1 2018-05-16 Mark Wielaard <mark@klomp.org>
3 * dwarf2out.c (count_index_strings): New function.
4 (output_indirect_strings): Call count_index_strings and generate
5 header for dwarf_version >= 5.
7 2018-05-16 Mark Wielaard <mark@klomp.org>
9 * dwarf2out.c (dwarf_FORM): New function.
10 (set_indirect_string): Use dwarf_FORM.
11 (reset_indirect_string): Likewise.
12 (size_of_die): Likewise.
13 (value_format): Likewise.
14 (output_die): Likewise.
15 (add_skeleton_AT_string): Likewise.
16 (output_macinfo_op): Likewise.
17 (index_string): Likewise.
18 (output_index_string_offset): Likewise.
19 (output_index_string): Likewise.
20 (count_index_strings): Likewise.
22 2018-05-16 Carl Love <cel@us.ibm.com>
24 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
25 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
27 2018-05-16 Martin Jambor <mjambor@suse.cz>
29 * ipa-prop.c (ipa_free_all_edge_args): Remove.
30 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
32 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
34 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
35 (fnma<mode>4): Likewise.
36 (fms<mode>4): Likewise.
37 (fnms<mode>4): Likewise.
38 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
39 (aarch64_fnma<mode>4): Likewise.
40 (aarch64_fms<mode>4): Likewise.
41 (aarch64_fnms<mode>4): Likewise.
42 (aarch64_fnmadd<mode>4): Likewise.
44 2018-05-16 Jason Merrill <jason@redhat.com>
46 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
48 2018-05-16 Richard Biener <rguenther@suse.de>
50 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
51 (dump_stmt_cost): Declare.
52 (add_stmt_cost): Dump cost we add.
53 (add_stmt_costs): New function.
54 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
56 (vect_analyze_stmt): Adjust prototype.
57 (vectorizable_condition): Likewise.
58 (vectorizable_live_operation): Likewise.
59 (vectorizable_reduction): Likewise.
60 (vectorizable_induction): Likewise.
61 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
62 cost vector to pass to vectorizable_ and record afterwards.
63 (vect_model_reduction_cost): Take cost vector argument and adjust.
64 (vect_model_induction_cost): Likewise.
65 (vectorizable_reduction): Likewise.
66 (vectorizable_induction): Likewise.
67 (vectorizable_live_operation): Likewise.
68 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
69 SLP_TREE_NUMBER_OF_VEC_STMTS.
70 (vect_analyze_slp_cost_1): Remove.
71 (vect_analyze_slp_cost): Likewise.
72 (vect_slp_analyze_node_operations): Take visited args and
73 a target cost vector. Avoid processing already visited stmt sets.
74 (vect_slp_analyze_operations): Use a local cost vector to gather
75 costs and register those of non-discarded instances.
76 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
77 (vect_schedule_slp_instance): Remove copying of
78 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
80 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
81 adding cost. Record cost entry location.
82 (vect_prologue_cost_for_slp_op): Function to compute cost of
83 a constant or invariant generated for SLP vect in the prologue,
84 split out from vect_analyze_slp_cost_1.
85 (vect_model_simple_cost): Make static. Adjust for SLP costing.
86 (vect_model_promotion_demotion_cost): Likewise.
87 (vect_model_store_cost): Likewise, make static.
88 (vect_model_load_cost): Likewise.
89 (vectorizable_bswap): Add cost vector arg and adjust.
90 (vectorizable_call): Likewise.
91 (vectorizable_simd_clone_call): Likewise.
92 (vectorizable_conversion): Likewise.
93 (vectorizable_assignment): Likewise.
94 (vectorizable_shift): Likewise.
95 (vectorizable_operation): Likewise.
96 (vectorizable_store): Likewise.
97 (vectorizable_load): Likewise.
98 (vectorizable_condition): Likewise.
99 (vectorizable_comparison): Likewise.
100 (can_vectorize_live_stmts): Likewise.
101 (vect_analyze_stmt): Likewise.
102 (vect_transform_stmt): Adjust calls to vectorizable_*.
103 * tree-vectorizer.c: Include gimple-pretty-print.h.
104 (dump_stmt_cost): New function.
106 2018-05-16 Richard Biener <rguenther@suse.de>
108 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
109 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
110 * tree-ssa-dse.c: Include tree-ssa-loop.h.
111 (check_name): New callback.
112 (dse_classify_store): Track cycles via a visited bitmap of PHI
113 defs and simplify handling of in-loop and across loop dead stores
114 and properly fail for loop-variant refs. Handle byte-tracking with
115 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
118 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
120 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
121 (vect_get_mask_type_for_stmt): Likewise.
122 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
124 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
125 to determine the statement's vector type and the vector type that
126 should be used for calculating nunits. Deal with cases in which
127 the type has to be deferred.
128 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
129 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
130 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
131 (vect_determine_vf_for_stmt): New functions, split out from...
132 (vect_determine_vectorization_factor): ...here.
133 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
134 (vect_get_mask_type_for_stmt): New functions, split out from
135 vect_determine_vectorization_factor.
137 2018-05-16 Richard Biener <rguenther@suse.de>
139 * tree-cfg.c (verify_gimple_assign_ternary): Properly
140 verify the [VEC_]COND_EXPR embedded comparison.
142 2018-05-15 Martin Sebor <msebor@redhat.com>
144 PR tree-optimization/85753
145 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
146 RECORD_TYPE in addition to ARRAY_TYPE.
148 2018-05-15 Martin Sebor <msebor@redhat.com>
151 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
153 2018-05-15 Richard Biener <rguenther@suse.de>
155 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
156 add by_clobber_p one. Change algorithm to collect all defs
157 representing uses we need to walk and try reducing them to
158 a single one before failing.
159 (dse_dom_walker::dse_optimize_stmt): Adjust.
161 2018-05-13 Mark Wielaard <mark@klomp.org>
163 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
164 (size_of_loc_descr): Likewise.
165 (output_loc_operands): Likewise.
166 (output_loc_operands_raw): Likewise.
167 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
168 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
169 (hash_loc_operands): Likewise.
170 (compare_loc_operands): Likewise.
172 2018-05-14 Mark Wielaard <mark@klomp.org>
174 * dwarf2out.c (count_index_addrs): New function.
175 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
177 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
179 PR tree-optimization/83648
180 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
181 return value as malloc candidate.
183 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
186 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
187 as true in call to suggest_attribute.
189 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
191 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
194 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
196 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
198 (aarch64_crypto_aesd_fused): Likewise.
200 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
202 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
203 (movsi_aarch64): Likewise.
204 (load_pairsi): Likewise.
205 (load_pairdi): Likewise.
206 (store_pairsi): Likewise.
207 (store_pairdi): Likewise.
208 (load_pairsf): Likewise.
209 (load_pairdf): Likewise.
210 (store_pairsf): Likewise.
211 (store_pairdf): Likewise.
212 (zero_extend): Likewise.
213 (trunc): Swap alternatives.
214 (fcvt_target): Add '?' to prefer w over r.
216 2018-05-14 Jakub Jelinek <jakub@redhat.com>
219 * config/i386/i386.md: Disallow non-commutative arithmetics in
220 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
221 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
222 in the peephole2 before it.
224 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
226 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
227 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
228 (ix86_handle_option): Handle -mcldemote.
229 * config.gcc: New header.
230 * config/i386/cldemoteintrin.h: New file.
231 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
232 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
234 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
235 OPTION_MASK_ISA_CLDEMOTE.
236 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
237 (ix86_valid_target_attribute_inner_p): Ditto.
238 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
239 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
240 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
241 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
242 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
244 * config/i386/i386.opt: Add -mcldemote.
245 * config/i386/x86intrin.h: New header.
246 * doc/invoke.texi: Add -mcldemote.
248 2018-05-14 Richard Biener <rguenther@suse.de>
250 * doc/match-and-simplify.texi: Adjust :s documentation.
252 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
254 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
255 intended memcpy size.
256 (REORDER_45): Likewise.
258 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
261 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
262 * vec.c (qsort_chk): Use gcc_qsort.
263 * Makefile.in (OBJS-libcommon): Add sort.o.
264 (build/sort.o): New target. Use it...
265 (BUILD_RTL): ... here, and...
266 (build/gencfn-macros): ... here, and...
267 (build/genmatch): ... here.
269 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
270 Chung-Ju Wu <jasonwucj@gmail.com>
272 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
273 * config/nds32/nds32-graywolf.md: New file.
274 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
275 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
277 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
278 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
279 * config/nds32/nds32.md (pipeline_model): Add graywolf.
280 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
281 * config/nds32/pipelines.md: Include n15 settings.
283 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
284 Chung-Ju Wu <jasonwucj@gmail.com>
286 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
287 * config/nds32/nds32-n13.md: New file.
288 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
289 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
291 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
292 * config/nds32/nds32.md (pipeline_model): Add n13.
293 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
294 * config/nds32/pipelines.md: Include n13 settings.
296 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
297 Chung-Ju Wu <jasonwucj@gmail.com>
299 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
300 * config/nds32/nds32-n10.md: New file.
301 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
302 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
304 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
305 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
306 * config/nds32/nds32.md (pipeline_model): Add n10.
307 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
308 * config/nds32/pipelines.md: Include n10 settings.
310 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
311 Kito Cheng <kito.cheng@gmail.com>
312 Chung-Ju Wu <jasonwucj@gmail.com>
314 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
315 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
316 Add enum values for DSP extension instructions.
317 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
319 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
320 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
322 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
323 * config/nds32/nds32-dspext.md: New file for DSP implementation.
324 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
325 * config/nds32/nds32-intrinsic.md: Likewise.
326 * config/nds32/nds32_intrinsic.h: Likewise.
327 * config/nds32/nds32-md-auxiliary.c: Likewise.
328 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
329 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
330 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
331 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
332 * config/nds32/nds32-protos.h: New declarations for DSP extension.
333 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
334 TYPE_DMAC in switch statement.
335 * config/nds32/nds32.c: New checking and implementation for DSP
336 extension instructions.
337 * config/nds32/nds32.h: Likewise.
338 * config/nds32/nds32.md: Likewise.
339 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
340 * config/nds32/predicates.md: Implement new predicates for DSP
343 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
345 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
346 Reformat alternatives and attributes so it is easier to identify
347 which constraints/attributes go with which instruction.
348 (mov<mode>_hardfloat32, FMOVE64): Likewise.
349 (mov<mode>_softfloat32, FMOVE64): Likewise.
350 (mov<mode>_hardfloat64, FMOVE64): Likewise.
351 (mov<mode>_softfloat64, FMOVE64): Likewise.
353 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
355 * doc/extend.texi (PowerPC Built-in Functions): Rename this
357 (Basic PowerPC Built-in Functions): The new name of the
358 subsection previously known as "PowerPC Built-in Functions".
359 (Basic PowerPC Built-in Functions Available on all Configurations):
361 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
362 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
363 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
364 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
366 2018-05-11 Martin Jambor <mjambor@suse.cz>
369 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
372 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
375 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
377 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
379 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
380 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
381 (ix86_handle_option): Handle -mwaitpkg.
382 * config.gcc: New header.
383 * config/i386/cpuid.h (bit_WAITPKG): New bit.
384 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
385 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
387 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
388 OPTION_MASK_ISA_WAITPKG.
389 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
390 (ix86_option_override_internal): Add PTA_WAITPKG.
391 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
392 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
394 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
395 __builtin_ia32_umwait and __builtin_ia32_tpause.
396 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
397 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
398 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
399 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
400 UNSPECV_TPAUSE): New.
401 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
402 * config/i386/i386.opt: Add -mwaitpkg.
403 * config/i386/waitpkgintrin.h: New file.
404 * config/i386/x86intrin.h: New header.
405 * doc/invoke.texi: Add -mwaitpkg.
407 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
410 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
412 (cortex-m0): Use armv6s-m isa.
413 (cortex-m0plus): Likewise.
414 (cortex-m1): Likewise.
415 (cortex-m0.small-multiply): Likewise.
416 (cortex-m0plus.small-multiply): Likewise.
417 (cortex-m1.small-multiply): Likewise.
419 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
420 Jakub Jelinek <jakub@redhat.com>
422 PR tree-optimization/85692
423 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
424 source permute as well.
426 2018-05-11 Martin Liska <mliska@suse.cz>
429 * doc/extend.texi: Document LLVM style format for no_sanitize
432 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
434 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
435 mode_supports_vsx_dform_quad to mode_supports_dq_form.
436 (mode_supports_vsx_dform_quad): Likewise.
437 (mode_supports_vmx_dform): Move these functions to be next to the
438 other mode_supports functions.
439 (mode_supports_dq_form): Likewise.
440 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
441 mode_supports_dq_form.
442 (reg_offset_addressing_ok_p): Likewise.
443 (offsettable_ok_by_alignment): Likewise.
444 (rs6000_legitimate_offset_address_p): Likewise.
445 (legitimate_lo_sum_address_p): Likewise.
446 (rs6000_legitimize_address): Likewise.
447 (rs6000_legitimize_reload_address): Likewise.
448 (rs6000_secondary_reload_inner): Likewise.
449 (rs6000_preferred_reload_class): Likewise.
450 (rs6000_output_move_128bit): Likewise.
452 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
454 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
455 Generate SImode target register for null target.
456 <case IX86_BUILTIN_XGETBV>: Ditto.
457 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
458 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
460 2018-05-10 Carl Love <cel@us.ibm.com>
462 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
463 dcbtt and dcbtstt if operands[2] is 0.
465 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
468 * config/i386/sse.md (usadv64qi): New expander.
470 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
472 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
473 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
474 -maltivec=be support.
475 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
476 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
477 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
478 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
479 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
480 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
481 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
482 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
483 altivec_vsumsws): Adjust.
484 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
485 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
486 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
488 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
489 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
490 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
491 (altivec_lve<VI_char>x): Delete expand.
492 (*altivec_lve<VI_char>x_internal): Rename to...
493 (altivec_lve<VI_char>x): ... this.
494 (altivec_lvxl_<mode>): Delete expand.
495 (*altivec_lvxl_<mode>_internal): Rename to ...
496 (altivec_lvxl_<mode>): ... this.
497 (altivec_stvxl_<mode>): Delete expand.
498 (*altivec_stvxl_<mode>_internal): Rename to ...
499 (altivec_stvxl_<mode>): ... this.
500 (altivec_stve<VI_char>x): Delete expand.
501 (*altivec_stve<VI_char>x_internal): Rename to ...
502 (altivec_stve<VI_char>x): ... this.
503 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
504 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
505 reduc_plus_scal_<mode>): Adjust.
506 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
508 (rs6000_cpu_cpp_builtins): Adjust.
