_intrinsic,
22773 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
22774 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
22775 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
22776 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
22777 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
22778 vfmsl_lane_high<mode>_intrinsic): New define_insns.
22780 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22782 * config/arm/arm-cpus.in (fp16fml): New feature.
22783 (ALL_SIMD): Add fp16fml.
22784 (armv8.2-a): Add fp16fml as an option.
22785 (armv8.3-a): Likewise.
22786 (armv8.4-a): Add fp16fml as part of fp16.
22787 * config/arm/arm.h (TARGET_FP16FML): Define.
22788 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
22790 * config/arm/arm-modes.def (V2HF): Define.
22791 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
22792 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
22793 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
22794 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
22795 vfmsl_low, vfmsl_high): New set of builtins.
22796 * config/arm/iterators.md (PLUSMINUS): New code iterator.
22797 (vfml_op): New code attribute.
22798 (VFMLHALVES): New int iterator.
22799 (VFML, VFMLSEL): New mode attributes.
22800 (V_reg): Define mapping for V2HF.
22801 (V_hi, V_lo): New mode attributes.
22802 (VF_constraint): Likewise.
22803 (vfml_half, vfml_half_selector): New int attributes.
22804 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
22806 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
22807 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
22809 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
22810 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
22811 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
22812 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
22814 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
22815 Document new effective target and option set.
22817 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22819 * config/arm/arm-cpus.in (armv8_4): New feature.
22820 (ARMv8_4a): New fgroup.
22821 (armv8.4-a): New arch.
22822 * config/arm/arm-tables.opt: Regenerate.
22823 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
22824 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
22825 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
22826 Add matching rules for -march=armv8.4-a and extensions.
22827 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
22829 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
22832 * config/rx/rx.md (BW): New mode attribute.
22833 (sync_lock_test_and_setsi): Add mode suffix to insn output.
22835 2018-01-11 Richard Biener <rguenther@suse.de>
22837 PR tree-optimization/83435
22838 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
22839 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
22840 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
22842 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22843 Alan Hayward <alan.hayward@arm.com>
22844 David Sherwood <david.sherwood@arm.com>
22846 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
22848 (aarch64_classify_address): Initialize it. Track polynomial offsets.
22849 (aarch64_print_address_internal): Use it to check for a zero offset.
22851 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22852 Alan Hayward <alan.hayward@arm.com>
22853 David Sherwood <david.sherwood@arm.com>
22855 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
22856 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
22857 Return a poly_int64 rather than a HOST_WIDE_INT.
22858 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
22859 rather than a HOST_WIDE_INT.
22860 * config/aarch64/aarch64.h (aarch64_frame): Protect with
22861 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
22862 hard_fp_offset, frame_size, initial_adjust, callee_offset and
22863 final_offset from HOST_WIDE_INT to poly_int64.
22864 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
22865 to_constant when getting the number of units in an Advanced SIMD
22867 (aarch64_builtin_vectorized_function): Check for a constant number
22869 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
22871 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
22872 attribute instead of GET_MODE_NUNITS.
22873 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
22874 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
22875 GET_MODE_SIZE for fixed-size registers.
22876 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
22877 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
22878 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
22879 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
22880 (aarch64_print_operand, aarch64_print_address_internal)
22881 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
22882 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
22883 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
22884 Handle polynomial GET_MODE_SIZE.
22885 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
22886 wider than SImode without modification.
22887 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
22888 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
22889 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
22890 passing and returning SVE modes.
22891 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
22892 rather than GEN_INT.
22893 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
22894 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
22895 (aarch64_allocate_and_probe_stack_space): Likewise.
22896 (aarch64_layout_frame): Cope with polynomial offsets.
22897 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
22898 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
22899 polynomial offsets.
22900 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
22901 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
22902 poly_int64 rather than a HOST_WIDE_INT.
22903 (aarch64_get_separate_components, aarch64_process_components)
22904 (aarch64_expand_prologue, aarch64_expand_epilogue)
22905 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
22906 (aarch64_anchor_offset): New function, split out from...
22907 (aarch64_legitimize_address): ...here.
22908 (aarch64_builtin_vectorization_cost): Handle polynomial
22909 TYPE_VECTOR_SUBPARTS.
22910 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
22912 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
22913 number of elements from the PARALLEL rather than the mode.
22914 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
22915 rather than GET_MODE_BITSIZE.
22916 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
22917 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
22918 (aarch64_expand_vec_perm_const_1): Handle polynomial
22919 d->perm.length () and d->perm elements.
22920 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
22921 Apply to_constant to d->perm elements.
22922 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
22923 polynomial CONST_VECTOR_NUNITS.
22924 (aarch64_move_pointer): Take amount as a poly_int64 rather
22926 (aarch64_progress_pointer): Avoid temporary variable.
22927 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
22928 the mode attribute instead of GET_MODE.
22930 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22931 Alan Hayward <alan.hayward@arm.com>
22932 David Sherwood <david.sherwood@arm.com>
22934 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
22935 x exists before using it.
22936 (aarch64_add_constant_internal): Rename to...
22937 (aarch64_add_offset_1): ...this. Replace regnum with separate
22938 src and dest rtxes. Handle the case in which they're different,
22939 including when the offset is zero. Replace scratchreg with an rtx.
22940 Use 2 additions if there is no spare register into which we can
22941 move a 16-bit constant.
22942 (aarch64_add_constant): Delete.
22943 (aarch64_add_offset): Replace reg with separate src and dest
22944 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
22945 Use aarch64_add_offset_1.
22946 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
22947 an rtx rather than an int. Take the delta as a poly_int64
22948 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
22949 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
22950 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
22951 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
22952 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
22953 and aarch64_add_sp.
22954 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
22955 aarch64_add_constant.
22957 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22959 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22960 Use scalar_float_mode.
22962 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22964 * config/aarch64/aarch64-simd.md
22965 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
22966 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
22967 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
22968 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
22969 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
22970 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
22971 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
22972 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
22973 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
22974 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
22976 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22979 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
22980 targ_options->x_arm_arch_string is non NULL.
22982 2018-01-11 Tamar Christina <tamar.christina@arm.com>
22984 * config/aarch64/aarch64.h
22985 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
22987 2018-01-11 Sudakshina Das <sudi.das@arm.com>
22990 * expmed.c (emit_store_flag_force): Swap if const op0
22991 and change VOIDmode to mode of op0.
22993 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22995 PR rtl-optimization/83761
22996 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
22997 than bytes to mode_for_size.
22999 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23001 PR middle-end/83189
23002 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23003 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23006 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23008 PR middle-end/83575
23009 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23010 when in layout mode.
23011 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23012 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23015 2018-01-10 Michael Collison <michael.collison@arm.com>
23017 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23018 * config/aarch64/aarch64-option-extension.def: Add
23019 AARCH64_OPT_EXTENSION of 'fp16fml'.
23020 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23021 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23022 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23023 * config/aarch64/constraints.md (Ui7): New constraint.
23024 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23025 (VFMLA_SEL_W): Ditto.
23028 (VFMLA16_LOW): New int iterator.
23029 (VFMLA16_HIGH): Ditto.
23030 (UNSPEC_FMLAL): New unspec.
23031 (UNSPEC_FMLSL): Ditto.
23032 (UNSPEC_FMLAL2): Ditto.
23033 (UNSPEC_FMLSL2): Ditto.
23034 (f16mac): New code attribute.
23035 * config/aarch64/aarch64-simd-builtins.def
23036 (aarch64_fmlal_lowv2sf): Ditto.
23037 (aarch64_fmlsl_lowv2sf): Ditto.
23038 (aarch64_fmlalq_lowv4sf): Ditto.
23039 (aarch64_fmlslq_lowv4sf): Ditto.
23040 (aarch64_fmlal_highv2sf): Ditto.
23041 (aarch64_fmlsl_highv2sf): Ditto.
23042 (aarch64_fmlalq_highv4sf): Ditto.
23043 (aarch64_fmlslq_highv4sf): Ditto.
23044 (aarch64_fmlal_lane_lowv2sf): Ditto.
23045 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23046 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23047 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23048 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23049 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23050 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23051 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23052 (aarch64_fmlal_lane_highv2sf): Ditto.
23053 (aarch64_fmlsl_lane_highv2sf): Ditto.
23054 (aarch64_fmlal_laneq_highv2sf): Ditto.
23055 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23056 (aarch64_fmlalq_lane_highv4sf): Ditto.
23057 (aarch64_fmlsl_lane_highv4sf): Ditto.
23058 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23059 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23060 * config/aarch64/aarch64-simd.md:
23061 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23062 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23063 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23064 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23065 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23066 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23067 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23068 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23069 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23070 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23071 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23072 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23073 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23074 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23075 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23076 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23077 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23078 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23079 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23080 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23081 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23082 (vfmlsl_low_u32): Ditto.
23083 (vfmlalq_low_u32): Ditto.
23084 (vfmlslq_low_u32): Ditto.
23085 (vfmlal_high_u32): Ditto.
23086 (vfmlsl_high_u32): Ditto.
23087 (vfmlalq_high_u32): Ditto.
23088 (vfmlslq_high_u32): Ditto.
23089 (vfmlal_lane_low_u32): Ditto.
23090 (vfmlsl_lane_low_u32): Ditto.
23091 (vfmlal_laneq_low_u32): Ditto.
23092 (vfmlsl_laneq_low_u32): Ditto.
23093 (vfmlalq_lane_low_u32): Ditto.
23094 (vfmlslq_lane_low_u32): Ditto.
23095 (vfmlalq_laneq_low_u32): Ditto.
23096 (vfmlslq_laneq_low_u32): Ditto.
23097 (vfmlal_lane_high_u32): Ditto.
23098 (vfmlsl_lane_high_u32): Ditto.
23099 (vfmlal_laneq_high_u32): Ditto.
23100 (vfmlsl_laneq_high_u32): Ditto.
23101 (vfmlalq_lane_high_u32): Ditto.
23102 (vfmlslq_lane_high_u32): Ditto.
23103 (vfmlalq_laneq_high_u32): Ditto.
23104 (vfmlslq_laneq_high_u32): Ditto.
23105 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23106 (AARCH64_FL_FOR_ARCH8_4): New.
23107 (AARCH64_ISA_F16FML): New ISA flag.
23108 (TARGET_F16FML): New feature flag for fp16fml.
23109 (doc/invoke.texi): Document new fp16fml option.
23111 2018-01-10 Michael Collison <michael.collison@arm.com>
23113 * config/aarch64/aarch64-builtins.c:
23114 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23115 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23116 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23117 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23118 (AARCH64_ISA_SHA3): New ISA flag.
23119 (TARGET_SHA3): New feature flag for sha3.
23120 * config/aarch64/iterators.md (sha512_op): New int attribute.
23121 (CRYPTO_SHA512): New int iterator.
23122 (UNSPEC_SHA512H): New unspec.
23123 (UNSPEC_SHA512H2): Ditto.
23124 (UNSPEC_SHA512SU0): Ditto.
23125 (UNSPEC_SHA512SU1): Ditto.
23126 * config/aarch64/aarch64-simd-builtins.def
23127 (aarch64_crypto_sha512hqv2di): New builtin.
23128 (aarch64_crypto_sha512h2qv2di): Ditto.
23129 (aarch64_crypto_sha512su0qv2di): Ditto.
23130 (aarch64_crypto_sha512su1qv2di): Ditto.
23131 (aarch64_eor3qv8hi): Ditto.
23132 (aarch64_rax1qv2di): Ditto.
23133 (aarch64_xarqv2di): Ditto.
23134 (aarch64_bcaxqv8hi): Ditto.
23135 * config/aarch64/aarch64-simd.md:
23136 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23137 (aarch64_crypto_sha512su0qv2di): Ditto.
23138 (aarch64_crypto_sha512su1qv2di): Ditto.
23139 (aarch64_eor3qv8hi): Ditto.
23140 (aarch64_rax1qv2di): Ditto.
23141 (aarch64_xarqv2di): Ditto.
23142 (aarch64_bcaxqv8hi): Ditto.
23143 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23144 (vsha512h2q_u64): Ditto.
23145 (vsha512su0q_u64): Ditto.
23146 (vsha512su1q_u64): Ditto.
23147 (veor3q_u16): Ditto.
23148 (vrax1q_u64): Ditto.
23149 (vxarq_u64): Ditto.
23150 (vbcaxq_u16): Ditto.
23151 * config/arm/types.md (crypto_sha512): New type attribute.
23152 (crypto_sha3): Ditto.
23153 (doc/invoke.texi): Document new sha3 option.
23155 2018-01-10 Michael Collison <michael.collison@arm.com>
23157 * config/aarch64/aarch64-builtins.c:
23158 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23159 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23160 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23161 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23162 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23163 (AARCH64_ISA_SM4): New ISA flag.
23164 (TARGET_SM4): New feature flag for sm4.
23165 * config/aarch64/aarch64-simd-builtins.def
23166 (aarch64_sm3ss1qv4si): Ditto.
23167 (aarch64_sm3tt1aq4si): Ditto.
23168 (aarch64_sm3tt1bq4si): Ditto.
23169 (aarch64_sm3tt2aq4si): Ditto.
23170 (aarch64_sm3tt2bq4si): Ditto.
23171 (aarch64_sm3partw1qv4si): Ditto.
23172 (aarch64_sm3partw2qv4si): Ditto.
23173 (aarch64_sm4eqv4si): Ditto.
23174 (aarch64_sm4ekeyqv4si): Ditto.
23175 * config/aarch64/aarch64-simd.md:
23176 (aarch64_sm3ss1qv4si): Ditto.
23177 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23178 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23179 (aarch64_sm4eqv4si): Ditto.
23180 (aarch64_sm4ekeyqv4si): Ditto.
23181 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23182 (sm3part_op): Ditto.
23183 (CRYPTO_SM3TT): Ditto.
23184 (CRYPTO_SM3PART): Ditto.
23185 (UNSPEC_SM3SS1): New unspec.
23186 (UNSPEC_SM3TT1A): Ditto.
23187 (UNSPEC_SM3TT1B): Ditto.
23188 (UNSPEC_SM3TT2A): Ditto.
23189 (UNSPEC_SM3TT2B): Ditto.
23190 (UNSPEC_SM3PARTW1): Ditto.
23191 (UNSPEC_SM3PARTW2): Ditto.
23192 (UNSPEC_SM4E): Ditto.
23193 (UNSPEC_SM4EKEY): Ditto.
23194 * config/aarch64/constraints.md (Ui2): New constraint.
23195 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23196 * config/arm/types.md (crypto_sm3): New type attribute.
23197 (crypto_sm4): Ditto.
23198 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23199 (vsm3tt1aq_u32): Ditto.
23200 (vsm3tt1bq_u32): Ditto.
23201 (vsm3tt2aq_u32): Ditto.
23202 (vsm3tt2bq_u32): Ditto.
23203 (vsm3partw1q_u32): Ditto.
23204 (vsm3partw2q_u32): Ditto.
23205 (vsm4eq_u32): Ditto.
23206 (vsm4ekeyq_u32): Ditto.
23207 (doc/invoke.texi): Document new sm4 option.
23209 2018-01-10 Michael Collison <michael.collison@arm.com>
23211 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23212 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23213 (AARCH64_FL_FOR_ARCH8_4): New.
23214 (AARCH64_FL_V8_4): New flag.
23215 (doc/invoke.texi): Document new armv8.4-a option.
23217 2018-01-10 Michael Collison <michael.collison@arm.com>
23219 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23220 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23221 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23222 * config/aarch64/aarch64-option-extension.def: Add
23223 AARCH64_OPT_EXTENSION of 'sha2'.
23224 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23225 (crypto): Disable sha2 and aes if crypto disabled.
23226 (crypto): Enable aes and sha2 if enabled.
23227 (simd): Disable sha2 and aes if simd disabled.
23228 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23230 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23231 (TARGET_SHA2): New feature flag for sha2.
23232 (TARGET_AES): New feature flag for aes.
23233 * config/aarch64/aarch64-simd.md:
23234 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23235 conditional on TARGET_AES.
23236 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23237 (aarch64_crypto_sha1hsi): Make pattern conditional
23239 (aarch64_crypto_sha1hv4si): Ditto.
23240 (aarch64_be_crypto_sha1hv4si): Ditto.
23241 (aarch64_crypto_sha1su1v4si): Ditto.
23242 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23243 (aarch64_crypto_sha1su0v4si): Ditto.
23244 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23245 (aarch64_crypto_sha256su0v4si): Ditto.
23246 (aarch64_crypto_sha256su1v4si): Ditto.
23247 (doc/invoke.texi): Document new aes and sha2 options.
23249 2018-01-10 Martin Sebor <msebor@redhat.com>
23251 PR tree-optimization/83781
23252 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23255 2018-01-11 Martin Sebor <msebor@gmail.com>
23256 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23258 PR tree-optimization/83501
23259 PR tree-optimization/81703
23261 * tree-ssa-strlen.c (get_string_cst): Rename...
23262 (get_string_len): ...to this. Handle global constants.
23263 (handle_char_store): Adjust.
23265 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23266 Jim Wilson <jimw@sifive.com>
23268 * config/riscv/riscv-protos.h (riscv_output_return): New.
23269 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23270 (riscv_attribute_table, riscv_output_return),
23271 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23272 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23273 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23275 (riscv_expand_prologue): Add early return for naked function.
23276 (riscv_expand_epilogue): Likewise.
23277 (riscv_function_ok_for_sibcall): Return false for naked function.
23278 (riscv_set_current_function): New.
23279 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23280 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23281 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23282 * doc/extend.texi (RISC-V Function Attributes): New.
23284 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23286 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23287 check for 128-bit long double before checking TCmode.
23288 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23289 128-bit long doubles before checking TFmode or TCmode.
23290 (FLOAT128_IBM_P): Likewise.
23292 2018-01-10 Martin Sebor <msebor@redhat.com>
23294 PR tree-optimization/83671
23295 * builtins.c (c_strlen): Unconditionally return zero for the empty
23297 Use -Warray-bounds for warnings.
23298 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23299 for non-constant array indices with COMPONENT_REF, arrays of
23300 arrays, and pointers to arrays.
23301 (gimple_fold_builtin_strlen): Determine and set length range for
23302 non-constant character arrays.
23304 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23306 PR middle-end/81897
23307 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23310 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23312 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23314 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23317 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23318 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23319 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23320 indexed_or_indirect_operand predicate.
23321 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23322 (*vsx_le_perm_load_v8hi): Likewise.
23323 (*vsx_le_perm_load_v16qi): Likewise.
23324 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23325 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23326 (*vsx_le_perm_store_v8hi): Likewise.
23327 (*vsx_le_perm_store_v16qi): Likewise.
23328 (eight unnamed splitters): Likewise.
23330 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23332 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23333 * config/rs6000/emmintrin.h: Likewise.
23334 * config/rs6000/mmintrin.h: Likewise.
23335 * config/rs6000/xmmintrin.h: Likewise.
23337 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23340 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23342 * tree.c (tree_nop_conversion): Return true for location wrapper
23344 (maybe_wrap_with_location): New function.
23345 (selftest::check_strip_nops): New function.
23346 (selftest::test_location_wrappers): New function.
23347 (selftest::tree_c_tests): Call it.
23348 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23349 (maybe_wrap_with_location): New decl.
23350 (EXPR_LOCATION_WRAPPER_P): New macro.
23351 (location_wrapper_p): New inline function.
23352 (tree_strip_any_location_wrapper): New inline function.
23354 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23357 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23358 stack_realign_offset for the largest alignment of stack slot
23360 (ix86_find_max_used_stack_alignment): New function.
23361 (ix86_finalize_stack_frame_flags): Use it. Set
23362 max_used_stack_alignment if we don't realign stack.
23363 * config/i386/i386.h (machine_function): Add
23364 max_used_stack_alignment.
23366 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23368 * config/arm/arm.opt (-mbranch-cost): New option.
23369 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23372 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23375 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23376 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23378 2018-01-10 Richard Biener <rguenther@suse.de>
23381 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23382 early out so it also covers the case where we have a non-NULL
23385 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23387 PR tree-optimization/83753
23388 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23389 for non-strided grouped accesses if the number of elements is 1.
23391 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23394 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23395 * i386.h (TARGET_USE_GATHER): Define.
23396 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23398 2018-01-10 Martin Liska <mliska@suse.cz>
23401 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23402 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23404 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23405 CLEANUP_NO_PARTITIONING is not set.
23407 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23409 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23410 for vectors, as a partial revert of r254296.
23411 * rtl.h (const_vec_p): Delete.
23412 (const_vec_duplicate_p): Don't test for vector CONSTs.
23413 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23414 * expmed.c (make_tree): Likewise.
23417 * common.md (E, F): Use CONSTANT_P instead of checking for
23419 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23420 checking for CONST_VECTOR.
23422 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23424 PR middle-end/83575
23425 * predict.c (force_edge_cold): Handle in more sane way edges
23426 with no prediction.
23428 2018-01-09 Carl Love <cel@us.ibm.com>
23430 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
23432 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
23433 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
23434 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
23435 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
23436 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
23437 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
23438 * config/rs6000/rs6000-protos.h: Add extern defition for
23439 rs6000_generate_float2_double_code.
23440 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
23442 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
23443 (float2_v2df): Add define_expand.