509 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
510 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
511 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
512 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
513 -maltivec=be support.
514 (rs6000_split_vec_extract_var): Adjust.
515 (rs6000_split_v4si_init): Adjust.
516 (swap_selector_for_mode): Delete.
517 (altivec_expand_lvx_be, altivec_expand_stvx_be,
518 altivec_expand_stvex_be): Delete.
519 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
520 -maltivec=be support.
521 (rs6000_gimple_fold_builtin): Ditto.
522 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
524 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
525 (TARGET_DIRECT_MOVE_64BIT): Adjust.
526 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
527 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
528 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
529 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
530 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
531 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
532 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
533 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
534 anonymous split): Adjust.
535 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
536 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
538 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
540 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
541 when --with-gxx-include-dir is also specified.
542 * configure: Regenerate.
544 2018-05-09 Jim Wilson <jimw@sifive.com>
547 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
548 * config/riscv/t-withmultilib: New.
549 * config/riscv/withmultilib.h: New.
550 * doc/install.texi: Document RISC-V --with-multilib-list support.
552 2018-05-09 Richard Biener <rguenther@suse.de>
554 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
556 (vect_bb_vectorization_profitable_p): Adjust. Compute
557 actual scalar cost using the cost vector and the add_stmt_cost
560 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
562 PR rtl-optimization/85645
563 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
564 in the REG_CFA_REGISTER note for LR, don't leave it empty.
566 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
568 PR rtl-optimization/85645
569 * shrink-wrap.c (spread_components): Return a boolean saying if
570 anything was changed.
571 (try_shrink_wrapping_separate): Iterate spread_components until
572 nothing changes anymore.
574 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
576 PR rtl-optimization/85645
577 * regrename.c (build_def_use): Also kill the chains that include the
578 destination of a REG_CFA_REGISTER note.
580 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
582 PR rtl-optimization/85645
583 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
584 insn that has a REG_CFA_REGISTER note.
586 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
588 * cfgexpand.c (expand_clobber): New function.
589 (expand_gimple_stmt_1): Use it.
590 * tree-vect-stmts.c (vect_clobber_variable): New function,
592 (vectorizable_simd_clone_call): ...here.
593 (vectorizable_store): Emit a clobber either side of an
594 IFN_STORE_LANES sequence.
595 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
597 2018-05-09 Tom de Vries <tom@codesourcery.com>
600 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
601 (define_insn "trap_if_false"): Add exit after trap.
603 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
605 PR rtl-optimization/85638
606 * bb-reorder.c: Include common/common-target.h.
607 (create_forwarder_block): New function extracted from...
608 (fix_up_crossing_landing_pad): ...here. Rename into...
609 (dw2_fix_up_crossing_landing_pad): ...this.
610 (sjlj_fix_up_crossing_landing_pad): New function.
611 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
612 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
613 from both partitions and exit the loop after one iteration.
615 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
618 * doc/extend.texi (PowerPC Built-in Functions): Rename this
620 (Basic PowerPC Built-in Functions): The new name of the
621 subsection previously known as "PowerPC Built-in Functions".
622 (Basic PowerPC Built-in Functions Available on all Configurations):
624 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
626 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
628 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
630 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
633 2018-05-08 Jim Wilson <jimw@sifive.com>
635 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
636 (LD_EMUL_SUFFIX): New.
639 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
641 * doc/extend.texi (PowerPC Built-in Functions): Rename this
643 (Basic PowerPC Built-in Functions): The new name of the
644 subsection previously known as "PowerPC Built-in Functions".
645 (Basic PowerPC Built-in Functions Available on all Configurations):
647 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
649 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
651 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
653 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
656 2018-05-08 Jakub Jelinek <jakub@redhat.com>
659 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
660 after cmpelim optimization.
662 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
664 * config.gcc: Support "goldmont".
665 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
666 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
668 * config/i386/i386.c (m_GOLDMONT): Define.
669 (processor_target_table): Add "goldmont".
670 (PTA_GOLDMONT): Define.
671 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
672 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
673 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
674 (fold_builtin_cpu): Add "goldmont".
675 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
676 (ix86_option_override_internal): Add "goldmont".
677 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
678 (processor_type): Add PROCESSOR_GOLDMONT.
679 * config/i386/i386.md: Add CPU "glm".
680 * config/i386/glm.md: New file.
681 * config/i386/x86-tune.def: Add m_GOLDMONT.
682 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
684 2018-05-08 Jakub Jelinek <jakub@redhat.com>
687 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
689 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
690 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
691 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
694 * config/i386/i386.c (ix86_fold_builtin): Handle
695 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
698 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
699 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
701 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
704 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
705 (check_arch): Likewise.
706 (check_fpu): Return the result rather than printing it.
707 (end arch): Fix operator precedence.
709 (END): Print the result from check_fpu.
711 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
712 Alan Hayward <alan.hayward@arm.com>
713 David Sherwood <david.sherwood@arm.com>
715 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
716 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
717 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
718 (*fcmuo<mode>_and): New patterns.
720 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
722 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
723 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
724 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
725 (cmp_op, sve_imm_con): New code attributes.
726 (SVE_COND_INT_CMP, imm_con): Delete.
727 (cmp_op): Remove above unspecs from int attribute.
728 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
730 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
731 comparison-specific unspecs.
732 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
733 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
734 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
735 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
736 (*vec_fcm<cmp_op><mode>): Rename to...
737 (*fcm<cmp_op><mode>): ...this and adjust likewise.
738 (*vec_fcmuo<mode>): Rename to...
739 (*fcmuo<mode>): ...this and adjust likewise.
740 (*pred_fcm<cmp_op><mode>): New pattern.
741 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
742 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
744 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
746 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
747 (aarch64_emit_sve_predicated_cond): New function.
748 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
749 (aarch64_emit_unspec_cond_or): Replace with...
750 (aarch64_emit_sve_or_conds): ...this new function. Use
751 aarch64_emit_sve_ptrue_op for the individual comparisons and
752 aarch64_emit_binop to OR them together.
753 (aarch64_emit_inverted_unspec_cond): Replace with...
754 (aarch64_emit_sve_inverted_cond): ...this new function. Use
755 aarch64_emit_sve_ptrue_op for the comparison and
756 aarch64_emit_unop to invert the result.
757 (aarch64_expand_sve_vec_cmp_float): Update after the above
758 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
760 2018-05-07 Nathan Sidwell <nathan@acm.org>
762 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
763 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
764 (Backwards Compatibility): Likewise.
766 2018-05-07 Luis Machado <luis.machado@linaro.org>
770 2018-05-07 Luis Machado <luis.machado@linaro.org>
772 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
773 <prefetch_dynamic_strides>: New const bool field.
774 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
775 prefetch_dynamic_strides.
776 (exynosm1_prefetch_tune): Likewise.
777 (thunderxt88_prefetch_tune): Likewise.
778 (thunderx_prefetch_tune): Likewise.
779 (thunderx2t99_prefetch_tune): Likewise.
780 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
781 (aarch64_override_options_internal): Update to set
782 PARAM_PREFETCH_DYNAMIC_STRIDES.
783 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
784 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
785 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
786 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
787 prefetch-dynamic-strides setting.
789 2018-05-07 Luis Machado <luis.machado@linaro.org>
791 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
792 <minimum_stride>: New const int field.
793 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
794 minimum_stride field.
795 (exynosm1_prefetch_tune): Likewise.
796 (thunderxt88_prefetch_tune): Likewise.
797 (thunderx_prefetch_tune): Likewise.
798 (thunderx2t99_prefetch_tune): Likewise.
799 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
800 (aarch64_override_options_internal): Update to set
801 PARAM_PREFETCH_MINIMUM_STRIDE.
802 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
803 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
804 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
805 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
806 stride is constant and is below the minimum stride threshold.
808 2018-05-07 Luis Machado <luis.machado@linaro.org>
810 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
813 2018-05-07 Luis Machado <luis.machado@linaro.org>
815 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
816 <prefetch_dynamic_strides>: New const bool field.
817 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
818 prefetch_dynamic_strides.
819 (exynosm1_prefetch_tune): Likewise.
820 (thunderxt88_prefetch_tune): Likewise.
821 (thunderx_prefetch_tune): Likewise.
822 (thunderx2t99_prefetch_tune): Likewise.
823 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
824 (aarch64_override_options_internal): Update to set
825 PARAM_PREFETCH_DYNAMIC_STRIDES.
826 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
827 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
828 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
829 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
830 prefetch-dynamic-strides setting.
832 2018-05-07 Luis Machado <luis.machado@linaro.org>
834 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
835 <minimum_stride>: New const int field.
836 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
837 minimum_stride field.
838 (exynosm1_prefetch_tune): Likewise.
839 (thunderxt88_prefetch_tune): Likewise.
840 (thunderx_prefetch_tune): Likewise.
841 (thunderx2t99_prefetch_tune): Likewise.
842 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
843 (aarch64_override_options_internal): Update to set
844 PARAM_PREFETCH_MINIMUM_STRIDE.
845 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
846 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
847 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
848 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
849 stride is constant and is below the minimum stride threshold.
851 2018-05-06 Jakub Jelinek <jakub@redhat.com>
854 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
855 the type is addressable. Don't force op into register if it has
858 2018-05-05 Roland McGrath <mcgrathr@google.com>
861 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
862 any section for which we don't know a specific type it should have,
863 regardless of name. Previously this was done only for the exact
864 names ".init_array", ".fini_array", and ".preinit_array".
865 (default_elf_asm_named_section): Add comment about
866 relationship with default_section_type_flags and SECTION_NOTYPE.
867 (get_section): Don't consider it a type conflict if one side has
868 SECTION_NOTYPE and the other doesn't, as long as neither has the
869 SECTION_BSS et al used in the default_section_type_flags logic.
871 2018-05-05 Tom de Vries <tom@codesourcery.com>
874 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
875 (workaround_barsyncs): New function.
876 (nvptx_reorg): Use workaround_barsyncs.
877 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
878 (define_expand "nvptx_membar_cta"): New define_expand.
879 (define_insn "*nvptx_membar_cta"): New insn.
881 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
883 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
884 To improve optimization opportunities.
885 * builtin-types.def: The new needed builtin types for the above.
887 2018-05-04 Richard Biener <rguenther@suse.de>
889 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
890 * gimple-ssa-store-merging.c
891 (imm_store_chain_info::output_merged_store): Remove redundant create,
892 release split_store vector contents on failure.
893 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
894 scalar stmt vector on cache hit.
896 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
898 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
900 * config.gcc (powerpc-xilinx-eabi*): Remove.
901 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
903 (fusion_addis_mem_combo_load): Ditto.
904 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
906 (rs6000_cpu_cpp_builtins): Ditto.
907 * config/rs6000/rs6000-linux.c
908 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
909 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
910 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
912 (rs6000_setup_reg_addr_masks): Ditto.
913 (rs6000_init_hard_regno_mode_ok): Ditto.
914 (rs6000_option_override_internal): Ditto.
915 (legitimate_lo_sum_address_p): Ditto.
916 (rs6000_legitimize_address): Ditto.
917 (rs6000_legitimize_reload_address): Ditto.
918 (rs6000_legitimate_address_p): Ditto.
919 (abi_v4_pass_in_fpr): Ditto.
920 (setup_incoming_varargs): Ditto.
921 (rs6000_gimplify_va_arg): Ditto.
922 (rs6000_split_multireg_move): Ditto.
923 (rs6000_savres_strategy): Ditto.
924 (rs6000_emit_prologue_components): Ditto.
925 (rs6000_emit_epilogue_components): Ditto.
926 (rs6000_emit_prologue): Ditto.
927 (rs6000_emit_epilogue): Ditto.
928 (rs6000_elf_file_end): Ditto.
929 (rs6000_function_value): Ditto.
930 (rs6000_libcall_value): Ditto.
931 * config/rs6000/rs6000.h: Ditto.
932 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
933 (TARGET_MINMAX): ... this. New.
934 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
935 * config/rs6000/rs6000.md: Remove Xilinx FP support.
936 (*movsi_internal1_single): Delete.
937 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
938 mfpu=, mxilinx-fpu): Delete.
939 * config/rs6000/singlefp.h: Delete.
940 * config/rs6000/sysv4.h: Remove Xilinx FP support.
941 * config/rs6000/t-rs6000: Ditto.
942 * config/rs6000/t-xilinx: Delete.
943 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
944 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
945 (VStype_simple): Delete.
946 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
947 * config/rs6000/xfpu.h: Delete.
948 * config/rs6000/xfpu.md: Delete.
949 * config/rs6000/xilinx.h: Delete.
950 * config/rs6000/xilinx.opt: Delete.
951 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
952 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
954 2018-05-04 Tom de Vries <tom@codesourcery.com>
957 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
960 2018-05-04 Richard Biener <rguenther@suse.de>
963 * tree-complex.c (update_complex_assignment): We are always in SSA form.
964 (expand_complex_div_wide): Likewise.
965 (expand_complex_operations_1): Likewise.
966 (expand_complex_libcall): Preserve EH info of the original stmt.
967 (tree_lower_complex): Handle removed blocks.
968 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
969 on complex multiplication and division libcall builtins.
971 2018-05-04 Richard Biener <rguenther@suse.de>
974 * fold-const.c (negate_expr_p): Restrict negation of operand
975 zero of a division to when we know that can happen without
977 (fold_negate_expr_1): Likewise.
979 2018-05-04 Jakub Jelinek <jakub@redhat.com>
982 * real.h (real_nextafter): Declare.
983 * real.c (real_nextafter): New function.
984 * fold-const-call.c (fold_const_nextafter): New function.
985 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
987 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
988 even when arg1_mode is different from arg0_mode.
990 2018-05-03 Nathan Sidwell <nathan@acm.org>
992 * doc/extend.texi (Deprecated Features): Remove
994 (Backwards Compatibility): Likewise.
995 * doc/invoke.texi (C++ Language Options): Likewise.
996 (C++ Dialect Options): Likewise.
998 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1001 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1002 _mm512_mask_mullox_epi64): New intrinsics.
1004 2018-05-03 Tom de Vries <tom@codesourcery.com>
1007 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1008 dump files): Add offload-tree.
1010 2018-05-03 Richard Biener <rguenther@suse.de>
1012 PR tree-optimization/85615
1013 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1014 to loops not nested in BBs loop father to avoid creating multi-entry
1017 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1019 PR tree-optimization/70291
1020 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1021 arguments. Change return type to tree. Emit libcall as a new
1022 statement rather than replacing existing one when inplace_p is true.
1023 (expand_complex_multiplication_components): New function.
1024 (expand_complex_multiplication): Expand floating-point complex
1025 multiplication using the above.
1026 (expand_complex_division): Rename inner_type parameter to type.
1027 Update expand_complex_libcall call-site.
1028 (expand_complex_operations_1): Update expand_complex_multiplication
1029 and expand_complex_division call-sites.
1031 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1034 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1035 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1036 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1037 the highest significant bit of the shift count mask is clear. In
1038 check whether and[sq]i3 is needed verify that all significant bits
1039 of the shift count other than the highest are set.
1041 2018-05-02 Tom de Vries <tom@codesourcery.com>
1044 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1045 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1046 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1047 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1048 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1049 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1050 __builtin_goacc_parlevel_size.
1052 2018-05-02 Richard Biener <rguenther@suse.de>
1054 PR tree-optimization/85597
1055 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1056 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1059 2018-05-02 Tom de Vries <tom@codesourcery.com>
1062 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1063 dump files): Add ltrans-tree.
1065 2018-05-02 Tom de Vries <tom@codesourcery.com>
1068 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1069 dump files): Add wpa-ipa.
1071 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1073 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1074 powerpc*-*-linux*paired* target.
1075 * config/rs6000/750cl.h: Delete.
1076 * config/rs6000/paired.h: Delete.
1077 * config/rs6000/paired.md: Delete.
1078 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1080 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1081 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1082 comment. Remove paired float support.
1083 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1084 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1086 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1087 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1089 * config/rs6000/rs6000.c: Remove paired float support.
1090 (paired_expand_vector_init, paired_expand_vector_move,
1091 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1092 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1093 paired_expand_builtin, paired_expand_predicate_builtin,
1094 paired_init_builtins): Delete.
1095 * config/rs6000/rs6000.h: Remove paired float support.
1096 * config/rs6000/rs6000.md: Remove paired float support.
1097 (move_from_CR_ov_bit): Delete.
1098 * config/rs6000/rs6000.opt (mpaired): Delete.
1099 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1100 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1102 2018-05-02 Richard Biener <rguenther@suse.de>
1105 * gimplify.c (gimplify_save_expr): When in SSA form allow
1106 SAVE_EXPRs to compute to SSA vars.
1108 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1111 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1112 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1113 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1114 clobber operands[2], instead use a new pseudo. Formatting fixes.
1116 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1118 PR tree-optimization/85586
1119 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1120 exit early for statements in the same group if the accesses are
1123 2018-05-02 Tom de Vries <tom@codesourcery.com>
1126 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1129 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1131 PR tree-optimization/85143
1132 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1134 2018-05-01 Tom de Vries <tom@codesourcery.com>
1137 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1138 not found" error message.
1140 2018-05-01 Tom de Vries <tom@codesourcery.com>
1143 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1144 * vec.c (test_ordered_remove_if): New function.
1145 (vec_c_tests): Call test_ordered_remove_if.
1146 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1147 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1148 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1149 VEC_ORDERED_REMOVE_IF.
1151 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1153 PR tree-optimization/82665
1154 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1155 pointer subtraction where arguments come from a memchr call.
1157 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1159 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1160 --push-state --as-needed and --pop-state instead of --as-needed and
1161 --no-as-needed if ld supports it.
1162 * configure: Regenerated.
1165 * doc/install.texi2html: Replace _002d with - and _002a with * in
1166 generated html files using sed.
1168 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1171 * gcc-rich-location.c (blank_line_before_p): New function.
1172 (use_new_line): New function.
1173 (gcc_rich_location::add_fixit_insert_formatted): New function.
1174 * gcc-rich-location.h
1175 (gcc_rich_location::add_fixit_insert_formatted): New function.
1177 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1179 * selftest.c (assert_streq): Rename "expected" and "actual" to
1180 "val1" and "val2". Extend NULL-handling to cover both inputs
1181 symmetrically, while still requiring both to be non-NULL for a pass.
1182 * selftest.h (assert_streq): Rename "expected" and "actual" to
1184 (ASSERT_EQ): Likewise.
1185 (ASSERT_EQ_AT): Likewise.
1186 (ASSERT_KNOWN_EQ): Likewise.
1187 (ASSERT_KNOWN_EQ_AT): Likewise.
1188 (ASSERT_NE): Likewise.
1189 (ASSERT_MAYBE_NE): Likewise.
1190 (ASSERT_MAYBE_NE_AT): Likewise.
1191 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1192 the assertion to pass.
1193 (ASSERT_STREQ_AT): Likewise.
1195 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1197 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1198 interaction with -pie.
1200 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1202 * selftest.h: Fix alphabetization of per-source-file selftest
1205 2018-04-30 Jason Merrill <jason@redhat.com>
1207 PR c++/61982 - dead stores to destroyed objects.
1208 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1211 2018-04-30 Jason Merrill <jason@redhat.com>
1213 * tree.c (build_clobber): New.
1214 * tree.h: Declare it.
1215 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1217 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1219 * diagnostic-show-locus.c (layout::layout): Update for
1220 location_get_source_line returning a char_span.
1221 (struct char_span): Move to input.h.
1222 (struct correction): Update for fields in char_span becoming
1224 (struct source_line): Update for location_get_source_line
1225 returning a char_span.
1226 (layout::print_line): Likewise.
1227 * edit-context.c (edited_file::print_content): Likewise.
1228 (edited_file::print_diff_hunk): Likewise.
1229 (edited_file::print_run_of_changed_lines): Likewise.
1230 (edited_file::get_num_lines): Likewise.
1231 (edited_line::edited_line): Likewise.
1232 * final.c (asm_show_source): Likewise.
1233 * input.c (location_get_source_line): Convert return type
1234 from const char * to char_span, losing the final "line_len"
1236 (dump_location_info): Update for the above.
1237 (get_substring_ranges_for_loc): Likewise. Use a char_span
1238 when handling the literal within the line.
1239 (test_reading_source_line): Update for location_get_source_line
1240 returning a char_span.
1241 * input.h (class char_span): Move here from
1242 diagnostic-show-locus.c, converting from a struct to a class.
1243 Make data members private.
1244 (char_span::operator bool): New.
1245 (char_span::length): New.
1246 (char_span::get_buffer): New.
1247 (char_span::operator[]): New.
1248 (char_span::subspan): Make const.
1249 (char_span::xstrdup): New.
1250 (location_get_source_line): Convert return type from const char *
1251 to char_span, losing the final "line_size" param.
1253 2018-04-30 Jan Hubicka <jh@suse.cz>
1255 * lto-wrapper.c (ltrans_priorities): New static var.
1256 (cmp_priority): New.
1257 (run_gcc): Read priorities and if doing parallel build order
1258 the Makefile by them.
1260 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1262 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1264 2018-04-30 Richard Biener <rguenther@suse.de>
1266 * tree-cfg.c (verify_address): Remove base argument, add
1267 flag whether to check TREE_ADDRESSABLE and do that.
1268 (verify_expr): Remove.
1269 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1270 (verify_gimple_assign_single): Likewise.
1271 (verify_gimple_switch): Likewise.
1272 (verify_expr_location_1): Dereference tp once. Add (disabled)
1273 piece from verify_expr.
1274 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1276 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1278 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1280 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1282 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1283 (small_data_pattern): Likewise.
1284 (arc_rewrite_small_data): Likewise.
1285 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1286 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1287 (get_symbol_alignment): New function.
1288 (legitimate_small_data_address_p): Likewise.
1289 (legitimate_scaled_address): Update, call
1290 legitimate_small_data_address_p.
1291 (output_sdata): New static variable.
1292 (arc_print_operand): Update how we handle small data operands.
1293 (arc_print_operand_address): Likewise.
1294 (arc_legitimate_address_p): Update, use
1295 legitimate_small_data_address_p.
1296 (arc_rewrite_small_data_p): Remove.
1297 (arc_rewrite_small_data_1): Likewise.
1298 (arc_rewrite_small_data): Likewise.
1299 (small_data_pattern): Likewise.
1300 (compact_sda_memory_operand): Update to use
1301 legitimate_small_data_address_p and get_symbol_alignment.
1302 (prepare_move_operands): Don't rewite sdata pattern.
1303 (prepare_extend_operands): Remove.
1304 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1306 (zero_extendqisi2): Likewise.
1307 (zero_extendhisi2): Likewise.
1308 (extendqihi2): Likewise.
1309 (extendqisi2): Likewise.
1310 (extendhisi2): Likewise.
1314 * config/arc/constraints.md (Usd): Change it to memory constraint.
1316 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1318 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1319 as source of std instructions.
1320 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1321 allow 6-bit constants as source for store instructions.
1322 (movdi_insn): Update instruction pattern to allow 6-bit constants
1323 as source for store instructions.
1325 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1327 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1329 2018-04-30 Nathan Sidwell <nathan@acm.org>
1330 Sandra Loosemore <sandra@codesourcery.com>
1332 * dumpfile.c (dump_open): Allow '-' for stdout.
1333 * doc/invoke.texi (Developer Options): Document dump filename
1334 determination early. Document stdin/stdout selection.
1336 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1338 Microblaze Target: PIC data text relative
1340 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1341 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1343 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1344 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1345 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1346 New addressing mode for data-text relative position indepenedent code.
1347 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1348 'ADDRESS_SYMBOLIC_TXT_REL'.
1349 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1350 (microblaze_legitimate_pic_operand): Exclude function calls from
1351 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1352 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1354 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1355 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1356 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1357 for 'address + offset'.
1358 (microblaze_expand_prologue): Add new function prologue call for
1360 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1361 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1362 table in case of TARGET_PIC_DATA_TEXT_REL.
1363 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1364 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1365 Add new macros 'UNSPEC_TEXT',
1366 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1367 + exclude function calls from 'UNSPEC_PLT' in case of data text
1369 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1370 new target hook for generating address diff vector tables in case of
1372 * doc/tm.texi : Regenerate.
1373 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1374 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1375 of addr diff vector generation.
1376 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1377 target hook definition.
1378 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1379 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1380 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1381 option for data text relative.
1383 2018-04-30 Richard Biener <rguenther@suse.de>
1385 * tree-chrec.h (evolution_function_is_constant_p): Remove
1387 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1389 2018-04-30 Richard Biener <rguenther@suse.de>
1392 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1394 2018-04-30 Richard Biener <rguenther@suse.de>
1396 PR tree-optimization/28364
1397 PR tree-optimization/85275
1398 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1399 copying first exit test.
1401 2018-04-28 Mark Wielaard <mark@klomp.org>
1403 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1405 (dwarf_AT): Handle DW_AT_addr_base.
1406 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1408 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1411 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1412 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1413 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1414 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1416 2018-04-28 Richard Biener <rguenther@suse.de>
1418 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1419 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1420 to reflect use. Only add interesting stmts.
1422 2018-04-27 Martin Jambor <mjambor@suse.cz>
1425 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1426 the jump function allows for passing through aggregate values.
1428 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1430 * input.h (in_system_header_at): Convert from macro to inline
1432 (from_macro_expansion_at): Likewise.
1433 (from_macro_definition_at): Likewise.
1435 2018-04-27 Jeff Law <law@redhat.com>
1437 * config.gcc: Mark tile* targets as deprecated/obsolete.
1439 2018-04-27 Richard Biener <rguenther@suse.de>
1441 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1444 2018-04-27 Richard Biener <rguenther@suse.de>
1446 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1448 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1450 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1451 with Yd constraint. Set "preferred_for_speed" attribute from
1452 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1454 (*movdi_internal): Ditto.
1455 (movti_interunit splitters): Remove
1456 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1457 (movdi_interunit splitters): Ditto.
1458 * config/i386/constraints.md (Ye): Remove.
1459 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1461 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1464 * config/aarch64/constraints.md (Usg): Limit to 31.
1467 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1469 PR tree-optimization/85529
1470 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1471 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1472 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1473 zero extension or masking of the MSB bit.
1474 (optimize_range_tests): Add FIRST_BB argument, pass it through
1475 to optimize_range_tests_var_bound.
1476 (maybe_optimize_range_tests, reassociate_bb): Adjust
1477 optimize_range_tests callers.
1479 2018-04-26 Richard Biener <rguenther@suse.de>
1480 Jakub Jelinek <jakub@redhat.com>
1482 * cgraph.h (symbol_table): Just declare debug method here.
1483 * symtab.c (symbol_table::debug): Define.
1485 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1487 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1489 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1491 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1492 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1493 (*movdi_internal): Substitute Yi and Yj constraint with x
1494 and Ym and Yn constraint with y constraint. Update "isa"
1495 attribute and set "preferred_for_speed" attribute from
1496 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1497 (*movsi_internal): Ditto.
1498 (*movdf_internal): Ditto.
1499 (*movsf_internal): Ditto.
1500 (*zero_extendsidi2): Ditto.
1501 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1502 (sse2_loadld): Ditto.
1503 (*vec_extract<ssevecmodelower>_0): Ditto.
1504 (*vec_extractv4si_0_zext_sse4): Ditto.
1505 (vec_concatv2di): Ditto.
1506 (*vec_dup<mode>): Ditto.
1507 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1508 * config/i386/constraints.md (Yi): Remove.
1513 2018-04-26 Nathan Sidwell <nathan@acm.org>
1515 * dumpfile.c (dump_open): New.
1516 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1517 (dump_finish): Detect stdio/stderr by value not name.
1519 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1521 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1523 2018-04-26 Tom de Vries <tom@codesourcery.com>
1526 * config/nvptx/nvptx.c (verify_neutering_jumps)
1527 (verify_neutering_labels): New function
1528 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1530 2018-04-26 Tom de Vries <tom@codesourcery.com>
1533 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1534 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1537 2018-04-26 Richard Biener <rguenther@suse.de>
1538 Tom de Vries <tom@codesourcery.com>
1541 * lto-streamer-out.c (output_function): Fixup loops if required to match
1542 discovery done in the reader.
1544 2018-04-26 Richard Biener <rguenther@suse.de>
1546 PR tree-optimization/85116
1547 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1548 have a loop exit from the single latch predecessor. Remove
1549 case of header with just condition.
1550 (ch_base::copy_headers): Exclude infinite loops from any
1552 (pass_ch::execute): Record exits.
1554 2018-04-26 Richard Biener <rguenther@suse.de>
1556 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1557 prologue cost vector and pass it to vect_get_load_cost.
1558 (vect_get_peeling_costs_all_drs): Likewise.
1559 (vect_peeling_hash_get_lowest_cost): Likewise.
1560 (vect_enhance_data_refs_alignment): Likewise.
1562 2018-04-26 Richard Biener <rguenther@suse.de>
1565 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1566 checking of integer<->pointer conversions.
1567 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1568 sign-/zero-extending pointer types.
1569 (expand_omp_for_static_chunk): Likewise.