23445 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
23448 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
23449 op_mode in the force_to_mode call.
23451 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23453 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23454 instead of checking each element individually.
23455 (aarch64_evpc_uzp): Likewise.
23456 (aarch64_evpc_zip): Likewise.
23457 (aarch64_evpc_ext): Likewise.
23458 (aarch64_evpc_rev): Likewise.
23459 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
23460 instead of checking each element individually. Return true without
23462 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
23463 whether all selected elements come from the same input, instead of
23464 checking each element individually. Remove calls to gen_rtx_REG,
23465 start_sequence and end_sequence and instead assert that no rtl is
23468 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23470 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
23471 order of HIGH and CONST checks.
23473 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23475 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
23476 if the destination isn't an SSA_NAME.
23478 2018-01-09 Richard Biener <rguenther@suse.de>
23480 PR tree-optimization/83668
23481 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
23483 (canonicalize_loop_form): ... here, renamed from ...
23484 (canonicalize_loop_closed_ssa_form): ... this and amended to
23485 swap successor edges for loop exit blocks to make us use
23486 the RPO order we need for initial schedule generation.
23488 2018-01-09 Joseph Myers <joseph@codesourcery.com>
23490 PR tree-optimization/64811
23491 * match.pd: When optimizing comparisons with Inf, avoid
23492 introducing or losing exceptions from comparisons with NaN.
23494 2018-01-09 Martin Liska <mliska@suse.cz>
23497 * asan.c (shadow_mem_size): Add gcc_assert.
23499 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
23501 Don't save registers in main().
23504 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
23505 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
23506 * config/avr/avr.c (avr_set_current_function): Don't error if
23507 naked, OS_task or OS_main are specified at the same time.
23508 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
23510 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
23512 * common/config/avr/avr-common.c (avr_option_optimization_table):
23513 Switch on -mmain-is-OS_task for optimizing compilations.
23515 2018-01-09 Richard Biener <rguenther@suse.de>
23517 PR tree-optimization/83572
23518 * graphite.c: Include cfganal.h.
23519 (graphite_transform_loops): Connect infinite loops to exit
23520 and remove fake edges at the end.
23522 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23524 * ipa-inline.c (edge_badness): Revert accidental checkin.
23526 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23529 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
23530 symbols; not inline clones.
23532 2018-01-09 Jakub Jelinek <jakub@redhat.com>
23535 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
23536 hard registers. Formatting fixes.
23538 PR preprocessor/83722
23539 * gcc.c (try_generate_repro): Pass
23540 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
23541 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
23544 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
23545 Kito Cheng <kito.cheng@gmail.com>
23547 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
23548 (riscv_leaf_function_p): Delete.
23549 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
23551 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23553 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
23555 (do_ifelse): New function.
23556 (do_isel): New function.
23557 (do_sub3): New function.
23558 (do_add3): New function.
23559 (do_load_mask_compare): New function.
23560 (do_overlap_load_compare): New function.
23561 (expand_compare_loop): New function.
23562 (expand_block_compare): Call expand_compare_loop() when appropriate.
23563 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
23564 option description.
23565 (-mblock-compare-inline-loop-limit): New option.
23567 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23570 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
23571 Reverse order of second and third operands in first alternative.
23572 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
23573 of first and second elements in UNSPEC_VPERMR vector.
23574 (altivec_expand_vec_perm_le): Likewise.
23576 2018-01-08 Jeff Law <law@redhat.com>
23578 PR rtl-optimizatin/81308
23579 * tree-switch-conversion.c (cfg_altered): New file scoped static.
23580 (process_switch): If group_case_labels makes a change, then set
23582 (pass_convert_switch::execute): If a switch is converted, then
23583 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
23585 PR rtl-optimization/81308
23586 * recog.c (split_all_insns): Conditionally cleanup the CFG after
23589 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
23591 PR target/83663 - Revert r255946
23592 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
23593 generation for cases where splatting a value is not useful.
23594 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
23595 across a vec_duplicate and a paradoxical subreg forming a vector
23596 mode to a vec_concat.
23598 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23600 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23601 -march=armv8.3-a variants.
23602 * config/arm/t-multilib: Likewise.
23603 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
23605 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23607 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23609 (cceq_ior_compare_complement): Give it a name so I can use it, and
23610 change boolean_or_operator predicate to boolean_operator so it can
23611 be used to generate a crand.
23612 (eqne): New code iterator.
23613 (bd/bd_neg): New code_attrs.
23614 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23615 a single define_insn.
23616 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23617 decrement (bdnzt/bdnzf/bdzt/bdzf).
23618 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23619 with the new names of the branch decrement patterns, and added the
23620 names of the branch decrement conditional patterns.
23622 2018-01-08 Richard Biener <rguenther@suse.de>
23624 PR tree-optimization/83563
23625 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23628 2018-01-08 Richard Biener <rguenther@suse.de>
23630 PR middle-end/83713
23631 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23633 2018-01-08 Richard Biener <rguenther@suse.de>
23635 PR tree-optimization/83685
23636 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23637 references to abnormals.
23639 2018-01-08 Richard Biener <rguenther@suse.de>
23642 * dwarf2out.c (output_indirect_strings): Handle empty
23643 skeleton_debug_str_hash.
23644 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23646 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23648 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23649 (emit_store_direct): Likewise.
23650 (arc_trampoline_adjust_address): Likewise.
23651 (arc_asm_trampoline_template): New function.
23652 (arc_initialize_trampoline): Use asm_trampoline_template.
23653 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23654 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23655 * config/arc/arc.md (flush_icache): Delete pattern.
23657 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23659 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23660 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23663 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23666 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23667 by not USED_FOR_TARGET.
23668 (make_pass_resolve_sw_modes): Likewise.
23670 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23672 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23675 2018-01-08 Richard Biener <rguenther@suse.de>
23677 PR middle-end/83580
23678 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23680 2018-01-08 Richard Biener <rguenther@suse.de>
23682 PR middle-end/83517
23683 * match.pd ((t * 2) / 2) -> t): Add missing :c.
23685 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
23687 PR middle-end/81897
23688 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23689 basic blocks with a small number of successors.
23690 (convert_control_dep_chain_into_preds): Improve handling of
23692 (dump_predicates): Split apart into...
23693 (dump_pred_chain): ...here...
23694 (dump_pred_info): ...and here.
23695 (can_one_predicate_be_invalidated_p): Add debugging printfs.
23696 (can_chain_union_be_invalidated_p): Improve check for invalidation
23698 (uninit_uses_cannot_happen): Avoid unnecessary if
23699 convert_control_dep_chain_into_preds yielded nothing.
23701 2018-01-06 Martin Sebor <msebor@redhat.com>
23703 PR tree-optimization/83640
23704 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23705 subtracting negative offset from size.
23706 (builtin_access::overlap): Adjust offset bounds of the access to fall
23707 within the size of the object if possible.
23709 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
23711 PR rtl-optimization/83699
23712 * expmed.c (extract_bit_field_1): Restrict the vector usage of
23713 extract_bit_field_as_subreg to cases in which the extracted
23714 value is also a vector.
23716 * lra-constraints.c (process_alt_operands): Test for the equivalence
23717 substitutions when detecting a possible reload cycle.
23719 2018-01-06 Jakub Jelinek <jakub@redhat.com>
23722 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23723 by default if flag_selective_schedling{,2}. Formatting fixes.
23725 PR rtl-optimization/83682
23726 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23727 if it has non-VECTOR_MODE element mode.
23728 (vec_duplicate_p): Likewise.
23730 PR middle-end/83694
23731 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
23732 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
23734 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23737 * config/i386/i386-builtin.def
23738 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
23739 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
23740 Require also OPTION_MASK_ISA_AVX512F in addition to
23741 OPTION_MASK_ISA_GFNI.
23742 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
23743 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
23744 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
23745 to OPTION_MASK_ISA_GFNI.
23746 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
23747 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
23748 OPTION_MASK_ISA_AVX512BW.
23749 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
23750 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
23751 addition to OPTION_MASK_ISA_GFNI.
23752 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
23753 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
23754 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
23755 to OPTION_MASK_ISA_GFNI.
23756 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
23757 a requirement for all ISAs rather than any of them with a few
23759 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
23761 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
23762 bitmasks to be enabled with 3 exceptions, instead of requiring any
23763 enabled ISA with lots of exceptions.
23764 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
23765 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
23766 Change avx512bw in isa attribute to avx512f.
23767 * config/i386/sgxintrin.h: Add license boilerplate.
23768 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
23769 to __AVX512F__ and __AVX512VL to __AVX512VL__.
23770 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
23771 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
23773 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
23774 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
23775 temporarily sse2 rather than sse if not enabled already.
23778 * config/i386/sse.md (VI248_VLBW): Rename to ...
23779 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
23780 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
23781 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
23782 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
23783 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
23784 mode iterator instead of VI248_VLBW.
23786 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
23788 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
23789 (record_modified): Skip clobbers; add debug output.
23790 (param_change_prob): Use sreal frequencies.
23792 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23794 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
23795 punt for user-aligned variables.
23797 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23799 * tree-chrec.c (chrec_contains_symbols): Return true for
23802 2018-01-05 Sudakshina Das <sudi.das@arm.com>
23805 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
23806 of (x|y) == x for BICS pattern.
23808 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23810 PR tree-optimization/83605
23811 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
23812 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
23815 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
23817 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
23818 * config/epiphany/rtems.h: New file.
23820 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23821 Uros Bizjak <ubizjak@gmail.com>
23824 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
23825 QIreg_operand instead of register_operand predicate.
23826 * config/i386/i386.c (ix86_rop_should_change_byte_p,
23827 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
23828 comments instead of -fmitigate[-_]rop.
23830 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23833 * cgraphunit.c (symbol_table::compile): Switch to text_section
23834 before calling assembly_start debug hook.
23835 * run-rtl-passes.c (run_rtl_passes): Likewise.
23838 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23840 * tree-vrp.c (extract_range_from_binary_expr_1): Check
23841 range_int_cst_p rather than !symbolic_range_p before calling
23842 extract_range_from_multiplicative_op_1.
23844 2018-01-04 Jeff Law <law@redhat.com>
23846 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
23847 redundant test in assertion.
23849 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23851 * doc/rtl.texi: Document machine_mode wrapper classes.
23853 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23855 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
23856 using tree_to_uhwi.
23858 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23860 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
23861 the VEC_PERM_EXPR fold to fail.
23863 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23866 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
23867 to switched_sections.
23869 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23872 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
23873 test for d.testing.
23875 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
23878 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
23879 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
23881 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23884 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
23885 is BLKmode and bitpos not zero or mode change is needed.
23887 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23890 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
23893 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
23896 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
23897 instead of MULT rtx. Update all corresponding splitters.