1571 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1572 Jean Lee <xiaoyur347@gmail.com>
1574 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1575 (TARGET_ASAN_SHADOW_OFFSET): Define.
1576 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1577 true for -fsanitize=address.
1579 2018-04-25 Mark Wielaard <mark@klomp.org>
1581 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1584 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1586 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1587 than "alu", remove explicit "memory" and "imm_disp" attributes.
1588 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1591 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1592 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1593 gen_lowpart_no_emit.
1595 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1598 * config/i386/i386.c (ix86_expand_builtin): Change memory
1599 operand to XI, extend p0 to Pmode.
1600 * config/i386/i386.md: Change unspec volatile and operand
1601 1 mode to XI, change operand 0 mode to P.
1603 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1605 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1606 GET_MODE_MASK before any checking.
1607 (nds32_can_use_bset_p): Likewise.
1608 (nds32_can_use_btgl_p): Likewise.
1610 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1612 * config/nds32/nds32-doubleword.md: New define_split pattern for
1613 illegal register number.
1615 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1617 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1619 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1621 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1623 2018-04-25 Richard Biener <rguenther@suse.de>
1625 * lto-streamer.h (LTO_major_version): Bump to 8.
1627 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1629 * BASE-VER: Set to 9.0.0.
1631 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1633 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1634 in __abskf2 and __powikf2.
1636 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1639 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1640 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1641 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1642 Use the above on operand 2. Reindent.
1643 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1645 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1648 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1650 * config/i386/i386.opt (mcet): Removed.
1651 * doc/install.texi: Remove -mcet documentation.
1652 * doc/invoke.texi: Likewise.
1654 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1657 * doc/install.texi: Remove -mcet from bootstrap-cet.
1659 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1662 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1663 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1667 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1668 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1669 containing a CONST_VECTOR.
1671 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1673 * doc/install.texi: Update newlib dependency for nvptx.
1675 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1678 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1679 instead of INTVAL when shifting x left.
1681 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1683 PR tree-optimization/85478
1684 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1685 vect_grouped_store_supported for single element vectors.
1687 2018-04-24 Richard Biener <rguenther@suse.de>
1690 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1691 load cost increase to the case of non-constant step.
1693 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1696 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1697 destination if any_malformed_asm.
1699 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1702 * expr.c (store_field): In the bitfield case, if the value comes from
1703 a function call and is returned in registers by means of a PARALLEL,
1704 do not change the mode of the temporary unless BLKmode and VOIDmode.
1706 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1708 PR rtl-optimization/85423
1709 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1710 dependencies to debug insns when the previous insn is non-debug.
1712 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1714 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1715 enums into a single definition.
1716 (fls): Fix predicates and printing.
1719 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1721 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1722 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1723 and short u6 immediate.
1724 (check_if_valid_sleep_operand): Remove.
1725 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1727 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1730 flag_always_save_lp condition.
1731 * config/nds32/nds32.opt (malways-save-lp): New option.
1733 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1735 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1736 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1737 * config/nds32/nds32.h
1738 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1739 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1741 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1743 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1744 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1746 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1747 Chung-Ju Wu <jasonwucj@gmail.com>
1749 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1751 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1752 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1754 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1756 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1758 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1760 * config/nds32/nds32-protos.h (nds32_data_alignment,
1761 nds32_local_alignment): Declare.
1762 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1763 nds32_local_alignment): New functions.
1764 (TARGET_CONSTANT_ALIGNMENT): Define.
1765 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1767 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1769 * config/nds32/nds32.c
1770 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1771 (TARGET_MODES_TIEABLE_P): Likewise.
1773 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1775 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1778 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1779 Chung-Ju Wu <jasonwucj@gmail.com>
1781 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1782 for unaligned access.
1783 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1785 * config/nds32/nds32-intrinsic.md: Likewise.
1786 * config/nds32/nds32_intrinsic.h: Likewise.
1787 * config/nds32/nds32.h (nds32_builtins): Likewise.
1788 * config/nds32/nds32.opt (munaligned-access): New option.
1789 * config/nds32/nds32.c (nds32_asm_file_start): Display
1790 flag_unaligned_access status.
1792 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1794 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1795 -mno-relax is present.
1796 * config/riscv/linux.h (LINK_SPEC): Ditto.
1798 2018-04-20 Martin Sebor <msebor@redhat.com>
1801 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1803 (gimple_fold_builtin_stxcpy_chk): Same.
1804 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1806 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1809 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1810 __powikf2 when long double is IEEE 128-bit.
1812 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1814 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1815 step to make sure stack always aligned.
1817 2018-04-20 Carl Love <cel@us.ibm.com>
1820 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1821 size check for arg0.
1823 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1824 Tom de Vries <tom@codesourcery.com>
1827 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1828 Emit insns for calls too.
1829 (nvptx_find_par): Always look for worker-level predecessor insn.
1830 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1832 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1833 (nvptx_process_pars): Propagate frames for calls.
1835 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1838 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1840 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1841 (ix86_handle_option): Don't handle OPT_mibt.
1842 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1844 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1846 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1847 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1848 (ix86_target_macros): Define __CET__ with flag_cf_protection
1849 for -fcf-protection.
1850 * config/i386/i386.c (isa2_opts): Remove -mibt.
1851 * config/i386/i386.h (TARGET_IBT): Removed.
1852 (TARGET_IBT_P): Likewise.
1853 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1854 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1855 * config/i386/i386.opt (mcet): Update help message.
1858 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1859 -mcet as an alias for -mshstk.
1861 2018-04-20 Richard Biener <rguenther@suse.de>
1864 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1865 complexity by forcing a single use of the multiply operand.
1867 2018-04-20 Martin Jambor <mjambor@suse.cz>
1870 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1871 recursion dependency to only apply to non-clones.
1873 2018-04-20 Martin Jambor <mjambor@suse.cz>
1876 * ipa-cp.c (create_specialized_node): Check that clones of
1877 self-recursive edges exist during IPA-CP.
1879 2018-04-19 Toon Moene <toon@moene.org>
1881 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1884 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1886 PR tree-optimization/85467
1887 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1888 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1889 VECTOR_CST element to type.
1891 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1894 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1895 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1896 (builtin_longjmp): Likewise.
1897 (save_stack_nonlocal): New pattern.
1898 (restore_stack_nonlocal): Likewise.
1900 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1903 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1904 Replace ASM_OUTPUT_LABEL with fprintf.
1906 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1909 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1910 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1911 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1912 define __IBT__ and __SHSTK__ for -fcf-protection.
1913 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1915 (ix86_trampoline_init): Likewise.
1916 (x86_output_mi_thunk): Likewise.
1917 (ix86_notrack_prefixed_insn_p): Likewise.
1918 (ix86_option_override_internal): Don't disallow -fcf-protection.
1919 * config/i386/i386.md (rdssp<mode>): Also enable for
1921 (incssp<mode>): Likewise.
1922 (nop_endbr): Likewise.
1923 * config/i386/i386.opt (mcet): Change help message to built-in
1927 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1928 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1929 enable CET built-in functions.
1931 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1933 * common/config/i386/i386-common.c
1934 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1935 OPTION_MASK_ISA_MOVDIRI_UNSET,
1936 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1937 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1938 * config.gcc (movdirintrin.h): New header.
1939 * config/i386/cpuid.h (bit_MOVDIRI,
1940 bit_MOVDIR64B): New bits.
1941 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1943 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1944 (VOID, PVOID, PCVOID)): New function types.
1945 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1946 __builtin_ia32_directstoreu_u64,
1947 __builtin_ia32_movdir64b): New builtins.
1948 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1949 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1951 (ix86_valid_target_attribute_inner_p): Ditto.
1952 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1953 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1954 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1955 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1956 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1957 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1958 (movdiri<mode>, movdir64b_<mode>): New.
1959 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1960 * config/i386/immintrin.h: Include movdirintrin.h.
1961 * config/i386/movdirintrin.h: New file.
1962 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1964 2018-04-19 Richard Biener <rguenther@suse.de>
1967 * cfg.c (clear_bb_flags): When loop state says we have
1968 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1970 2018-04-19 Richard Biener <rguenther@suse.de>
1972 PR tree-optimization/84737
1973 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1974 copying restrict info.
1975 (vect_setup_realignment): Use it.
1976 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1977 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1978 the first DR to all generated stores.
1979 (vectorizable_load): Likewise for loads.
1981 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1983 PR tree-optimization/85446
1984 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1985 the integral and pointer types to have the same precision.
1987 * doc/install.texi: Document --disable-cet being the default and
1990 2018-04-18 Martin Liska <mliska@suse.cz>
1992 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1995 2018-04-18 Martin Liska <mliska@suse.cz>
1998 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2001 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2002 arguments if they are comparable.
2004 2018-04-18 Martin Liska <mliska@suse.cz>
2007 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2010 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2013 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2016 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2017 ENDBR after calling __morestack.
2019 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2022 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2023 by using gcc_base_ver to generate a gcc_driver_version, and use
2024 it when generating GCC_DRIVER_NAME.
2025 * configure: Regenerate.
2027 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2030 * config.gcc: Obsolete powerpc*-*-*spe*.
2032 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2035 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2036 (stabstr_D): Change type of unum from unsigned int to
2037 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2040 2018-04-17 Jim Wilson <jimw@sifive.com>
2043 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2044 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2045 Set arg_pointer_offset after using pretend_args_size.
2047 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2049 PR rtl-optimization/85431
2050 * dse.c (record_store): Ignore zero width stores.
2053 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2054 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2055 __builtin_stack_restore rather than after it.
2056 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2057 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2058 argument instead of virtual_dynamic_stack_rtx.
2060 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2062 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2064 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2065 Add note to error message to explain internal mapping of overloaded
2066 built-in function name to non-overloaded built-in function name.
2067 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2070 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2073 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2074 where the inputs overlap with the output.
2076 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2078 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2079 (=v, v) alternative and explicit "memory" attribute.
2080 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2081 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2083 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2084 "sselog1" type instead of "sselog".
2085 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2086 "sselog". Remove explicit "memory" attribute.
2087 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2088 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2090 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2091 "sselog1" type instead of "sselog". Remove explicit "memory"
2093 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2094 use "sselog1" type instead of "sselog". Remove explicit "memory"
2096 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2097 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2099 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2100 "sselog1" type instead of "sselog". Remove explicit "memory"
2102 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2103 use "sselog1" type instead of "sselog". Remove explicit "memory"
2107 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2110 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2113 2018-04-17 Martin Jambor <mjambor@suse.cz>
2116 * ipa-cp.c (create_specialized_node): Call
2117 expand_all_artificial_thunks if necessary.
2119 2018-04-17 Martin Liska <mliska@suse.cz>
2122 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2123 in message, remote space in between '_G' and '('.
2125 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2128 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2129 avx512f_vmcmp<mode>3<round_saeonly_name>,
2130 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2131 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2132 avx512f_rndscale<mode><round_saeonly_name>,
2133 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2134 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2135 Use %<iptr>2 instead of %2 for -masm=intel.
2136 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2137 avx512f_vcvttss2usi<round_saeonly_name>,
2138 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2140 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2141 avx512f_vcvttsd2usi<round_saeonly_name>,
2142 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2143 Use %q1 instead of %1 for -masm=intel.
2144 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2145 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2146 of %3 for -masm=intel.
2147 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2149 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2151 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2153 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2154 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2156 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2158 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2159 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2160 %g1 and one with %0 and %1.
2161 (avx512er_vmrcp28<mode><round_saeonly_name>,
2162 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2164 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2165 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2166 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2167 of %0 and %{%4%} for -masm=intel.
2168 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2169 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2170 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2171 order of %0 and %{%5%}%{z%} for -masm=intel.
2173 2018-04-17 Jan Hubicka <jh@suse.cz>
2176 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2178 2018-04-17 Martin Liska <mliska@suse.cz>
2181 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2182 for target_clone error message. Make default implementation
2183 clone to be a local declaration.
2184 (separate_attrs): Add new argument and check for an empty
2186 (expand_target_clones): Handle it.
2187 (ipa_target_clone): Make redirection just for target_clones
2190 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2191 Tom de Vries <tom@codesourcery.com>
2194 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2195 tiled basic blocks without omp continue statements.
2197 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2200 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2201 vec_extract expression as having side effects to make sure it gets
2204 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2207 * config/i386/i386.c (get_builtin_code_for_version): Check
2210 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2213 * gcc/config.gcc: Support "skylake".
2214 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2216 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2217 (processor_target_table): Add "skylake".
2218 (ix86_option_override_internal): Add "skylake".
2219 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2220 PROCESSOR_CANNONLAKE.
2221 (get_builtin_code_for_version): Fix priority for
2222 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2223 PROCESSOR_SKYLAKE-AVX512.
2224 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2225 (processor_type): Add PROCESSOR_SKYLAKE.
2227 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2228 Jason Merrill <jason@redhat.com>
2231 * convert.c (convert_to_integer_1): Use direct recursion for
2232 enumeral types and types with a precision less than the number
2233 of bits in their mode.
2235 2018-04-16 Julia Koval <julia.koval@intel.com>
2238 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2239 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2241 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2244 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2245 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2246 and -mno-direct-move.
2248 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2251 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2252 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2253 Ensure negative shifts result in {0}.
2255 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2257 PR rtl-optimization/79916
2258 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2259 regs (if any) to define how to gnerate SD moves when LRA is in
2262 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2264 PR rtl-optimization/85393
2265 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2266 * except.c (expand_dw2_landing_pad_for_region): Make static.
2267 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2268 a label and unconditional jump to old_bb, rather than
2269 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2272 PR rtl-optimization/85376
2273 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2274 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2275 instead of a specific value.
2277 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2278 Bin Cheng <bin.cheng@arm.com>
2280 PR tree-optimization/82965
2281 PR tree-optimization/83991
2282 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2283 by_profile_only parameter.
2284 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2285 information if the loop was predicted to iterate too many times.
2286 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2288 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2291 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2294 2018-04-13 Martin Liska <mliska@suse.cz>
2295 Jakub Jelinek <jakub@redhat.com>
2298 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2299 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2300 * builtins.c (expand_builtin_memory_copy_args): Use
2301 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2302 handle dest_addr == pc_rtx.
2304 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2307 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2308 asked to not generate direct moves.
2309 (fix_trunc<mode>si2_stfiwx): Similar.
2310 (fix_trunc<mode>si2_internal): Similar.
2312 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2315 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2316 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2317 lookup if dest in some wider mode is known to be const0_rtx and
2318 if so, record permanent equivalence for it to be ZERO_EXTEND of
2319 the narrower mode destination.
2321 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2323 * lto-streamer-out.c (output_function): Revert 259346.
2324 * omp-expand.c (expand_oacc_for): Likewise.
2326 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2328 PR rtl-optimization/85354
2329 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2330 * sel-sched.c (sel_global_init): ... here.
2332 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2335 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2336 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2337 mode for PE-COFF targets.
2338 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2339 (i386_pe_asm_lto_end): Likewise.
2340 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2341 (TARGET_ASM_LTO_END): Likewise.
2342 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2343 (i386_pe_asm_lto_start): New function.
2344 (i386_pe_asm_lto_end): Likewise.
2346 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2347 Richard Biener <rguenther@suse.de>
2350 * lto-streamer-out.c (output_function): Fix CFG loop state before
2352 * omp-expand.c (expand_oacc_for): Handle calls to internal
2353 functions like regular functions.
2355 2018-04-12 Richard Biener <rguenther@suse.de>
2358 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2359 for the early LTO debug to properly generate references to it
2360 during DIE emission. Do not re-use that for the skeleton for
2362 (dwarf2out_early_finish): Likewise.
2364 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2367 * config/i386/sse.md
2368 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2369 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2370 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2371 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2372 and output is a reg, avoid creating invalid lowpart subreg, but
2373 instead split into a 512-bit move. Don't split if not AVX512VL,
2374 input is xmm16+ reg and output is a mem.
2375 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2376 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2377 xmm16+ reg and output is a mem.
2379 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2381 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2382 also for flag_dwarf2_cfi_asm.
2384 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2386 PR rtl-optimization/85342
2387 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2388 a bool scalar var inside of the loop instead. Don't try to update
2389 recog_data.operand after failed apply_change_group.
2391 2018-04-12 Tom de Vries <tom@codesourcery.com>
2394 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2395 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2396 array with flexible array member as array without given dimension.
2397 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2398 argument for undefined param to true.
2400 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2403 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2404 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2405 from PowerPC section.
2406 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2407 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2408 help text that is too long.
2409 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2410 help text that is too long.
2411 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2412 help text that is too long.
2414 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2416 * config/alpha/alpha.md (stack_probe_internal): Rename
2417 from "probe_stack". Update all callers.
2419 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2421 PR rtl-optimization/84566
2422 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2423 sched_macro_fuse_insns.
2425 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2428 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2429 (compute_block_dependences): ... from here.
2431 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2433 PR tree-optimization/85331
2434 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2435 from int to HOST_WIDE_INT.
2437 2018-04-11 Martin Jambor <mjambor@suse.cz>
2440 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2441 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2442 not the same as the source val.
2443 (cgraph_edge_brings_value_p): New parameter.
2444 (gather_edges_for_value): Pass destination value to
2445 cgraph_edge_brings_value_p.
2446 (perhaps_add_new_callers): Likewise.
2447 (get_info_about_necessary_edges): Likewise and exclude values brought
2448 only by self-recursive edges.
2449 (create_specialized_node): Redirect only clones of self-calling edges.
2450 (+self_recursive_pass_through_p): New function.
2451 (find_more_scalar_values_for_callers_subset): Use it.
2452 (find_aggregate_values_for_callers_subset): Likewise.
2453 (known_aggs_to_agg_replacement_list): Removed.
2454 (decide_whether_version_node): Re-calculate known constants for all
2455 remaining context clones.
2457 2018-04-11 Richard Biener <rguenther@suse.de>
2460 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2461 from early DWARF output.
2462 (dwarf2out_early_finish): Output line info unconditionally into
2463 early DWARF and add reference to it.
2465 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2468 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2469 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2470 other than V2DFmode using iptr mode attribute.
2471 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2473 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2475 PR rtl-optimization/84659
2476 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2478 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2481 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2483 (output_loc_list): Pass address of a dummy size variable even in the
2484 locview handling loop.
2485 (index_location_lists): Add comment on why skip_loc_list_entry can't
2488 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2491 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2494 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2497 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2498 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2499 and -mstring-compare-inline-limit.
2501 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2504 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2505 for stack clash protection in a register whenever we need it to be in
2508 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2510 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2511 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2513 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2516 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2518 (mlong-double-): Ditto.
2519 * config/rs6000/sysv4.opt (msdata=): Ditto.
2520 (mtls-size=): Ditto.
2522 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2524 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2525 erroneous entries for
2526 "vector int vec_ldl (int, long int *)", and
2527 "vector unsigned int vec_ldl (int, unsigned long int *)".
2528 Add comments and entries for
2529 "vector bool char vec_ldl (int, bool char *)",
2530 "vector bool short vec_ldl (int, bool short *)",
2531 "vector bool int vec_ldl (int, bool int *)",
2532 "vector bool long long vec_ldl (int, bool long long *)",
2533 "vector pixel vec_ldl (int, pixel *)",
2534 "vector long long vec_ldl (int, long long *)",
2535 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2536 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2537 type tree bool_long_long_type_node and correct definition of
2538 bool_V2DI_type_node to make reference to this new type tree.
2539 (rs6000_mangle_type): Replace erroneous reference to
2540 bool_long_type_node with bool_long_long_type_node.
2541 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2542 comments to emphasize sign distinctions for char and int types and
2543 replace RS6000_BTI_bool_long constant with
2544 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2545 use of RS6000_BTI_pixel.
2546 (bool_long_type_node): Remove this macro definition.
2547 (bool_long_long_type_node): New macro definition
2549 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2551 PR rtl-optimization/85300
2552 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2553 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2554 simplify_unary_operation fails.
2556 2018-04-10 Martin Liska <mliska@suse.cz>
2558 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2559 cgraph_edge and ipa_ref.
2561 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2565 * config/i386/sse.md
2566 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2567 computation of the VEC_MERGE selector from mask.
2568 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2569 Fix decoding of the VEC_MERGE selector into mask.
2571 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2573 PR tree-optimization/85286
2574 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2576 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2578 * final.c (final_1): Set insn_last_address as well as
2579 insn_current_address.
2581 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2584 * explow.c (emit_stack_probe): Call validize_mem on memory location
2585 before passing it to gen_probe_stack. Create address operand and
2586 legitimize it for the probe_stack_address case.
2588 2018-04-09 Jan Hubicka <jh@suse.cz>
2591 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2592 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2593 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2595 (free_lang_data): Rebuild type inheritance graph.
2597 2018-04-09 Martin Sebor <msebor@redhat.com>
2599 * invoke.texi (-finline-small-functions): Mention other optimization
2601 (-findirect-inlining, -fpartial-inlining): Same.
2602 (-finline-functions-called-once): Same.
2603 (-freorder-blocks-and-partition): Same.
2605 2018-04-09 Jan Hubicka <jh@suse.cz>
2608 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2609 jumps; choose last target that matches the criteria (i.e.
2610 no partition changes for non-crossing jumps).
2611 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2612 support for redirecting crossing jumps to non-crossing.
2614 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2616 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2617 also for naked functions.
2619 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2621 * config/arc/arc.md (add_shift): New pattern.
2622 (add_shift2): Likewise.
2623 (sub_shift): Likewise.
2624 (sub_shift_cmp0_noout): Likewise.
2625 (compare_si_ashiftsi): Likewise.
2626 (xbfu_cmp0_noout): New combine pattern.
2627 (xbfu_cmp0"): Likewise.
2628 (movsi_set_cc_insn): Place the predicable variant first.
2629 (commutative_binary_cmp0_noout): Remove clobber.
2630 (commutative_binary_cmp0): New pattern.
2631 (noncommutative_binary_cmp0): Likewise.
2632 (noncommutative_binary_cmp0_noout): Likewise.
2633 (noncommutative_binary_comparison_result_used): Removed.
2634 (rsub_cmp0): New pattern.
2635 (rsub_cmp0_noout): Likewise.
2636 (extzvsi): Changed, keep only meaningful variants.
2637 (SQH, SEZ): New iterators.
2638 (SQH_postfix): New mode attribute.
2639 (SEZ_prefix): New code attribute.
2640 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2641 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2642 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2644 (noncommutative_operator): Check the availability of barrel
2647 2018-04-09 Richard Biener <rguenther@suse.de>
2649 PR tree-optimization/85284
2650 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2651 Only use the niter constraining form of simple_iv when the exit
2654 2018-04-09 Tom de Vries <tom@codesourcery.com>
2657 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2658 (define_expand "*memory_barrier"): New define_expand.
2659 (define_insn "memory_barrier"): New insn.
2661 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2663 PR rtl-optimization/80463
2664 PR rtl-optimization/83972
2665 PR rtl-optimization/83480
2667 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2668 correct producer for the insn.
2669 (tidy_control_flow): Fixup seqnos in case of debug insns.
2671 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2673 PR rtl-optimization/83913
2675 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2676 different sched-times when merging exprs.
2678 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2680 PR rtl-optimization/83962
2682 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2683 tidy_fallthru_edge and tidy_control_flow.
2685 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2687 PR rtl-optimization/83530
2689 * sel-sched.c (force_next_insn): New global variable.
2690 (remove_insn_for_debug): When force_next_insn is true, also leave only
2691 next insn in the ready list.
2692 (sel_sched_region): When the region wasn't scheduled, make another pass
2693 over it with force_next_insn set to 1.
2695 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2697 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2699 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2700 for interrupt control.
2701 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2702 functions for interrupt control.
2703 * config/nds32/nds32-intrinsic.md: Likewise.
2704 * config/nds32/nds32_intrinsic.h: Likewise.
2705 * config/nds32/nds32.h (nds32_builtins): Likewise.
2707 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2709 * config/nds32/nds32.c (nds32_init_machine_status,
2710 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2711 strict_aligned_p field.
2712 (nds32_expand_to_rtl_hook): New function.
2713 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2714 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2716 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2717 Chung-Ju Wu <jasonwucj@gmail.com>
2719 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2720 * config/nds32/nds32-n7.md: New file.
2721 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2722 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2724 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2725 * config/nds32/nds32.md (pipeline_model): Add n7.
2726 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2727 * config/nds32/pipelines.md: Include n7 settings.
2729 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2730 Chung-Ju Wu <jasonwucj@gmail.com>
2732 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2733 * config/nds32/nds32-e8.md: New file.
2734 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2735 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2737 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2738 * config/nds32/nds32.md (pipeline_model): Add e8.
2739 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2740 * config/nds32/pipelines.md: Include e8 settings.
2742 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2743 Chung-Ju Wu <jasonwucj@gmail.com>
2745 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2746 * config/nds32/nds32-n8.md: New file.
2747 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2748 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2750 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2751 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2752 * config/nds32/nds32.md (pipeline_model): Add n8.
2753 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2754 * config/nds32/pipelines.md: Include n8 settings.
2756 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2757 Chung-Ju Wu <jasonwucj@gmail.com>
2759 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2760 * config/nds32/nds32-n9-2r1w.md: New file.
2761 * config/nds32/nds32-n9-3r2w.md: New file.
2762 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2763 nds32_register_ports): New or modify for cpu n9.
2764 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2766 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2767 * config/nds32/nds32-utils.c: New file.
2768 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2769 TARGET_MUL_SLOW): Define.
2770 * config/nds32/nds32.md (pipeline_model): New attribute.
2771 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2772 New options that support cpu n9.
2773 * config/nds32/pipelines.md: Include n9 settings.
2774 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2776 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2778 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2779 information if necessary.
2780 (output_cond_branch_compare_zero): Likewise.
2781 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2782 (nds32_target_alignment): Refine for alignment.
2783 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2784 (FUNCTION_BOUNDARY): Modify.
2785 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2787 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2789 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2791 * config/nds32/constants.md (unspec_volatile_element): Add values for
2792 TLB operation and data prefetch.
2793 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2794 functions for TLB operation and data prefetch.
2795 * config/nds32/nds32-intrinsic.md: Likewise.
2796 * config/nds32/nds32_intrinsic.h: Likewise.
2797 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2798 (nds32_print_operand): Likewise.
2799 * config/nds32/nds32.h (nds32_builtins): Likewise.
2801 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2802 Andrew Pinski <pinsika@gcc.gnu.org>
2805 * match.pd: Use constant_boolean_node of correct type instead of
2806 boolean_true_node or boolean_false_node for simplifying
2807 pointer comparisons to zero.
2809 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2811 PR tree-optimization/80021
2812 * tree.c (verify_type_variant): Make error call in verify_variant_match
2813 translatable and remove final full stop.
2815 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2817 * config/nds32/constants.md (unspec_volatile_element): Add
2818 UNSPEC_VOLATILE_EH_RETURN.
2819 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2820 nds32_output_stack_pop): Support dwarf exception handling process.
2821 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2822 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2823 exception handling process.
2824 (nds32_compute_stack_frame): Likewise.
2825 (nds32_return_addr_rtx): Likewise.
2826 (nds32_initial_elimination_offset): Likewise.
2827 (nds32_expand_prologue): Likewise.
2828 (nds32_expand_epilogue): Likewise.
2829 (nds32_dynamic_chain_address): New function.
2830 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2832 (DYNAMIC_CHAIN_ADDRESS): Define.
2833 (EH_RETURN_DATA_REGNO): Define.
2834 (EH_RETURN_STACKADJ_RTX): Define.
2835 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2836 patterns for dwarf exception handling.
2838 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2840 * config/nds32/nds32.h: Clean up obsolete macros.
2842 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2844 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2845 Add enum values for particular instructions.
2846 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2847 particular intrinsic functions.
2848 * config/nds32/nds32-intrinsic.md: Likewise.
2849 * config/nds32/nds32_intrinsic.h: Likewise.
2850 * config/nds32/nds32.h (nds32_builtins): Likewise.
2851 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2852 (btst, ave): New patterns for particular instructions.
2854 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2856 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2857 Add enum values for atomic load/store and memory sync.
2858 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2860 * config/nds32/nds32-intrinsic.md: Likewise.
2861 * config/nds32/nds32_intrinsic.h: Likewise.
2862 * config/nds32/nds32.h (nds32_builtins): Likewise.
2864 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2866 PR tree-optimization/85257
2867 * fold-const.c (native_encode_vector): If not all elts could fit
2868 and off is -1, return 0 rather than offset.
2869 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2870 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2871 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2872 adjust buffer in native_interpret_expr call.
2874 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2876 * config/nds32/constants.md (unspec_volatile_element): Add cache
2877 control enum values.
2878 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2879 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2880 * config/nds32/nds32.c (nds32_cctl_names): New.
2881 (nds32_print_operand): Handle cache control register names.
2882 * config/nds32/nds32.h (nds32_builtins): New enum values.
2883 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2885 * config/nds32/nds32.md (type): Add mmu.
2886 * config/nds32/pipelines.md (simple_insn): Add mmu.
2888 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2890 * config/nds32/nds32.md (type): Remove call.
2891 * config/nds32/pipelines.md (simple_insn): Likewise.