23898 (*saddl_se): Ditto.
23899 (*ssub<modesuffix>): Ditto.
23900 (*ssubl_se): Ditto.
23901 (*cmp_sadd_di): Update split patterns.
23902 (*cmp_sadd_si): Ditto.
23903 (*cmp_sadd_sidi): Ditto.
23904 (*cmp_ssub_di): Ditto.
23905 (*cmp_ssub_si): Ditto.
23906 (*cmp_ssub_sidi): Ditto.
23907 * config/alpha/predicates.md (const23_operand): New predicate.
23908 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
23909 Look for ASHIFT, not MULT inner operand.
23910 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
23912 2018-01-04 Martin Liska <mliska@suse.cz>
23914 PR gcov-profile/83669
23915 * gcov.c (output_intermediate_file): Add version to intermediate
23917 * doc/gcov.texi: Document new field 'version' in intermediate
23918 file format. Fix location of '-k' option of gcov command.
23920 2018-01-04 Martin Liska <mliska@suse.cz>
23923 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
23925 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23927 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
23929 2018-01-03 Martin Sebor <msebor@redhat.com>
23931 PR tree-optimization/83655
23932 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
23933 checking calls with invalid arguments.
23935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23937 * tree-vect-stmts.c (vect_get_store_rhs): New function.
23938 (vectorizable_mask_load_store): Delete.
23939 (vectorizable_call): Return false for masked loads and stores.
23940 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
23941 instead of gimple_assign_rhs1.
23942 (vectorizable_load): Handle IFN_MASK_LOAD.
23943 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
23945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23947 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
23949 (vectorizable_mask_load_store): ...here.
23950 (vectorizable_load): ...and here.
23952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23954 * tree-vect-stmts.c (vect_build_all_ones_mask)
23955 (vect_build_zero_merge_argument): New functions, split out from...
23956 (vectorizable_load): ...here.
23958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23960 * tree-vect-stmts.c (vect_check_store_rhs): New function,
23962 (vectorizable_mask_load_store): ...here.
23963 (vectorizable_store): ...and here.
23965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23967 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
23969 (vectorizable_mask_load_store): ...here.
23971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23973 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
23974 (vect_model_store_cost): Take a vec_load_store_type instead of a
23976 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
23977 (vect_model_store_cost): Take a vec_load_store_type instead of a
23979 (vectorizable_mask_load_store): Update accordingly.
23980 (vectorizable_store): Likewise.
23981 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
23983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23985 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
23986 IFN_MASK_LOAD calls here rather than...
23987 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
23989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23990 Alan Hayward <alan.hayward@arm.com>
23991 David Sherwood <david.sherwood@arm.com>
23993 * expmed.c (extract_bit_field_1): For vector extracts,
23994 fall back to extract_bit_field_as_subreg if vec_extract
23997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23998 Alan Hayward <alan.hayward@arm.com>
23999 David Sherwood <david.sherwood@arm.com>
24001 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24002 they are variable or constant sized.
24003 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24004 slots for constant-sized data.
24006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24007 Alan Hayward <alan.hayward@arm.com>
24008 David Sherwood <david.sherwood@arm.com>
24010 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24011 handling COND_EXPRs with boolean comparisons, try to find a better
24012 basis for the mask type than the boolean itself.
24014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24016 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24017 is calculated and how it can be overridden.
24018 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24019 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24021 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24025 Alan Hayward <alan.hayward@arm.com>
24026 David Sherwood <david.sherwood@arm.com>
24028 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24029 Remove the mode argument.
24030 (aarch64_simd_valid_immediate): Remove the mode and inverse
24032 * config/aarch64/iterators.md (bitsize): New iterator.
24033 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24034 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24035 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24036 aarch64_simd_valid_immediate.
24037 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24038 (aarch64_reg_or_bic_imm): Likewise.
24039 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24040 with an insn_type enum and msl with a modifier_type enum.
24041 Replace element_width with a scalar_mode. Change the shift
24042 to unsigned int. Add constructors for scalar_float_mode and
24043 scalar_int_mode elements.
24044 (aarch64_vect_float_const_representable_p): Delete.
24045 (aarch64_can_const_movi_rtx_p)
24046 (aarch64_simd_scalar_immediate_valid_for_move)
24047 (aarch64_simd_make_constant): Update call to
24048 aarch64_simd_valid_immediate.
24049 (aarch64_advsimd_valid_immediate_hs): New function.
24050 (aarch64_advsimd_valid_immediate): Likewise.
24051 (aarch64_simd_valid_immediate): Remove mode and inverse
24052 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24053 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24054 and aarch64_float_const_representable_p on the result.
24055 (aarch64_output_simd_mov_immediate): Remove mode argument.
24056 Update call to aarch64_simd_valid_immediate and use of
24057 simd_immediate_info.
24058 (aarch64_output_scalar_simd_mov_immediate): Update call
24061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24062 Alan Hayward <alan.hayward@arm.com>
24063 David Sherwood <david.sherwood@arm.com>
24065 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24066 (mode_nunits): Likewise CONST_MODE_NUNITS.
24067 * machmode.def (ADJUST_NUNITS): Document.
24068 * genmodes.c (mode_data::need_nunits_adj): New field.
24069 (blank_mode): Update accordingly.
24070 (adj_nunits): New variable.
24071 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24073 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24074 listed in adj_nunits.
24075 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24076 listed in adj_nunits. Don't emit case statements for such modes.
24077 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24078 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24079 nothing if adj_nunits is nonnull.
24080 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24081 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24082 (emit_mode_fbit): Update use of print_maybe_const_decl.
24083 (emit_move_size): Likewise. Treat the array as non-const
24085 (emit_mode_adjustments): Handle adj_nunits.
24087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24089 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24090 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24091 (VECTOR_MODES): Use it.
24092 (make_vector_modes): Take the prefix as an argument.
24094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24095 Alan Hayward <alan.hayward@arm.com>
24096 David Sherwood <david.sherwood@arm.com>
24098 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24099 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24100 for MODE_VECTOR_BOOL.
24101 * machmode.def (VECTOR_BOOL_MODE): Document.
24102 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24103 (make_vector_bool_mode): New function.
24104 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24106 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24107 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24109 * stor-layout.c (int_mode_for_mode): Likewise.
24110 * tree.c (build_vector_type_for_mode): Likewise.
24111 * varasm.c (output_constant_pool_2): Likewise.
24112 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24113 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24114 for MODE_VECTOR_BOOL.
24115 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24116 of mode class checks.
24117 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24118 instead of a list of mode class checks.
24119 (expand_vector_scalar_condition): Likewise.
24120 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24123 Alan Hayward <alan.hayward@arm.com>
24124 David Sherwood <david.sherwood@arm.com>
24126 * machmode.h (mode_size): Change from unsigned short to
24128 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24129 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24130 or if measurement_type is not polynomial.
24131 (fixed_size_mode::includes_p): Check for constant-sized modes.
24132 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24133 return a poly_uint16 rather than an unsigned short.
24134 (emit_mode_size): Change the type of mode_size from unsigned short
24135 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24136 (emit_mode_adjustments): Cope with polynomial vector sizes.
24137 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24139 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24141 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24142 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24143 * caller-save.c (setup_save_areas): Likewise.
24144 (replace_reg_with_saved_mem): Likewise.
24145 * calls.c (emit_library_call_value_1): Likewise.
24146 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24147 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24148 (gen_lowpart_for_combine): Likewise.
24149 * convert.c (convert_to_integer_1): Likewise.
24150 * cse.c (equiv_constant, cse_insn): Likewise.
24151 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24152 (cselib_subst_to_values): Likewise.
24153 * dce.c (word_dce_process_block): Likewise.
24154 * df-problems.c (df_word_lr_mark_ref): Likewise.
24155 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24156 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24157 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24158 (rtl_for_decl_location): Likewise.
24159 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24160 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24161 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24162 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24163 (expand_expr_real_1): Likewise.
24164 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24165 (pad_below): Likewise.
24166 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24167 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24168 * ira.c (get_subreg_tracking_sizes): Likewise.
24169 * ira-build.c (ira_create_allocno_objects): Likewise.
24170 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24171 (ira_sort_regnos_for_alter_reg): Likewise.
24172 * ira-costs.c (record_operand_costs): Likewise.
24173 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24174 (resolve_simple_move): Likewise.
24175 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24176 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24177 (lra_constraints): Likewise.
24178 (CONST_POOL_OK_P): Reject variable-sized modes.
24179 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24180 (add_pseudo_to_slot, lra_spill): Likewise.
24181 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24182 * optabs-query.c (get_best_extraction_insn): Likewise.
24183 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24184 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24185 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24186 * recog.c (offsettable_address_addr_space_p): Likewise.
24187 * regcprop.c (maybe_mode_change): Likewise.
24188 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24189 * regrename.c (build_def_use): Likewise.
24190 * regstat.c (dump_reg_info): Likewise.
24191 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24192 (find_reloads, find_reloads_subreg_address): Likewise.
24193 * reload1.c (eliminate_regs_1): Likewise.
24194 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24195 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24196 (simplify_binary_operation_1, simplify_subreg): Likewise.
24197 * targhooks.c (default_function_arg_padding): Likewise.
24198 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24199 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24200 (verify_gimple_assign_ternary): Likewise.
24201 * tree-inline.c (estimate_move_cost): Likewise.
24202 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24203 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24204 (get_address_cost_ainc): Likewise.
24205 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24206 (vect_supportable_dr_alignment): Likewise.
24207 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24208 (vectorizable_reduction): Likewise.
24209 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24210 (vectorizable_operation, vectorizable_load): Likewise.
24211 * tree.c (build_same_sized_truth_vector_type): Likewise.
24212 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24213 * var-tracking.c (emit_note_insn_var_location): Likewise.
24214 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24215 (ADDR_VEC_ALIGN): Likewise.
24217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24218 Alan Hayward <alan.hayward@arm.com>
24219 David Sherwood <david.sherwood@arm.com>
24221 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24223 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24224 or if measurement_type is polynomial.
24225 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24226 * combine.c (make_extraction): Likewise.
24227 * dse.c (find_shift_sequence): Likewise.
24228 * dwarf2out.c (mem_loc_descriptor): Likewise.
24229 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24230 (extract_bit_field, extract_low_bits): Likewise.
24231 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24232 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24233 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24234 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24235 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24236 * reload.c (find_reloads): Likewise.
24237 * reload1.c (alter_reg): Likewise.
24238 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24239 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24240 * tree-if-conv.c (predicate_mem_writes): Likewise.
24241 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24242 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24243 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24244 * valtrack.c (dead_debug_insert_temp): Likewise.
24245 * varasm.c (mergeable_constant_section): Likewise.