2893 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2895 * config/nds32/constants.md (unspec_volatile_element): Add
2896 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2897 UNSPEC_VOLATILE_FMFCFG.
2898 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2899 description for fmfcfg and fmfcsr.
2900 (bdesc_1arg): Add fmtcsr.
2901 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2902 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2903 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2904 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2905 unspec_fmfcfg): New patterns.
2906 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2907 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2908 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2909 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2910 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2911 __nds32__fmfcfg): Define.
2913 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2915 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2916 intrinsic register names.
2917 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2918 intrinsic register enum values and macros.
2920 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2922 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2923 for load/store addressing form.
2924 (nds32_print_operand_address): Likewise.
2926 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2929 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2930 based on LABEL_REF. Remove useless assertion.
2931 (pic_address_needs_scratch): Fix formatting.
2932 (sparc_legitimize_pic_address): Minor tweaks.
2933 (sparc_delegitimize_address): Adjust assertion accordingly.
2934 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2935 into symbolic_operand.
2936 (movsi_high_pic_label_ref): Likewise.
2937 (movsi_lo_sum_pic_label_ref): Likewise.
2938 (movdi_pic_label_ref): Likewise.
2939 (movdi_high_pic_label_ref): Likewise.
2940 (movdi_lo_sum_pic_label_ref): Likewise.
2942 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2944 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2945 custom LIB_SPEC setup.
2947 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2948 Kito Cheng <kito.cheng@gmail.com>
2950 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2951 * config/riscv/freebsd.h: New.
2953 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2955 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2956 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2959 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2960 Kito Cheng <kito.cheng@gmail.com>
2962 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2963 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2964 * config/nds32/nds32-protos.h (nds32_output_call,
2965 nds32_output_return): Declare.
2966 * config/nds32/nds32.md: Refine all the call and return patterns.
2968 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2971 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2972 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2974 PR rtl-optimization/84872
2975 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2976 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2979 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2981 * expr.c (copy_blkmode_to_reg): Revert 254862.
2982 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2984 2018-04-06 Richard Biener <rguenther@suse.de>
2987 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2988 after seeing a component reference with an adjacent field. Treat
2989 refs to arrays at struct end of external decls similar to
2990 refs to unconstrained commons.
2992 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2995 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2996 look through SAVE_EXPRs with non-side-effects argument. Adjust
2998 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3001 2018-04-06 Richard Biener <rguenther@suse.de>
3004 * alias.c (find_base_term): New wrapper around find_base_term
3005 unwinding CSELIB_VAL_PTR changes.
3006 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3009 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3011 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3013 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3014 constant definitions.
3015 ("nop"): lr 0,0 -> nopr r0
3016 ("nop_lr0", "nop_lr1"): New insn definitions.
3018 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3020 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3021 NDS32_V3PUSH_AVAILABLE_P macro.
3023 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3024 Chung-Ju Wu <jasonwucj@gmail.com>
3026 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3027 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3028 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3029 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3030 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3031 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3032 * config/nds32/constraints.md: New constraints and checking for hard
3033 float configuration.
3034 * config/nds32/iterators.md: New mode iterator and attribute for hard
3035 float configuration.
3036 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3038 * config/nds32/nds32-fpu.md: New file.
3039 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3040 deal with hard float code generation.
3041 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3043 (abi_type, float_reg_number): New enum type.
3044 * config/nds32/nds32-predicates.c: New predicates for hard float.
3045 * config/nds32/nds32-protos.h: Declare functions for hard float.
3046 * config/nds32/nds32.c: Implementation for hard float configuration.
3047 * config/nds32/nds32.h: Definitions for hard float configuration.
3048 * config/nds32/nds32.md: Include hard float machine description and
3049 modify patterns for hard float configuration.
3050 * config/nds32/nds32.opt: New options for hard float configuration.
3051 * config/nds32/predicates.md: New predicates for hard float
3054 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3056 * common/config/nds32/nds32-common.c
3057 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3059 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3062 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3063 CONSTRUCTOR_ELT (ctor, ...)->value.
3065 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3068 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3070 2018-04-05 Tom de Vries <tom@codesourcery.com>
3073 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3076 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3077 Kito Cheng <kito.cheng@gmail.com>
3079 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3080 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3081 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3082 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3084 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3085 Kito Cheng <kito.cheng@gmail.com>
3087 * config/nds32/constraints.md (Ufe): New memory constraint.
3088 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3089 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3090 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3092 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3093 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3095 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3097 * config/nds32/nds32.md: Use optimize_size in the condition for
3098 alu-shift instructions.
3100 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3102 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3104 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3106 * config/nds32/nds32.md (negsi2): Refine pattern.
3108 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3109 Chung-Ju Wu <jasonwucj@gmail.com>
3111 * config/nds32/iterators.md (shift_rotate): New code iterator.
3112 (shift): New code attribute.
3113 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3114 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3115 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3116 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3117 bit-wise operations.
3118 (andsi3, *andsi3): Ditto.
3119 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3120 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3121 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3122 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3123 nds32_ior_operand, nds32_xor_operand): New predicates.
3125 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3127 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3128 (addsi3, subsi3): ... this.
3130 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3132 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3134 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3136 * config/nds32/nds32.md: Adjust indention.
3138 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3140 * config/nds32/nds32.md (feature): New attribute.
3142 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3144 * config/nds32/nds32.md (subtype): New attribute.
3146 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3149 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3150 expansion to perform a bitwise AND of the argument followed by a
3151 boolean negation of the result.
3153 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3155 PR rtl-optimization/84878
3156 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3157 the basic block. Assert the use reference is not artificial and that
3158 it has an associated insn.
3160 2018-04-04 Michael Matz <matz@suse.de>
3162 * builtins.c (compute_objsize): Pass correct operand
3163 to array_at_struct_end_p.
3165 2018-04-04 Richard Biener <rguenther@suse.de>
3168 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3169 from contexts for DINFO_LEVEL_TERSE and below.
3171 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3173 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3174 resiter_operand condition.
3175 * config/nds32/nds32.md (*move<mode>): Ditto.
3177 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3178 Monk Chiang <sh.chiang04@gmail.com>
3180 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3182 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3184 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3186 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3187 Kito Cheng <kito.cheng@gmail.com>
3189 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3190 nds32_cond_code_str, output_cond_branch,
3191 output_cond_branch_compare_zero, nds32_expand_cbranch,
3192 nds32_expand_cstore, nds32_expand_movcc,
3193 nds32_output_cbranchsi4_equality_zero,
3194 nds32_output_cbranchsi4_equality_reg,
3195 nds32_output_cbranchsi4_equality_reg_or_const_int,
3196 nds32_output_cbranchsi4_greater_less_zero: New functions.
3197 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3198 nds32_expand_cstore, nds32_expand_movcc,
3199 nds32_output_cbranchsi4_equality_zero,
3200 nds32_output_cbranchsi4_equality_reg,
3201 nds32_output_cbranchsi4_equality_reg_or_const_int,
3202 nds32_output_cbranchsi4_greater_less_zero): Declare.
3203 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3204 nds32_rimm11s_operand): New predicates.
3205 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3206 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3209 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3211 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3212 * config/nds32/nds32.md: Ditto.
3213 * config/nds32/pipelines.md: Ditto.
3215 2018-04-04 Richard Biener <rguenther@suse.de>
3217 PR tree-optimization/85168
3218 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3219 propagating abnormals.
3221 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3223 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3225 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3226 Kito Cheng <kito.cheng@gmail.com>
3228 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3229 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3230 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3231 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3232 * config/nds32/nds32.md (sibcall_internal): New.
3233 (sibcall_register): Remove.
3234 (sibcall_immediate): Remove.
3235 (sibcall_value_internal): New.
3236 (sibcall_value_register): Remove.
3237 (sibcall_value_immediate): Remove.
3238 * config/nds32/predicates.md (nds32_general_register_operand): New.
3239 (nds32_call_address_operand): New.
3241 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3243 PR rtl-optimization/85167
3244 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3245 bb_defs if *split_p, instead preinitialize it to NULL.
3247 PR tree-optimization/85156
3248 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3249 evaluating the argument multiple times.
3251 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3253 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3255 (_mm_cvtpd_ps): Likewise.
3256 (_mm_cvttpd_epi32): Likewise.
3257 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3258 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3259 vector, pixel, and bool following altivec.h include.
3261 2018-04-03 Martin Sebor <msebor@redhat.com>
3263 * doc/extend.texi (Common Function Attributes): Clarify.
3264 (const attribute): Likewise.
3265 (pure attribute): Likewise.
3267 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3270 * config/i386/i386.c (ix86_expand_vector_set): Use
3271 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3273 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3275 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3276 instructions when changing rounding bits to preserve precision bits
3277 in the x87 control word.
3279 2018-04-03 Martin Liska <mliska@suse.cz>
3281 PR tree-optimization/82491
3282 * rtl.h (strip_offset_and_add): Replace += suboffset with
3283 poly_uint64 () + suboffset.
3285 2018-03-29 Martin Liska <mliska@suse.cz>
3286 Martin Jambor <mjambor@suse.cz>
3289 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3290 param_type is not an integral or pointer type.
3292 2018-04-03 Richard Biener <rguenther@suse.de>
3294 * sese.h (recompute_all_dominators): Remove.
3296 2018-04-02 Martin Sebor <msebor@redhat.com>
3298 * doc/invoke.texi (-Wrestrict): Fix typos.
3300 2018-04-02 Jim Wilson <jimw@sifive.com>
3302 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3303 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3304 (<optab>di3, <optab>si3_extend): Likewise.
3305 (<optab>si3_mask, <optab>si3_mask_1): New.
3306 (<optab>di3_mask, <optab>di3_mask_1): New.
3307 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3308 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3309 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3311 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3313 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3316 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3318 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3319 (nds32_canonicalize_comparison): New function.
3321 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3322 Kito Cheng <kito.cheng@gmail.com>
3323 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3325 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3326 * config/nds32/constants.md (unspec_volatile_element): Add
3327 UNSPEC_VOLATILE_RELAX_GROUP.
3328 * config/nds32/nds32-relax-opt.c: New file.
3329 * config/nds32/nds32-predicates.c
3330 (nds32_symbol_load_store_p): New function.
3331 * config/nds32/nds32-protos.h
3332 (nds32_symbol_load_store_p): Declare function.
3333 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3334 * config/nds32/nds32.c
3335 (nds32_register_pass): New function to register pass.
3336 (nds32_register_passes): New function to register passes.
3337 * config/nds32/nds32.md (relax_group): New pattern.
3338 * config/nds32/nds32.opt (mrelax-hint): New option.
3339 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3341 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3343 * config/nds32/t-nds32: Modify files dependency.
3345 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3347 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3348 (PROFILE_HOOK): Define its implementation.
3350 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3352 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3353 type and 32-bit size.
3355 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3358 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3359 (V_128_256): New mode iterator.
3360 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3361 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3362 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3364 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3365 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3367 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3370 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3371 NaN inputs correctly.
3373 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3376 * config/rs6000/vsx.md (??r): New mode attribute.
3377 (*vsx_mov<mode>_64bit): Use it.
3378 (*vsx_mov<mode>_32bit): Likewise.
3380 2018-03-30 Martin Sebor <msebor@redhat.com>
3382 PR tree-optimization/84818
3383 * builtins.c (check_access): Use warning_n.
3385 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3388 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3390 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3393 2018-03-30 Julia Koval <julia.koval@intel.com>
3396 * x86-tune.def (movx, partial_reg_dependency): Enable for
3399 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3402 * lra-constraints.c (process_alt_operands): Move setting
3403 this_alternative_matches below.
3405 2018-03-29 Martin Liska <mliska@suse.cz>
3408 * doc/invoke.texi: Document how LTO works with debug info.
3409 Describe auto-load support of binutils. Mention 'x86-64'
3410 as valid option value of -march option.
3412 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3414 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3417 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3418 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3419 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3422 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3425 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3426 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3427 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3428 for RS6000_BTM_POWERPC64.
3429 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3430 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3431 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3436 2018-03-28 Carl Love <cel@us.ibm.com>
3439 2017-09-27 Carl Love <cel@us.ibm.com>
3441 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3442 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3443 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3446 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3448 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3449 instead of __vector bool.
3450 (_mm_max_pu8): Likewise.
3451 (_mm_min_pi16): Likewise.
3453 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3456 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3457 (DIVWEUO): Likewise.
3459 (DIVDEUO): Likewise.
3460 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3461 DIVWEUO and DIVDEUO.
3462 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3463 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3464 (div_extend): Likewise.
3465 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3467 (__builtin_divweuo): Likewise.
3468 (__builtin_divdeo): Likewise.
3469 (__builtin_divdeuo): Likewise.
3471 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3474 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3475 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3477 PR tree-optimization/82004
3478 * gimple-match-head.c (optimize_pow_to_exp): New function.
3479 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3480 Don't fold to exp if optimize_pow_to_exp is false.
3482 2018-03-28 Martin Liska <mliska@suse.cz>
3485 * calls.c (initialize_argument_information): Fix trailing space.
3486 * common.opt: Fix typo and provide better explanation for
3487 -fsanitize-coverage option.
3488 * config/i386/i386.opt: Fix typo.
3490 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3491 Martin Liska <mliska@suse.cz>
3494 * gimplify.c (asan_poison_variable): Don't do the check for
3495 gimplify_omp_ctxp here.
3496 (gimplify_decl_expr): Do it here.
3497 (gimplify_target_expr): Likewise.
3499 2018-03-28 Martin Liska <mliska@suse.cz>
3502 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3503 chkp_type_bounds_count if MPX is not enabled.
3505 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3507 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3509 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3512 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3513 function to create the function decl for complex long double
3514 multiply and divide for -mabi=ieeelongdouble.
3515 (init_float128_ieee): Call it.
3517 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3520 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3521 -fcf-protection=branch -mibt.
3522 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3524 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3527 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3529 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3532 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3533 extern array declarations.
3535 2018-03-27 Richard Biener <rguenther@suse.de>
3538 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3539 explicit single_use checks.
3541 2018-03-27 Richard Biener <rguenther@suse.de>
3543 PR tree-optimization/85082
3544 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3547 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3549 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3550 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3551 Turn on fasynchronous-unwind-tables and funwind-tables.
3553 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3556 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3557 (*bmi_blsr_<mode>_ccz): Ditto.
3559 2018-03-26 Tom de Vries <tom@codesourcery.com>
3561 PR tree-optimization/85063
3562 * omp-general.c (offloading_function_p): New function. Factor out
3564 * omp-offload.c (pass_omp_target_link::gate): ... here.
3565 * omp-general.h (offloading_function_p): Declare.