24246 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24249 Alan Hayward <alan.hayward@arm.com>
24250 David Sherwood <david.sherwood@arm.com>
24252 * expr.c (expand_assignment): Cope with polynomial mode sizes
24253 when assigning to a CONCAT.
24255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24256 Alan Hayward <alan.hayward@arm.com>
24257 David Sherwood <david.sherwood@arm.com>
24259 * machmode.h (mode_precision): Change from unsigned short to
24261 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24263 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24264 or if measurement_type is not polynomial.
24265 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24266 in which the mode is already known to be a scalar_int_mode.
24267 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24268 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24270 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24271 for GET_MODE_PRECISION.
24272 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24273 for GET_MODE_PRECISION.
24274 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24276 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24277 (expand_field_assignment, make_extraction): Likewise.
24278 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24279 (get_last_value): Likewise.
24280 * convert.c (convert_to_integer_1): Likewise.
24281 * cse.c (cse_insn): Likewise.
24282 * expr.c (expand_expr_real_1): Likewise.
24283 * lra-constraints.c (simplify_operand_subreg): Likewise.
24284 * optabs-query.c (can_atomic_load_p): Likewise.
24285 * optabs.c (expand_atomic_load): Likewise.
24286 (expand_atomic_store): Likewise.
24287 * ree.c (combine_reaching_defs): Likewise.
24288 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24289 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24290 * tree.h (type_has_mode_precision_p): Likewise.
24291 * ubsan.c (instrument_si_overflow): Likewise.
24293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24294 Alan Hayward <alan.hayward@arm.com>
24295 David Sherwood <david.sherwood@arm.com>
24297 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24298 polynomial numbers of units.
24299 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24300 (valid_vector_subparts_p): New function.
24301 (build_vector_type): Remove temporary shim and take the number
24302 of units as a poly_uint64 rather than an int.
24303 (build_opaque_vector_type): Take the number of units as a
24304 poly_uint64 rather than an int.
24305 * tree.c (build_vector_from_ctor): Handle polynomial
24306 TYPE_VECTOR_SUBPARTS.
24307 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24308 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24309 (build_vector_from_val): If the number of units is variable,
24310 use build_vec_duplicate_cst for constant operands and
24311 VEC_DUPLICATE_EXPR otherwise.
24312 (make_vector_type): Remove temporary is_constant ().
24313 (build_vector_type, build_opaque_vector_type): Take the number of
24314 units as a poly_uint64 rather than an int.
24315 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24317 * cfgexpand.c (expand_debug_expr): Likewise.
24318 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24319 (store_constructor, expand_expr_real_1): Likewise.
24320 (const_scalar_mask_from_tree): Likewise.
24321 * fold-const-call.c (fold_const_reduction): Likewise.
24322 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24323 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24324 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24325 (fold_relational_const): Likewise.
24326 (native_interpret_vector): Likewise. Change the size from an
24327 int to an unsigned int.
24328 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24329 TYPE_VECTOR_SUBPARTS.
24330 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24331 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24332 duplicating a non-constant operand into a variable-length vector.
24333 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24334 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24335 * ipa-icf.c (sem_variable::equals): Likewise.
24336 * match.pd: Likewise.
24337 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24338 * print-tree.c (print_node): Likewise.
24339 * stor-layout.c (layout_type): Likewise.
24340 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24341 * tree-cfg.c (verify_gimple_comparison): Likewise.
24342 (verify_gimple_assign_binary): Likewise.
24343 (verify_gimple_assign_ternary): Likewise.
24344 (verify_gimple_assign_single): Likewise.
24345 * tree-pretty-print.c (dump_generic_node): Likewise.
24346 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24347 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24348 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24349 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24350 (vect_shift_permute_load_chain): Likewise.
24351 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24352 (expand_vector_condition, optimize_vector_constructor): Likewise.
24353 (lower_vec_perm, get_compute_type): Likewise.
24354 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24355 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24356 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24357 (vect_recog_mask_conversion_pattern): Likewise.
24358 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24359 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24360 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24361 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24362 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24363 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24364 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24365 (supportable_widening_operation): Likewise.
24366 (supportable_narrowing_operation): Likewise.
24367 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24369 * varasm.c (output_constant): Likewise.
24371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24372 Alan Hayward <alan.hayward@arm.com>
24373 David Sherwood <david.sherwood@arm.com>
24375 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24376 so that both the length == 3 and length != 3 cases set up their
24377 own permute vectors. Add comments explaining why we know the
24378 number of elements is constant.
24379 (vect_permute_load_chain): Likewise.
24381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24382 Alan Hayward <alan.hayward@arm.com>
24383 David Sherwood <david.sherwood@arm.com>
24385 * machmode.h (mode_nunits): Change from unsigned char to
24387 (ONLY_FIXED_SIZE_MODES): New macro.
24388 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24389 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24390 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24392 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24393 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24394 or if measurement_type is not polynomial.
24395 * genmodes.c (ZERO_COEFFS): New macro.
24396 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24398 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24399 Use ZERO_COEFFS when emitting initializers.
24400 * data-streamer.h (bp_pack_poly_value): New function.
24401 (bp_unpack_poly_value): Likewise.
24402 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24403 for GET_MODE_NUNITS.
24404 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24405 for GET_MODE_NUNITS.
24406 * tree.c (make_vector_type): Remove temporary shim and make
24407 the real function take the number of units as a poly_uint64
24408 rather than an int.
24409 (build_vector_type_for_mode): Handle polynomial nunits.
24410 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24411 * emit-rtl.c (const_vec_series_p_1): Likewise.
24412 (gen_rtx_CONST_VECTOR): Likewise.
24413 * fold-const.c (test_vec_duplicate_folding): Likewise.
24414 * genrecog.c (validate_pattern): Likewise.
24415 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24416 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24417 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24418 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24419 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24420 * rtlanal.c (subreg_get_info): Likewise.
24421 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24422 (vect_grouped_load_supported): Likewise.
24423 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24424 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24425 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24426 (simplify_const_unary_operation, simplify_binary_operation_1)
24427 (simplify_const_binary_operation, simplify_ternary_operation)
24428 (test_vector_ops_duplicate, test_vector_ops): Likewise.
24429 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
24430 instead of CONST_VECTOR_NUNITS.
24431 * varasm.c (output_constant_pool_2): Likewise.
24432 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
24433 explicit-encoded elements in the XVEC for variable-length vectors.
24435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24437 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
24439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24440 Alan Hayward <alan.hayward@arm.com>
24441 David Sherwood <david.sherwood@arm.com>
24443 * coretypes.h (fixed_size_mode): Declare.
24444 (fixed_size_mode_pod): New typedef.
24445 * builtins.h (target_builtins::x_apply_args_mode)
24446 (target_builtins::x_apply_result_mode): Change type to
24447 fixed_size_mode_pod.
24448 * builtins.c (apply_args_size, apply_result_size, result_vector)
24449 (expand_builtin_apply_args_1, expand_builtin_apply)
24450 (expand_builtin_return): Update accordingly.
24452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24454 * cse.c (hash_rtx_cb): Hash only the encoded elements.
24455 * cselib.c (cselib_hash_rtx): Likewise.
24456 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
24457 CONST_VECTOR encoding.
24459 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24460 Jeff Law <law@redhat.com>
24463 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
24464 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
24465 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
24466 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
24469 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
24470 explicitly probe *sp in a noreturn function if there were any callee
24471 register saves or frame pointer is needed.
24473 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24476 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
24477 BLKmode for ternary, binary or unary expressions.
24480 * var-tracking.c (delete_vta_debug_insn): New inline function.
24481 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
24482 insns from get_insns () to NULL instead of each bb separately.
24483 Use delete_vta_debug_insn. No longer static.
24484 (vt_debug_insns_local, variable_tracking_main_1): Adjust
24485 delete_vta_debug_insns callers.
24486 * rtl.h (delete_vta_debug_insns): Declare.
24487 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
24488 instead of variable_tracking_main.
24490 2018-01-03 Martin Sebor <msebor@redhat.com>
24492 PR tree-optimization/83603
24493 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
24494 arguments past the endof the argument list in functions declared
24495 without a prototype.
24496 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
24497 Avoid checking when arguments are null.
24499 2018-01-03 Martin Sebor <msebor@redhat.com>
24502 * doc/extend.texi (attribute const): Fix a typo.
24503 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
24504 issuing -Wsuggest-attribute for void functions.
24506 2018-01-03 Martin Sebor <msebor@redhat.com>
24508 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
24509 offset_int::from instead of wide_int::to_shwi.
24510 (maybe_diag_overlap): Remove assertion.
24511 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
24512 * gimple-ssa-sprintf.c (format_directive): Same.
24513 (parse_directive): Same.
24514 (sprintf_dom_walker::compute_format_length): Same.
24515 (try_substitute_return_value): Same.
24517 2018-01-03 Jeff Law <law@redhat.com>
24519 PR middle-end/83654
24520 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
24521 non-constant residual for zero at runtime and avoid probing in
24522 that case. Reorganize code for trailing problem to mirror handling
24525 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24527 PR tree-optimization/83501
24528 * tree-ssa-strlen.c (get_string_cst): New.
24529 (handle_char_store): Call get_string_cst.
24531 2018-01-03 Martin Liska <mliska@suse.cz>
24533 PR tree-optimization/83593
24534 * tree-ssa-strlen.c: Include tree-cfg.h.
24535 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
24536 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
24537 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
24539 (strlen_dom_walker::before_dom_children): Call
24540 gimple_purge_dead_eh_edges. Dump tranformation with details
24542 (strlen_dom_walker::before_dom_children): Update call by adding
24543 new argument cleanup_eh.
24544 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
24546 2018-01-03 Martin Liska <mliska@suse.cz>
24549 * cif-code.def (VARIADIC_THUNK): New enum value.
24550 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
24553 2018-01-03 Jan Beulich <jbeulich@suse.com>
24555 * sse.md (mov<mode>_internal): Tighten condition for when to use
24556 vmovdqu<ssescalarsize> for TI and OI modes.
24558 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24560 Update copyright years.
24562 2018-01-03 Martin Liska <mliska@suse.cz>
24565 * ipa-visibility.c (function_and_variable_visibility): Skip
24566 functions with noipa attribure.
24568 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24570 * gcc.c (process_command): Update copyright notice dates.
24571 * gcov-dump.c (print_version): Ditto.
24572 * gcov.c (print_version): Ditto.
24573 * gcov-tool.c (print_version): Ditto.
24574 * gengtype.c (create_file): Ditto.
24575 * doc/cpp.texi: Bump @copying's copyright year.
24576 * doc/cppinternals.texi: Ditto.
24577 * doc/gcc.texi: Ditto.
24578 * doc/gccint.texi: Ditto.
24579 * doc/gcov.texi: Ditto.
24580 * doc/install.texi: Ditto.
24581 * doc/invoke.texi: Ditto.
24583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24585 * vector-builder.h (vector_builder::m_full_nelts): Change from
24586 unsigned int to poly_uint64.
24587 (vector_builder::full_nelts): Update prototype accordingly.
24588 (vector_builder::new_vector): Likewise.
24589 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
24590 (vector_builder::operator ==): Likewise.
24591 (vector_builder::finalize): Likewise.
24592 * int-vector-builder.h (int_vector_builder::int_vector_builder):
24593 Take the number of elements as a poly_uint64 rather than an
24595 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
24596 from unsigned int to poly_uint64.
24597 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
24598 (vec_perm_indices::new_vector): Likewise.
24599 (vec_perm_indices::length): Likewise.
24600 (vec_perm_indices::nelts_per_input): Likewise.
24601 (vec_perm_indices::input_nelts): Likewise.
24602 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24603 number of elements per input as a poly_uint64 rather than an
24604 unsigned int. Use the original encoding for variable-length
24605 vectors, rather than clamping each individual element.
24606 For the second and subsequent elements in each pattern,
24607 clamp the step and base before clamping their sum.
24608 (vec_perm_indices::series_p): Handle polynomial element counts.
24609 (vec_perm_indices::all_in_range_p): Likewise.
24610 (vec_perm_indices_to_tree): Likewise.
24611 (vec_perm_indices_to_rtx): Likewise.
24612 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24613 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24614 (tree_vector_builder::new_binary_operation): Handle polynomial
24615 element counts. Return false if we need to know the number
24616 of elements at compile time.
24617 * fold-const.c (fold_vec_perm): Punt if the number of elements
24618 isn't known at compile time.
24620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24622 * vec-perm-indices.h (vec_perm_builder): Change element type
24623 from HOST_WIDE_INT to poly_int64.
24624 (vec_perm_indices::element_type): Update accordingly.
24625 (vec_perm_indices::clamp): Handle polynomial element_types.
24626 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24627 (vec_perm_indices::all_in_range_p): Likewise.
24628 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24630 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24631 polynomial vec_perm_indices element types.
24632 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24633 * fold-const.c (fold_vec_perm): Likewise.
24634 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24635 * tree-vect-generic.c (lower_vec_perm): Likewise.
24636 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24637 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24638 element type to HOST_WIDE_INT.
24640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24641 Alan Hayward <alan.hayward@arm.com>
24642 David Sherwood <david.sherwood@arm.com>
24644 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24645 rather than an int. Use plus_constant.
24646 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24647 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24650 Alan Hayward <alan.hayward@arm.com>
24651 David Sherwood <david.sherwood@arm.com>
24653 * calls.c (emit_call_1, expand_call): Change struct_value_size from
24654 a HOST_WIDE_INT to a poly_int64.
24656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24657 Alan Hayward <alan.hayward@arm.com>
24658 David Sherwood <david.sherwood@arm.com>
24660 * calls.c (load_register_parameters): Cope with polynomial
24661 mode sizes. Require a constant size for BLKmode parameters
24662 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
24663 forces a parameter to be padded at the lsb end in order to
24664 fill a complete number of words, require the parameter size
24665 to be ordered wrt UNITS_PER_WORD.
24667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24668 Alan Hayward <alan.hayward@arm.com>
24669 David Sherwood <david.sherwood@arm.com>
24671 * reload1.c (spill_stack_slot_width): Change element type
24672 from unsigned int to poly_uint64_pod.
24673 (alter_reg): Treat mode sizes as polynomial.
24675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24676 Alan Hayward <alan.hayward@arm.com>
24677 David Sherwood <david.sherwood@arm.com>
24679 * reload.c (complex_word_subreg_p): New function.
24680 (reload_inner_reg_of_subreg, push_reload): Use it.
24682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24683 Alan Hayward <alan.hayward@arm.com>
24684 David Sherwood <david.sherwood@arm.com>
24686 * lra-constraints.c (process_alt_operands): Reject matched
24687 operands whose sizes aren't ordered.
24688 (match_reload): Refer to this check here.
24690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24691 Alan Hayward <alan.hayward@arm.com>
24692 David Sherwood <david.sherwood@arm.com>
24694 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24695 that the mode size is in the set {1, 2, 4, 8, 16}.
24697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24698 Alan Hayward <alan.hayward@arm.com>
24699 David Sherwood <david.sherwood@arm.com>
24701 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24702 Use plus_constant instead of gen_rtx_PLUS.
24704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24705 Alan Hayward <alan.hayward@arm.com>
24706 David Sherwood <david.sherwood@arm.com>
24708 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24709 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24710 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24711 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24712 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24713 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24714 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24715 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24716 * config/i386/i386.c (ix86_push_rounding): ...this new function.
24717 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24719 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24720 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24721 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24722 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24723 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24724 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24725 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24726 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24727 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24728 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24730 * expr.c (emit_move_resolve_push): Treat the input and result
24731 of PUSH_ROUNDING as a poly_int64.
24732 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
24733 (emit_push_insn): Likewise.
24734 * lra-eliminations.c (mark_not_eliminable): Likewise.
24735 * recog.c (push_operand): Likewise.
24736 * reload1.c (elimination_effects): Likewise.
24737 * rtlanal.c (nonzero_bits1): Likewise.
24738 * calls.c (store_one_arg): Likewise. Require the padding to be
24739 known at compile time.
24741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24742 Alan Hayward <alan.hayward@arm.com>
24743 David Sherwood <david.sherwood@arm.com>
24745 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
24746 Use plus_constant instead of gen_rtx_PLUS.
24748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24749 Alan Hayward <alan.hayward@arm.com>
24750 David Sherwood <david.sherwood@arm.com>
24752 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
24753 rather than an int.
24755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24756 Alan Hayward <alan.hayward@arm.com>
24757 David Sherwood <david.sherwood@arm.com>
24759 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
24760 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
24761 via stack temporaries. Treat the mode size as polynomial too.
24763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24764 Alan Hayward <alan.hayward@arm.com>
24765 David Sherwood <david.sherwood@arm.com>
24767 * expr.c (expand_expr_real_2): When handling conversions involving
24768 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
24769 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
24770 as a poly_uint64 too.
24772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24773 Alan Hayward <alan.hayward@arm.com>
24774 David Sherwood <david.sherwood@arm.com>
24776 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
24778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24779 Alan Hayward <alan.hayward@arm.com>
24780 David Sherwood <david.sherwood@arm.com>
24782 * combine.c (can_change_dest_mode): Handle polynomial
24783 REGMODE_NATURAL_SIZE.
24784 * expmed.c (store_bit_field_1): Likewise.
24785 * expr.c (store_constructor): Likewise.
24786 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
24787 and polynomial REGMODE_NATURAL_SIZE.
24788 (gen_lowpart_common): Likewise.
24789 * reginfo.c (record_subregs_of_mode): Likewise.
24790 * rtlanal.c (read_modify_subreg_p): Likewise.
24792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24793 Alan Hayward <alan.hayward@arm.com>
24794 David Sherwood <david.sherwood@arm.com>
24796 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
24797 numbers of elements.
24799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24800 Alan Hayward <alan.hayward@arm.com>
24801 David Sherwood <david.sherwood@arm.com>
24803 * match.pd: Cope with polynomial numbers of vector elements.
24805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24806 Alan Hayward <alan.hayward@arm.com>
24807 David Sherwood <david.sherwood@arm.com>
24809 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
24810 in a POINTER_PLUS_EXPR.
24812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24813 Alan Hayward <alan.hayward@arm.com>
24814 David Sherwood <david.sherwood@arm.com>
24816 * omp-simd-clone.c (simd_clone_subparts): New function.
24817 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
24818 (ipa_simd_modify_function_body): Likewise.
24820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24821 Alan Hayward <alan.hayward@arm.com>
24822 David Sherwood <david.sherwood@arm.com>
24824 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
24825 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
24826 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
24827 (expand_vector_condition, vector_element): Likewise.
24828 (subparts_gt): New function.
24829 (get_compute_type): Use subparts_gt.
24830 (count_type_subparts): Delete.
24831 (expand_vector_operations_1): Use subparts_gt instead of
24832 count_type_subparts.
24834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24835 Alan Hayward <alan.hayward@arm.com>
24836 David Sherwood <david.sherwood@arm.com>
24838 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
24839 (vect_compile_time_alias): ...this new function. Do the calculation
24840 on poly_ints rather than trees.
24841 (vect_prune_runtime_alias_test_list): Update call accordingly.
24843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24844 Alan Hayward <alan.hayward@arm.com>
24845 David Sherwood <david.sherwood@arm.com>
24847 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
24849 (vect_schedule_slp_instance): Likewise.
24851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24852 Alan Hayward <alan.hayward@arm.com>
24853 David Sherwood <david.sherwood@arm.com>
24855 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
24856 constant and extern definitions for variable-length vectors.
24857 (vect_get_constant_vectors): Note that the number of units
24858 is known to be constant.
24860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24861 Alan Hayward <alan.hayward@arm.com>
24862 David Sherwood <david.sherwood@arm.com>
24864 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
24865 of units as polynomial. Choose between WIDE and NARROW based
24868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24869 Alan Hayward <alan.hayward@arm.com>
24870 David Sherwood <david.sherwood@arm.com>
24872 * tree-vect-stmts.c (simd_clone_subparts): New function.
24873 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
24875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24876 Alan Hayward <alan.hayward@arm.com>
24877 David Sherwood <david.sherwood@arm.com>
24879 * tree-vect-stmts.c (vectorizable_call): Treat the number of
24880 vectors as polynomial. Use build_index_vector for
24881 IFN_GOMP_SIMD_LANE.
24883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24884 Alan Hayward <alan.hayward@arm.com>
24885 David Sherwood <david.sherwood@arm.com>
24887 * tree-vect-stmts.c (get_load_store_type): Treat the number of
24888 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
24889 for variable-length vectors.
24890 (vectorizable_mask_load_store): Treat the number of units as
24891 polynomial, asserting that it is constant if the condition has
24892 already been enforced.
24893 (vectorizable_store, vectorizable_load): Likewise.
24895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24896 Alan Hayward <alan.hayward@arm.com>
24897 David Sherwood <david.sherwood@arm.com>
24899 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
24900 of units as polynomial. Punt if we can't tell at compile time
24901 which vector contains the final result.
24903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24904 Alan Hayward <alan.hayward@arm.com>
24905 David Sherwood <david.sherwood@arm.com>
24907 * tree-vect-loop.c (vectorizable_induction): Treat the number
24908 of units as polynomial. Punt on SLP inductions. Use an integer
24909 VEC_SERIES_EXPR for variable-length integer reductions. Use a
24910 cast of such a series for variable-length floating-point
24913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24914 Alan Hayward <alan.hayward@arm.com>
24915 David Sherwood <david.sherwood@arm.com>
24917 * tree.h (build_index_vector): Declare.