3566 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3567 with attribute omp declare target for offloading functions.
3569 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3571 PR tree-optimization/84005
3572 * tree-data-ref.h (get_base_for_alignment): Declare.
3573 * tree-data-ref.c (get_base_for_alignment_1): New function.
3574 (get_base_for_alignment): Likewise.
3575 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3576 get_base_for_alignment to find a suitable base object, instead
3577 of always using drb->base_address.
3579 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3582 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3583 known size by default.
3585 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3588 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3589 and non BLKmode operands.
3591 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3594 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3595 Clean up attributes.
3597 2018-03-23 Richard Biener <rguenther@suse.de>
3600 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3601 we are going to emit early debug for LTO.
3603 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3606 * function.c (match_asm_constraints_1): Don't optimize if input
3607 doesn't satisfy general_operand predicate for output's mode.
3610 * alias.c (write_dependence_p): Don't require for x_canonicalized
3611 non-VOIDmode if x has VOIDmode.
3614 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3615 just don't try to optimize it rather than assert it never happens.
3617 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3619 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3620 macro expansions for definition of ST_INTERNAL_<mode> and
3621 LD_INTERNAL_<mode> builtins.
3622 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3624 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3626 (altivec_expand_st_builtin): Likewise.
3627 (altivec_expand_builtin): Remove calls to deleted functions.
3628 (rs6000_address_for_altivec): Delete this function.
3629 * config/rs6000/vector.md: Remove expands for
3630 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3632 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3635 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3636 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3637 re-computing once computed.
3638 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3639 (arm_init_machine_status): Initialize
3640 machine->static_chain_stack_bytes.
3642 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3645 * doc/extend.texi: Add four new prototypes for vec_ld.
3646 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3647 definitions for more logical presentation.
3648 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3649 entries for V1TI variants of __builtin_altivec_ld builtin.
3650 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3651 handling of V1TI variant of LVX icode pattern.
3652 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3653 (rs6000_gimple_fold_builtin): Likewise.
3654 (altivec_init_builtins): Add code to define
3655 __builtin_altivec_lvx_v1ti function.
3657 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3660 * function.c (match_asm_constraints_1): Don't do the optimization
3661 if input isn't a REG, SUBREG, MEM or constant.
3663 2018-03-22 Tom de Vries <tom@codesourcery.com>
3665 PR tree-optimization/84956
3666 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3667 bb_has_abnormal_pred.
3669 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3672 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3673 DECL_INITIAL (decl) to decl at the end.
3674 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3677 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3679 * doc/extend.texi (__builtin_tgmath): Document when complex
3680 integer types are treated as _Complex _Float64.
3682 2018-03-21 Tom de Vries <tom@codesourcery.com>
3684 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3686 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3688 PR tree-optimization/84960
3689 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3690 if it is ENTRY block, move them into single succ of ENTRY in that case.
3692 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3694 PR tree-optimization/84811
3695 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3696 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3697 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3698 (known_subrange_p): Update accordingly. Cast each value involved
3699 in the size comparison, rather than casting the result of the
3702 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3704 PR tree-optimization/84982
3705 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3706 by flipping the least significant bit rather than all bits from
3707 bitpos to bitpos + bitsize - 1.
3709 2018-03-21 Nathan Sidwell <nathan@acm.org>
3711 * doc/extend.texi (Deprecated Features): Remove mention of
3712 long-deleted deprecations.
3714 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3717 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3718 * configure: Regenerate.
3720 2018-03-21 Tom de Vries <tom@codesourcery.com>
3722 PR tree-optimization/83126
3723 * tree-parloops.c (num_phis): New function.
3724 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3726 2018-03-21 Nathan Sidwell <nathan@acm.org>
3728 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3729 mention anon-struct/union members and trailing attributes.
3731 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3733 PR tree-optimization/84969
3734 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3735 builtin memset partitions if they set different rhs values.
3737 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3739 PR rtl-optimization/84989
3740 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3741 VEC_DUPLICATE with scalar result mode.
3743 2018-03-21 Martin Liska <mliska@suse.cz>
3746 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3747 not intended return statement.
3749 2018-03-21 Martin Liska <mliska@suse.cz>
3752 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3753 (chkp_find_bound_slots_1): Limit number of iterations.
3755 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3758 * Minor grammar fixes for x86 options.
3760 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3763 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3764 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3766 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3769 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3770 (altivec_lvx_<mode>_1op): Likewise.
3771 (altivec_stvx_<mode>_2op): Likewise.
3772 (altivec_stvx_<mode>_1op): Likewise.
3773 (altivec_lvx_<VM2:mode>): New define_expand.
3774 (altivec_stvx_<VM2:mode>): Likewise.
3775 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3776 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3777 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3778 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3779 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3780 (rs6000_gen_lvx): Likewise.
3781 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3782 (altivec_expand_stv_builtin): Likewise.
3783 (altivec_expand_builtin): Likewise.
3784 * config/rs6000/vector.md: Likewise.
3786 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3789 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3792 2018-03-20 Richard Biener <rguenther@suse.de>
3795 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3796 sign-conversions as zero, fall back to standard scalar_stmt
3799 2018-03-20 Martin Liska <mliska@suse.cz>
3802 * predict.c (rebuild_frequencies): Handle case when we have
3803 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3805 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3808 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3809 flag_section_anchors.
3810 * varasm.c (use_blocks_for_decl_p): Remove hack for
3811 dw2_force_const_mem.
3814 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3816 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3817 be created, use lowpart_subreg of operands[0] rather than operands[0]
3819 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3820 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3821 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3822 and n constraint instead of aarch64_shift_imm_di and Usd.
3823 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3824 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3826 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3829 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3830 to favor GPR over NEON registers.
3831 (<shift>di3_neon): Likewise.
3833 2018-03-20 Tom de Vries <tom@codesourcery.com>
3836 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3837 (nvptx_process_pars): Emit bar.sync asap and alap.
3839 2018-03-20 Tom de Vries <tom@codesourcery.com>
3842 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3843 seen_label if seen_label is already set.
3845 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3848 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3849 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3850 Use 1U instead of 1. Formatting fixes.
3853 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3854 instead of TREE_TYPE (s1) for the return value.
3856 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3858 PR tree-optimization/84946
3859 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3860 bitsize + bitsize in poly_uint64 rather than poly_int64.
3863 * dwarf2asm.c: Include fold-const.c.
3864 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3865 of decl rather than decl itself.
3867 PR rtl-optimization/84643
3868 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3870 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3873 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3874 calling assemble_variable.
3876 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3879 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3880 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3882 2018-03-19 Jim Wilson <jimw@sifive.com>
3885 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3886 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3887 (riscv_first_stack_step): Likewise.
3888 (riscv_option_override): Use STACK_BOUNDARY instead of
3890 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3892 (BIGGEST_ALIGNMENT): Set to 128.
3893 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3894 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3897 2018-03-19 Richard Biener <rguenther@suse.de>
3899 PR tree-optimization/84933
3900 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3901 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3903 2018-03-19 Richard Biener <rguenther@suse.de>
3905 PR tree-optimization/84859
3906 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3907 (cond_if_else_store_replacement): Perform sinking operation on
3908 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3909 Generalize what a BB with a single eligible store is.
3911 2018-03-19 Richard Biener <rguenther@suse.de>
3913 PR tree-optimization/84929
3914 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3915 chrec_is_positive against non-chrec arg.
3917 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3920 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3922 2018-03-18 Martin Liska <mliska@suse.cz>
3924 PR rtl-optimization/84635
3925 * regrename.c (build_def_use): Use matches_mode only when
3928 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3930 PR tree-optimization/84913
3931 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3932 vectorize chains of COND_EXPRs.
3934 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3936 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3938 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3940 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3942 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3944 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3946 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3947 Kito Cheng <kito.cheng@gmail.com>
3949 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3950 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3951 (nds32_adjust_reg_alloc_order): New function.
3952 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3954 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3956 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3957 nds32_print_operand, nds32_print_operand_address): Use
3958 HOST_WIDE_INT_PRINT_DEC instead.
3960 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3962 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3964 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3967 * config/i386/i386.c (initial_ix86_tune_features,
3968 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3970 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3971 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3972 rather than 1u << ix86_tune. Formatting fix.
3973 (ix86_option_override_internal): Change ix86_arch_mask from
3974 unsigned int to unsigned HOST_WIDE_INT, initialize to
3975 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3976 (ix86_function_specific_restore): Likewise.
3978 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3981 * postreload.c (reload_combine_recognize_pattern): Perform
3982 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3983 truncate_int_for_mode the result for the destination's mode.
3986 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3987 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3989 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3992 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3995 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3996 regno_allocno_class_array and sorted_pseudos.
3997 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3998 insns where regno is used.
4000 2018-03-16 Martin Liska <mliska@suse.cz>
4003 * multiple_target.c (create_dispatcher_calls): Redirect
4004 reference in the symbol table.
4006 2018-03-16 Martin Liska <mliska@suse.cz>
4009 * multiple_target.c (create_dispatcher_calls): Redirect also
4012 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4016 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4017 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4018 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4020 2018-03-16 Julia Koval <julia.koval@intel.com>
4022 * doc/invoke.texi (Skylake Server): Add CLWB.
4023 Cannonlake): Remove CLWB.
4025 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4027 PR tree-optimization/84841
4028 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4030 (FLOAT_ONE_CONST_TYPE): Define.
4031 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4032 (sort_by_operand_rank): Put entries with higher constant_type last
4033 rather than first to match comments.
4035 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4037 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4040 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4043 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4044 check and use address of target always.
4046 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4049 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4050 (indirect_thunk_bnd_needed): Likewise.
4051 (indirect_thunks_used): Likewise.
4052 (indirect_thunks_bnd_used): Likewise.
4053 (indirect_return_needed): New.
4054 (indirect_return_bnd_needed): Likewise.
4055 (output_indirect_thunk_function): Add a bool argument for
4057 (output_indirect_thunk_function): Don't generate alias for
4058 function return thunk.
4059 (ix86_code_end): Call output_indirect_thunk_function to generate
4060 function return thunks.
4061 (ix86_output_function_return): Set indirect_return_bnd_needed
4062 and indirect_return_needed instead of indirect_thunk_bnd_needed
4063 and indirect_thunk_needed.
4065 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4067 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4068 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4069 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4071 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4072 Paul Hua <paul.hua.gm@gmail.com>
4075 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4077 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4079 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4080 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4083 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4086 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4087 instead of GET_MODE_SIZE when comparing Units.
4089 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4092 * varasm.c (hash_section): Return an unchangeble hash value
4093 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4094 Return !aarch64_can_use_per_function_literal_pools_p ().
4096 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4099 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4100 rather than address of cmode as last argument to prepare_cmp_insn.
4102 2018-03-15 Julia Koval <julia.koval@intel.com>
4104 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4105 F_AVX512VNNI, F_AVX512BITALG): New.
4107 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4110 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4111 insn for floating-point loads and stores.
4113 2018-03-14 Carl Love <cel@us.ibm.com>
4115 * config/rs6000/rs6000-c.c: Add macro definitions for
4116 ALTIVEC_BUILTIN_VEC_PERMXOR.
4117 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4118 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4119 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4120 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4122 * config/doc/extend.texi: Add prototypes for vec_permxor.
4124 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4127 * diagnostic-show-locus.c (class layout_point): Convert m_line
4128 from int to linenum_type.
4129 (line_span::comparator): Use linenum "compare" function when
4130 comparing line numbers.
4131 (test_line_span): New function.
4132 (layout_range::contains_point): Convert param "row" from int to
4134 (layout_range::intersects_line_p): Likewise.
4135 (layout::will_show_line_p): Likewise.
4136 (layout::print_source_line): Likewise.
4137 (layout::should_print_annotation_line_p): Likewise.
4138 (layout::print_annotation_line): Likewise.
4139 (layout::print_leading_fixits): Likewise.
4140 (layout::annotation_line_showed_range_p): Likewise.
4141 (struct line_corrections): Likewise for field m_row.
4142 (line_corrections::line_corrections): Likewise for param "row".
4143 (layout::print_trailing_fixits): Likewise.
4144 (layout::get_state_at_point): Likewise.
4145 (layout::get_x_bound_for_row): Likewise.
4146 (layout::print_line): Likewise.
4147 (diagnostic_show_locus): Likewise for locals "last_line" and
4149 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4150 * input.c (selftest::test_linenum_comparisons): New function.
4151 (selftest::input_c_tests): Call it.
4152 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4153 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4154 * selftest.h (ASSERT_GT): New macro.
4155 (ASSERT_GT_AT): New macro.
4156 (ASSERT_LT): New macro.
4157 (ASSERT_LT_AT): New macro.
4159 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4161 PR rtl-optimization/84780
4162 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4164 2018-03-14 Martin Liska <mliska@suse.cz>
4166 * tree.c (record_node_allocation_statistics): Use
4167 get_stats_node_kind.
4168 (get_stats_node_kind): New function extracted from
4169 record_node_allocation_statistics.
4170 (free_node): Use get_stats_node_kind.
4172 2018-03-14 Richard Biener <rguenther@suse.de>
4174 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4175 that the value-set of ANTIC_IN doesn't grow.
4178 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4180 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4181 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4183 2018-03-14 Julia Koval <julia.koval@intel.com>
4185 * config.gcc (icelake-client, icelake-server): New.
4187 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4188 (initial_ix86_arch_features): Ditto.
4189 (PTA_SKYLAKE): Add SGX.
4190 (PTA_ICELAKE): Remove.
4191 (PTA_ICELAKE_CLIENT): New.
4192 (PTA_ICELAKE_SERVER): New.
4193 (ix86_option_override_internal): Split up icelake on icelake client and
4195 (get_builtin_code_for_version): Ditto.
4196 (fold_builtin_cpu): Ditto.
4197 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4198 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4199 * config/i386/i386.h (processor_type): Ditto.
4200 * doc/invoke.texi: Ditto.
4202 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4205 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4206 INTEGER_CST offset, add it together with bitpos / 8 and
4207 sign extend based on POINTER_SIZE.
4211 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4214 * config/i386/constraints.md (Yc): New register constraint.
4215 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4216 Use Yc constraint for alternative 2 of operand 0. Remove
4217 preferred_for_speed attribute.
4219 2018-03-14 Richard Biener <rguenther@suse.de>
4221 PR tree-optimization/84830
4222 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4223 with the old one to avoid oscillations.
4225 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4228 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4230 (assign_by_spills): Return a flag of reload assignment failure.
4231 Do not process the reload assignment failures. Do not spill other
4232 reload pseudos if they has the same reg class. Update n if
4234 (lra_assign): Add a return arg. Set up from the result of
4235 assign_by_spills call.