24918 * tree.c (build_index_vector): New function.
24919 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
24920 of units as polynomial, forcibly converting it to a constant if
24921 vectorizable_reduction has already enforced the condition.
24922 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
24923 to create a {1,2,3,...} vector.
24924 (vectorizable_reduction): Treat the number of units as polynomial.
24925 Choose vectype_in based on the largest scalar element size rather
24926 than the smallest number of units. Enforce the restrictions
24929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24930 Alan Hayward <alan.hayward@arm.com>
24931 David Sherwood <david.sherwood@arm.com>
24933 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
24934 number of units as polynomial.
24936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24937 Alan Hayward <alan.hayward@arm.com>
24938 David Sherwood <david.sherwood@arm.com>
24940 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
24941 * target.def (autovectorize_vector_sizes): Return the vector sizes
24942 by pointer, using vector_sizes rather than a bitmask.
24943 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
24944 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
24945 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
24947 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
24948 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
24949 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
24950 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
24951 * omp-general.c (omp_max_vf): Likewise.
24952 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24953 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
24954 * tree-vect-loop.c (vect_analyze_loop): Likewise.
24955 * tree-vect-slp.c (vect_slp_bb): Likewise.
24956 * doc/tm.texi: Regenerate.
24957 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
24959 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
24960 the vector size as a poly_uint64 rather than an unsigned int.
24961 (current_vector_size): Change from an unsigned int to a poly_uint64.
24962 (get_vectype_for_scalar_type): Update accordingly.
24963 * tree.h (build_truth_vector_type): Take the size and number of
24964 units as a poly_uint64 rather than an unsigned int.
24965 (build_vector_type): Add a temporary overload that takes
24966 the number of units as a poly_uint64 rather than an unsigned int.
24967 * tree.c (make_vector_type): Likewise.
24968 (build_truth_vector_type): Take the number of units as a poly_uint64
24969 rather than an unsigned int.
24971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24972 Alan Hayward <alan.hayward@arm.com>
24973 David Sherwood <david.sherwood@arm.com>
24975 * target.def (get_mask_mode): Take the number of units and length
24976 as poly_uint64s rather than unsigned ints.
24977 * targhooks.h (default_get_mask_mode): Update accordingly.
24978 * targhooks.c (default_get_mask_mode): Likewise.
24979 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
24980 * doc/tm.texi: Regenerate.
24982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24983 Alan Hayward <alan.hayward@arm.com>
24984 David Sherwood <david.sherwood@arm.com>
24986 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
24987 * omp-general.c (omp_max_vf): Likewise.
24988 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
24989 (expand_omp_simd): Handle polynomial safelen.
24990 * omp-low.c (omplow_simd_context): Add a default constructor.
24991 (omplow_simd_context::max_vf): Change from int to poly_uint64.
24992 (lower_rec_simd_input_clauses): Update accordingly.
24993 (lower_rec_input_clauses): Likewise.
24995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24996 Alan Hayward <alan.hayward@arm.com>
24997 David Sherwood <david.sherwood@arm.com>
24999 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25000 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25001 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25002 (vect_analyze_slp_cost): Likewise.
25003 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25004 (vect_model_load_cost): Likewise.
25006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25007 Alan Hayward <alan.hayward@arm.com>
25008 David Sherwood <david.sherwood@arm.com>
25010 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25011 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25012 from an unsigned int * to a poly_uint64_pod *.
25013 (calculate_unrolling_factor): New function.
25014 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25017 Alan Hayward <alan.hayward@arm.com>
25018 David Sherwood <david.sherwood@arm.com>
25020 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25021 from an unsigned int to a poly_uint64.
25022 (_loop_vec_info::slp_unrolling_factor): Likewise.
25023 (_loop_vec_info::vectorization_factor): Change from an int
25025 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25026 (vect_get_num_vectors): New function.
25027 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25028 (vect_get_num_copies): Use vect_get_num_vectors.
25029 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25030 to an unsigned int *.
25031 (vect_analyze_data_refs): Change min_vf from an int * to a
25033 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25034 than an unsigned HOST_WIDE_INT.
25035 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25036 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25037 to an unsigned int *.
25038 (vect_analyze_data_ref_dependences): Likewise.
25039 (vect_compute_data_ref_alignment): Handle polynomial vf.
25040 (vect_enhance_data_refs_alignment): Likewise.
25041 (vect_prune_runtime_alias_test_list): Likewise.
25042 (vect_shift_permute_load_chain): Likewise.
25043 (vect_supportable_dr_alignment): Likewise.
25044 (dependence_distance_ge_vf): Take the vectorization factor as a
25045 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25046 (vect_analyze_data_refs): Change min_vf from an int * to a
25048 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25049 vfm1 as a poly_uint64 rather than an int. Make the same change
25050 for the returned bound_scalar.
25051 (vect_gen_vector_loop_niters): Handle polynomial vf.
25052 (vect_do_peeling): Likewise. Update call to
25053 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25054 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25056 * tree-vect-loop.c (vect_determine_vectorization_factor)
25057 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25058 (vect_get_known_peeling_cost): Likewise.
25059 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25060 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25061 (vect_transform_loop): Likewise. Use the lowest possible VF when
25062 updating the upper bounds of the loop.
25063 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25064 rather than an int.
25065 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25066 polynomial unroll factors.
25067 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25068 (vect_make_slp_decision): Likewise.
25069 (vect_supported_load_permutation_p): Likewise, and polynomial
25071 (vect_analyze_slp_cost): Handle polynomial vf.
25072 (vect_slp_analyze_node_operations): Likewise.
25073 (vect_slp_analyze_bb_1): Likewise.
25074 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25075 than an unsigned HOST_WIDE_INT.
25076 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25077 (vectorizable_load): Handle polynomial vf.
25078 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25080 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25083 Alan Hayward <alan.hayward@arm.com>
25084 David Sherwood <david.sherwood@arm.com>
25086 * match.pd: Handle bit operations involving three constants
25087 and try to fold one pair.
25089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25091 * tree-vect-loop-manip.c: Include gimple-fold.h.
25092 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25093 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25094 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25095 Add a path that uses a step of VF instead of 1, but disable it
25097 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25098 and niters_no_overflow parameters. Update calls to
25099 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25100 Create a new SSA name if the latter choses to use a ste other
25101 than zero, and return it via niters_vector_mult_vf_var.
25102 * tree-vect-loop.c (vect_transform_loop): Update calls to
25103 vect_do_peeling, vect_gen_vector_loop_niters and
25104 slpeel_make_loop_iterate_ntimes.
25105 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25106 (vect_gen_vector_loop_niters): Update declarations after above changes.
25108 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25110 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25111 128-bit round to integer instructions.
25112 (ceil<mode>2): Likewise.
25113 (btrunc<mode>2): Likewise.
25114 (round<mode>2): Likewise.
25116 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25118 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25119 unaligned VSX load/store on P8/P9.
25120 (expand_block_clear): Allow the use of unaligned VSX
25121 load/store on P8/P9.
25123 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25125 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25127 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25128 swap associated with both a load and a store.
25130 2018-01-02 Andrew Waterman <andrew@sifive.com>
25132 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25133 * config/riscv/riscv.md (clear_cache): Use it.
25135 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25137 * web.c: Remove out-of-date comment.
25139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25141 * expr.c (fixup_args_size_notes): Check that any existing
25142 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25143 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25144 (emit_single_push_insn): ...here.
25146 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25148 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25149 (const_vector_encoded_nelts): New function.
25150 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25151 (const_vector_int_elt, const_vector_elt): Declare.
25152 * emit-rtl.c (const_vector_int_elt_1): New function.
25153 (const_vector_elt): Likewise.
25154 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25155 of CONST_VECTOR_ELT.
25157 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25159 * expr.c: Include rtx-vector-builder.h.
25160 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25161 directly on the tree encoding.
25162 (const_vector_from_tree): Likewise.
25163 * optabs.c: Include rtx-vector-builder.h.
25164 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25165 sequence of "u" values.
25166 * vec-perm-indices.c: Include rtx-vector-builder.h.
25167 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25168 directly on the vec_perm_indices encoding.
25170 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25172 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25173 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25174 * rtx-vector-builder.h: New file.
25175 * rtx-vector-builder.c: Likewise.
25176 * rtl.h (rtx_def::u2): Add a const_vector field.
25177 (CONST_VECTOR_NPATTERNS): New macro.
25178 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25179 (CONST_VECTOR_DUPLICATE_P): Likewise.
25180 (CONST_VECTOR_STEPPED_P): Likewise.
25181 (CONST_VECTOR_ENCODED_ELT): Likewise.
25182 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25183 (unwrap_const_vec_duplicate): Likewise.
25184 (const_vec_series_p): Check for a non-duplicated vector encoding.
25185 Say that the function only returns true for integer vectors.
25186 * emit-rtl.c: Include rtx-vector-builder.h.
25187 (gen_const_vec_duplicate_1): Delete.
25188 (gen_const_vector): Call gen_const_vec_duplicate instead of
25189 gen_const_vec_duplicate_1.
25190 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25191 (gen_const_vec_duplicate): Use rtx_vector_builder.
25192 (gen_const_vec_series): Likewise.
25193 (gen_rtx_CONST_VECTOR): Likewise.
25194 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25195 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25196 Build a new vector rather than modifying a CONST_VECTOR in-place.
25197 (handle_special_swappables): Update call accordingly.
25198 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25199 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25200 Build a new vector rather than modifying a CONST_VECTOR in-place.
25201 (handle_special_swappables): Update call accordingly.
25203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25205 * simplify-rtx.c (simplify_const_binary_operation): Use
25206 CONST_VECTOR_ELT instead of XVECEXP.
25208 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25210 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25211 the selector elements to be different from the data elements
25212 if the selector is a VECTOR_CST.
25213 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25214 ssizetype for the selector.
25216 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25218 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25219 before testing each element individually.
25220 * tree-vect-generic.c (lower_vec_perm): Likewise.
25222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25224 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25225 * selftest-run-tests.c (selftest::run_tests): Call it.
25226 * vector-builder.h (vector_builder::operator ==): New function.
25227 (vector_builder::operator !=): Likewise.
25228 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25229 (vec_perm_indices::all_from_input_p): New function.
25230 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25231 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25232 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25233 instead of reading the VECTOR_CST directly. Detect whether both
25234 vector inputs are the same before constructing the vec_perm_indices,
25235 and update the number of inputs argument accordingly. Use the
25236 utility functions added above. Only construct sel2 if we need to.
25238 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25240 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25241 the broadcast of the low byte.
25242 (expand_mult_highpart): Use an explicit encoding for the permutes.
25243 * optabs-query.c (can_mult_highpart_p): Likewise.
25244 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25245 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25246 (vectorizable_bswap): Likewise.
25247 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25248 explicit encoding for the power-of-2 permutes.
25249 (vect_permute_store_chain): Likewise.
25250 (vect_grouped_load_supported): Likewise.
25251 (vect_permute_load_chain): Likewise.
25253 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25255 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25256 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25257 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25258 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25259 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25260 (vect_gen_perm_mask_any): Likewise.
25262 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25264 * int-vector-builder.h: New file.
25265 * vec-perm-indices.h: Include int-vector-builder.h.
25266 (vec_perm_indices): Redefine as an int_vector_builder.
25267 (auto_vec_perm_indices): Delete.
25268 (vec_perm_builder): Redefine as a stand-alone class.
25269 (vec_perm_indices::vec_perm_indices): New function.
25270 (vec_perm_indices::clamp): Likewise.
25271 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25272 (vec_perm_indices::new_vector): New function.
25273 (vec_perm_indices::new_expanded_vector): Update for new
25274 vec_perm_indices class.
25275 (vec_perm_indices::rotate_inputs): New function.
25276 (vec_perm_indices::all_in_range_p): Operate directly on the
25277 encoded form, without computing elided elements.
25278 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25279 encoding. Update for new vec_perm_indices class.
25280 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25281 the given vec_perm_builder.
25282 (expand_vec_perm_var): Update vec_perm_builder constructor.
25283 (expand_mult_highpart): Use vec_perm_builder instead of
25284 auto_vec_perm_indices.
25285 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25286 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25287 or double series encoding as appropriate.
25288 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25289 vec_perm_indices instead of auto_vec_perm_indices.
25290 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25291 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25292 (vect_permute_store_chain): Likewise.
25293 (vect_grouped_load_supported): Likewise.
25294 (vect_permute_load_chain): Likewise.
25295 (vect_shift_permute_load_chain): Likewise.
25296 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25297 (vect_transform_slp_perm_load): Likewise.
25298 (vect_schedule_slp_instance): Likewise.
25299 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25300 (vectorizable_mask_load_store): Likewise.
25301 (vectorizable_bswap): Likewise.
25302 (vectorizable_store): Likewise.
25303 (vectorizable_load): Likewise.
25304 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25305 vec_perm_indices instead of auto_vec_perm_indices. Use
25306 tree_to_vec_perm_builder to read the vector from a tree.
25307 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25308 vec_perm_builder instead of a vec_perm_indices.
25309 (have_whole_vector_shift): Use vec_perm_builder and
25310 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25311 truncation to calc_vec_perm_mask_for_shift.
25312 (vect_create_epilog_for_reduction): Likewise.
25313 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25314 from auto_vec_perm_indices to vec_perm_indices.
25315 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25316 instead of changing individual elements.
25317 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25318 the vector in d.perm.
25319 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25320 from auto_vec_perm_indices to vec_perm_indices.
25321 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25322 instead of changing individual elements.
25323 (arm_vectorize_vec_perm_const): Use new_vector to install
25324 the vector in d.perm.
25325 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25326 Update vec_perm_builder constructor.
25327 (rs6000_expand_interleave): Likewise.
25328 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25329 (rs6000_expand_interleave): Likewise.
25331 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25333 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25334 to qimode could truncate the indices.
25335 * optabs.c (expand_vec_perm_var): Likewise.
25337 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25339 * Makefile.in (OBJS): Add vec-perm-indices.o.
25340 * vec-perm-indices.h: New file.
25341 * vec-perm-indices.c: Likewise.
25342 * target.h (vec_perm_indices): Replace with a forward class
25344 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25345 * optabs.h: Include vec-perm-indices.h.
25346 (expand_vec_perm): Delete.
25347 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25348 (expand_vec_perm_const): Declare.
25349 * target.def (vec_perm_const_ok): Replace with...
25350 (vec_perm_const): ...this new hook.
25351 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25352 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25353 * doc/tm.texi: Regenerate.
25354 * optabs.def (vec_perm_const): Delete.
25355 * doc/md.texi (vec_perm_const): Likewise.
25356 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25357 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25358 expand_vec_perm for constant permutation vectors. Assert that
25359 the mode of variable permutation vectors is the integer equivalent
25360 of the mode that is being permuted.
25361 * optabs-query.h (selector_fits_mode_p): Declare.
25362 * optabs-query.c: Include vec-perm-indices.h.
25363 (selector_fits_mode_p): New function.
25364 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25365 is defined, instead of checking whether the vec_perm_const_optab
25366 exists. Use targetm.vectorize.vec_perm_const instead of
25367 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25368 fit in the vector mode before using a variable permute.
25369 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25370 vec_perm_indices instead of an rtx.
25371 (expand_vec_perm): Replace with...
25372 (expand_vec_perm_const): ...this new function. Take the selector
25373 as a vec_perm_indices rather than an rtx. Also take the mode of
25374 the selector. Update call to shift_amt_for_vec_perm_mask.
25375 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25376 Use vec_perm_indices::new_expanded_vector to expand the original
25377 selector into bytes. Check whether the indices fit in the vector
25378 mode before using a variable permute.
25379 (expand_vec_perm_var): Make global.
25380 (expand_mult_highpart): Use expand_vec_perm_const.
25381 * fold-const.c: Includes vec-perm-indices.h.
25382 * tree-ssa-forwprop.c: Likewise.
25383 * tree-vect-data-refs.c: Likewise.
25384 * tree-vect-generic.c: Likewise.
25385 * tree-vect-loop.c: Likewise.
25386 * tree-vect-slp.c: Likewise.
25387 * tree-vect-stmts.c: Likewise.
25388 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25390 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25391 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25392 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25393 (aarch64_vectorize_vec_perm_const): ...this new function.
25394 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25395 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25396 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25397 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25398 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25399 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25400 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25402 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25403 check for NEON modes.
25404 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25405 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25406 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25407 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25409 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25410 the old VEC_PERM_CONST conditions.
25411 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25412 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25413 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25414 (ia64_vectorize_vec_perm_const_ok): Merge into...
25415 (ia64_vectorize_vec_perm_const): ...this new function.
25416 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25417 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25418 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25419 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25420 * config/mips/mips.c (mips_expand_vec_perm_const)
25421 (mips_vectorize_vec_perm_const_ok): Merge into...
25422 (mips_vectorize_vec_perm_const): ...this new function.
25423 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25424 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25425 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25426 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25427 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25428 (rs6000_expand_vec_perm_const): Delete.
25429 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
25431 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25432 (altivec_expand_vec_perm_const_le): Take each operand individually.
25433 Operate on constant selectors rather than rtxes.
25434 (altivec_expand_vec_perm_const): Likewise. Update call to
25435 altivec_expand_vec_perm_const_le.
25436 (rs6000_expand_vec_perm_const): Delete.
25437 (rs6000_vectorize_vec_perm_const_ok): Delete.
25438 (rs6000_vectorize_vec_perm_const): New function.
25439 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25440 an element count and rtx array.
25441 (rs6000_expand_extract_even): Update call accordingly.
25442 (rs6000_expand_interleave): Likewise.
25443 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
25444 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
25445 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
25446 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
25447 (rs6000_expand_vec_perm_const): Delete.
25448 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25449 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25450 (altivec_expand_vec_perm_const_le): Take each operand individually.
25451 Operate on constant selectors rather than rtxes.
25452 (altivec_expand_vec_perm_const): Likewise. Update call to
25453 altivec_expand_vec_perm_const_le.
25454 (rs6000_expand_vec_perm_const): Delete.
25455 (rs6000_vectorize_vec_perm_const_ok): Delete.
25456 (rs6000_vectorize_vec_perm_const): New function. Remove stray
25457 reference to the SPE evmerge intructions.
25458 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25459 an element count and rtx array.
25460 (rs6000_expand_extract_even): Update call accordingly.
25461 (rs6000_expand_interleave): Likewise.
25462 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
25463 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
25465 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25467 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25469 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
25470 vector mode and that that mode matches the mode of the data
25472 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
25473 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
25474 directly using expand_vec_perm_1 when forcing selectors into
25476 (expand_vec_perm_var): New function, split out from expand_vec_perm.
25478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25480 * optabs-query.h (can_vec_perm_p): Delete.
25481 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
25482 * optabs-query.c (can_vec_perm_p): Split into...
25483 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
25484 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
25485 particular selector is valid.
25486 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25487 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25488 (vect_grouped_load_supported): Likewise.
25489 (vect_shift_permute_load_chain): Likewise.
25490 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25491 (vect_transform_slp_perm_load): Likewise.
25492 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25493 (vectorizable_bswap): Likewise.
25494 (vect_gen_perm_mask_checked): Likewise.
25495 * fold-const.c (fold_ternary_loc): Likewise. Don't take
25496 implementations of variable permutation vectors into account
25497 when deciding which selector to use.
25498 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
25499 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
25500 with a false third argument.
25501 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
25502 to test whether the constant selector is valid and can_vec_perm_var_p
25503 to test whether a variable selector is valid.
25505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25507 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
25508 * optabs-query.c (can_vec_perm_p): Likewise.
25509 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
25510 instead of vec_perm_indices.
25511 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
25512 (vect_gen_perm_mask_checked): Likewise,
25513 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
25514 (vect_gen_perm_mask_checked): Likewise,
25516 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25518 * optabs-query.h (qimode_for_vec_perm): Declare.
25519 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
25520 (qimode_for_vec_perm): ...this new function.
25521 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
25523 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25525 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
25526 does not have a conditional at the top.
25528 2018-01-02 Richard Biener <rguenther@suse.de>
25530 * ipa-inline.c (big_speedup_p): Fix expression.
25532 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25535 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
25538 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25542 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
25543 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
25544 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
25545 cond_taken_branch_cost 3->4.
25547 2018-01-01 Jakub Jelinek <jakub@redhat.com>
25549 PR tree-optimization/83581
25550 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
25551 TODO_cleanup_cfg if any changes have been made.
25553 PR middle-end/83608
25554 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
25555 convert_modes if target mode has the right side, but different mode
25558 PR middle-end/83609
25559 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
25560 last argument when extracting from CONCAT. If either from_real or
25561 from_imag is NULL, use expansion through memory. If result is not
25562 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
25563 the parts directly to inner mode, if even that fails, use expansion
25566 PR middle-end/83623
25567 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
25568 check for bswap in mode rather than HImode and use that in expand_unop
25571 Copyright (C) 2018 Free Software Foundation, Inc.
25573 Copying and distribution of this file, with or without modification,
25574 are permitted in any medium without royalty provided the copyright
25575 notice and this notice are preserved.