4236 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4237 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4238 usage_insns if it is not NULL.
4239 (spill_hard_reg_in_range): New function.
4240 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4241 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4242 function prototypes.
4243 (lra_assign): Change prototype.
4244 * lra.c (lra): Add code to deal with fails by splitting hard reg
4247 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4249 * config/riscv/riscv.opt (mrelax): New option.
4250 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4251 "norelax" when riscv_mrelax is disabled.
4252 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4254 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4257 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4258 reassociation for int modes.
4260 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4262 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4263 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4265 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4266 * config/aarch64/aarch64-sve.md
4267 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4268 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4269 (*extend<mode><Vwide>2): Rename to...
4270 (aarch64_sve_extend<mode><Vwide>2): ...this.
4271 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4272 renaming the old pattern to...
4273 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4275 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4276 define_expand, renaming the old pattern to...
4277 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4278 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4279 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4280 account when deciding which SVE instruction the optab should use.
4281 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4283 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4285 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4286 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4287 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4288 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4289 (tlsdesc_small_<mode>): Turn a define_expand and use
4290 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4291 (tlsdesc_small_advsimd_<mode>): ...this.
4292 (tlsdesc_small_sve_<mode>): New pattern.
4294 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4296 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4297 (UNSPEC_UMUL_HIGHPART): New constants.
4298 (MUL_HIGHPART): New int iteraor.
4299 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4300 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4302 (*<su>mul<mode>3_highpart): New define_insn.
4304 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4307 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4310 2018-03-13 Martin Liska <mliska@suse.cz>
4313 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4315 (sem_item_optimizer::~sem_item_optimizer): Release it.
4316 (sem_item_optimizer::merge_classes): Register variable aliases.
4317 (sem_item_optimizer::fixup_pt_set): New function.
4318 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4319 * ipa-icf.h: Declare new variables and functions.
4321 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4324 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4325 integer_pow2p@2 and test integer_pow2p in condition.
4326 (A < 0 ? C : 0): Similarly for @1.
4329 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4330 characters starting at p contain '\0' character, don't look beyond
4334 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4335 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4338 * reg-stack.c (change_stack): Change update_end var from int to
4339 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4340 also call set_block_for_insn on the newly added insns and rescan.
4343 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4344 on the last operand.
4347 * tree.c (stabilize_reference_1): Return save_expr (e) for
4348 STATEMENT_LIST even if it doesn't have side-effects.
4350 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4352 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4354 2018-03-12 Renlin Li <renlin.li@arm.com>
4356 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4357 aarch64_output_scalar_simd_mov_immediate.
4359 2018-03-12 Martin Sebor <msebor@redhat.com>
4361 PR tree-optimization/83456
4362 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4363 for perfectly overlapping calls to memcpy.
4364 (gimple_fold_builtin_memory_chk): Same.
4365 (gimple_fold_builtin_strcpy): Handle no-warning.
4366 (gimple_fold_builtin_stxcpy_chk): Same.
4367 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4369 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4371 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4372 parameter. Use it for SFmode.
4373 (rs6000_function_arg_advance_1): Adjust.
4374 (rs6000_function_arg): Adjust.
4375 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4377 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4379 PR rtl-optimization/84169
4380 PR rtl-optimization/84780
4381 * combine.c (can_combine_p): Check for a 2-insn combination whether
4382 the destination register is used between the two insns, too.
4384 2018-03-12 Richard Biener <rguenther@suse.de>
4386 PR tree-optimization/84803
4387 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4388 for refs DR analysis didn't process.
4390 2018-03-12 Richard Biener <rguenther@suse.de>
4392 PR tree-optimization/84777
4393 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4394 force-vectorize loops ignore whether we are optimizing for size.
4396 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4398 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4399 (TARGET_MD_ASM_ADJUST): Define.
4401 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4402 Kito Cheng <kito.cheng@gmail.com>
4403 Chung-Ju Wu <jasonwucj@gmail.com>
4405 * config/nds32/nds32.c (nds32_compute_stack_frame,
4406 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4407 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4408 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4409 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4410 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4411 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4412 * config/nds32/nds32.md (prologue, epilogue): Use macro
4413 NDS32_V3PUSH_AVAILABLE_P to do checking.
4415 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4418 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4419 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4420 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4421 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4422 addition of most attributes on !orig_type_die or the attribute not
4423 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4425 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4426 Chung-Ju Wu <jasonwucj@gmail.com>
4428 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4430 * config/nds32/nds32.opt (mvh): New option.
4432 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4433 Chung-Ju Wu <jasonwucj@gmail.com>
4435 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4437 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4438 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4441 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4442 Chung-Ju Wu <jasonwucj@gmail.com>
4444 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4446 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4447 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4449 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4450 Kito Cheng <kito.cheng@gmail.com>
4451 Chung-Ju Wu <jasonwucj@gmail.com>
4453 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4454 UNSPEC_FFMISM and UNSPEC_FLMISM.
4455 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4456 for ffb, ffmism and flmism.
4457 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4458 (unspec_ffmism): Ditto.
4459 (unspec_flmism): Ditto.
4460 (nds32_expand_builtin_impl): Check if string extension is available.
4461 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4462 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4464 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4467 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4470 * lra-assigns.c (assign_by_spills): Return a flag of reload
4471 assignment failure. Do not process the reload assignment
4472 failures. Do not spill other reload pseudos if they has the same
4474 (lra_assign): Add a return arg. Set up from the result of
4475 assign_by_spills call.
4476 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4477 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4478 usage_insns if it is not NULL.
4479 (spill_hard_reg_in_range): New function.
4480 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4481 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4482 function prototypes.
4483 (lra_assign): Change prototype.
4484 * lra.c (lra): Add code to deal with fails by splitting hard reg
4487 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4490 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4492 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4495 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4496 (dw_val_node): Add val_symbolic_view.
4497 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4498 (symview_upper_bound): New.
4499 (new_line_info_table): Initialize symviews_since_reset.
4500 (dwarf2out_source_line): Count symviews_since_reset and set
4501 symview_upper_bound.
4502 (dw_val_equal_p): Handle symview.
4503 (add_AT_symview): New.
4504 (print_dw_val): Handle symview.
4505 (attr_checksum, attr_checksum_ordered): Likewise.
4506 (same_dw_val_p, size_of_die): Likewise.
4507 (value_format, output_die): Likewise.
4508 (add_high_low_attributes): Use add_AT_symview for entry_view.
4509 (dwarf2out_finish): Reset symview_upper_bound, clear
4512 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4515 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4516 Add strict argument and use it.
4517 (rs6000_split_multireg_move): Update for new strict argument.
4518 (mem_operand_gpr): Disallow all non-offsettable addresses.
4519 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4521 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4524 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4525 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4526 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4529 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4530 decl, use remap_type if we want to use the type.
4532 2018-03-09 Martin Sebor <msebor@redhat.com>
4534 PR tree-optimization/84526
4535 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4537 (builtin_access::generic_overlap): Be prepared to handle non-array
4540 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4542 PR rtl-optimization/84682
4543 * lra-constraints.c (process_address_1): Check is_address flag
4544 for address constraints.
4545 (process_alt_operands): Likewise.
4546 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4547 preprocess_constraints.
4548 * recog.h (preprocess_constraints): Add oploc parameter.
4550 * recog.c (preprocess_constraints): Test address_operand for
4551 CT_ADDRESS constraints.
4553 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4556 * lra-assigns.c (assign_by_spills): Return a flag of reload
4557 assignment failure. Do not process the reload assignment
4558 failures. Do not spill other reload pseudos if they has the same
4560 (lra_assign): Add a return arg. Set up from the result of
4561 assign_by_spills call.
4562 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4563 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4564 usage_insns if it is not NULL.
4565 (spill_hard_reg_in_range): New function.
4566 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4567 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4568 function prototypes.
4569 (lra_assign): Change prototype.
4570 * lra.c (lra): Add code to deal with fails by splitting hard reg
4573 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4576 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4577 Accept complain bool parameter. Only emit errors if it is true.
4578 (arm_parse_cpu_option_name): Likewise.
4579 (arm_target_thumb_only): Adjust callers of the above.
4580 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4581 prototype to take a default true bool parameter.
4582 (arm_parse_arch_option_name): Likewise.
4584 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4585 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4589 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4590 * configure: Regenerate.
4591 * configure.ac ("linker --version-script option"): New.
4592 ("linker soname option"): New.
4594 2018-03-09 Richard Biener <rguenther@suse.de>
4596 PR tree-optimization/84775
4597 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4598 immediate uses of predicate stmts and mark them modified.
4601 PR tree-optimization/84178
4602 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4604 (version_loop_for_if_conversion): Delay update_ssa call.
4605 (tree_if_conversion): Delay update_ssa until after predicate
4608 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4611 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4612 when the function accesses prior frames.
4614 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4617 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4618 gen_llsym, otherwise call maybe_gen_llsym.
4621 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4622 has ',' character inside of it.
4624 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4627 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4628 as clobbering CC_REGNUM.
4630 2018-03-08 Richard Biener <rguenther@suse.de>
4633 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4634 (follow_copies_to_constant): Do not follow SSA names registered
4637 2018-03-08 Richard Biener <rguenther@suse.de>
4639 PR tree-optimization/84178
4640 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4642 (version_loop_for_if_conversion): Delay update_ssa call.
4643 (tree_if_conversion): Delay update_ssa until after predicate
4646 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4648 PR tree-optimization/84178
4649 * tree-if-conv.c (release_bb_predicate): Remove the
4650 the assertion that the stmts have NULL use_ops.
4651 Discard the statements, asserting that they haven't
4652 yet been added to a BB.
4654 2018-03-08 Richard Biener <rguenther@suse.de>
4656 PR tree-optimization/84746
4657 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4658 (phi_translate): Pass in destination ANTIC_OUT set.
4659 (phi_translate_1): Likewise. For a simplified result lookup
4660 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4661 (phi_translate_set): Adjust.
4662 (do_pre_regular_insertion): Likewise.
4663 (do_pre_partial_partial_insertion): Likewise.
4665 2018-03-08 Martin Liska <mliska@suse.cz>
4667 PR gcov-profile/84735
4668 * doc/gcov.texi: Document usage of profile files.
4669 * gcov-io.h: Document changes in the format.
4671 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4675 * dwarf2out.c (struct dw_line_info_table): Update comments for
4677 (FORCE_RESET_NEXT_VIEW): New.
4678 (FORCE_RESETTING_VIEW_P): New.
4679 (RESETTING_VIEW_P): Check for -1 too.
4680 (ZERO_VIEW_P): Likewise.
4681 (new_line_info_table): Force-reset next view.
4682 (dwarf2out_begin_function): Likewise.
4683 (dwarf2out_source_line): Simplify zero_view_p initialization.
4684 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4685 view directly. Omit view when omitting .loc at line 0.
4687 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4689 PR tree-optimization/84740
4690 * tree-switch-conversion.c (process_switch): Call build_constructors
4691 only if info.phi_count is non-zero.
4693 PR tree-optimization/84739
4694 * tree-tailcall.c (find_tail_calls): Check call arguments against
4695 DECL_ARGUMENTS (current_function_decl) rather than
4696 DECL_ARGUMENTS (func) when checking for tail recursion.
4698 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4700 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4701 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4702 Volker Reichelt's entry and add entries for people that perform
4703 GCC fuzzy testing and report numerous bugs.
4705 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4708 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4709 readonly data in sdata, if that is disabled.
4710 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4711 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4712 -mreadonly-in-sdata option.
4714 2018-03-07 Martin Sebor <msebor@redhat.com>
4716 PR tree-optimization/84468
4717 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4718 basic block when looking for nul assignment.
4720 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4723 * except.h (output_function_exception_table): Adjust prototype.
4724 * except.c (output_function_exception_table): Remove FNNAME parameter
4725 and add SECTION parameter. Ouput one part of the table at a time.
4726 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4727 the first part of the exception table and emit unwind directives.
4728 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4729 (i386_pe_seh_cold_init): Likewise.
4730 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4731 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4732 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4733 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4734 * config/i386/winnt.c: Include except.h.
4735 (struct seh_frame_state): Add reg_offset, after_prologue and
4736 in_cold_section fields.
4737 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4738 (i386_pe_seh_cold_init): New function.
4739 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4740 to seh->in_cold_section.
4741 (seh_emit_push): Record the offset of the push.
4742 (seh_emit_save): Record the offet of the save.
4743 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4744 Test seh->after_prologue to disregard the epilogue.
4745 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4746 (i386_pe_end_cold_function): New function.
4748 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4751 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4752 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4755 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4757 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4758 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4761 * multiple_target.c: Include tree-inline.h and intl.h.
4762 (expand_target_clones): Diagnose and fail if node->definition and
4763 !tree_versionable_function_p (node->decl).
4765 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4767 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4769 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4770 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4771 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4773 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4776 * combine.c (try_combine): Use reg_or_subregno instead of handling
4777 just paradoxical SUBREGs and REGs.
4779 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4781 * config/arc/arc.c (arc_finalize_pic): Remove function.
4782 (arc_must_save_register): We use single base PIC register, remove
4783 checks to save/restore the PIC register.
4784 (arc_expand_prologue): Likewise.
4785 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4787 (arc_verify_short): Likewise.
4788 (arc_attr_type): Likewise.
4789 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4790 (walk_stores): Likewise.
4791 (arc_address_cost): Make it static.
4792 (arc_verify_short): Likewise.
4793 (branch_dest): Likewise.
4794 (arc_attr_type): Likewise.
4795 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4796 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4797 (arc_final_prescan_insn): Remove inserting the nops due to
4798 hardware hazards. It is done in reorg step.
4799 (insn_length_variant_t): Remove.
4800 (insn_length_parameters_t): Likewise.
4801 (arc_insn_length_parameters): Likewise.
4802 (arc_get_insn_variants): Likewise.
4803 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4805 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4808 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4811 PR tree-optimization/84687
4812 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4814 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4816 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4818 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4819 Rename to ppc_speculation_barrier.
4820 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4821 __builtin_ppc_speculation_barrier.
4823 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4826 * combine.c (combine_simplify_rtx): Don't try to simplify if
4827 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4830 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4832 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4833 to 32 bytes when compiling for POWER9.
4835 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4838 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4839 regparm >= 3 with no arg reg available also for calls with
4840 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4843 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4845 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4847 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4850 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4852 2018-03-05 Richard Biener <rguenther@suse.de>
4854 PR tree-optimization/84486
4855 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4856 When inserting a __builtin_assume_aligned call set the LHS
4857 SSA name alignment info accordingly.
4859 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4861 PR tree-optimization/84114
4862 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4863 Avoid reassociation of FLOAT_MODE addition.
4865 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4867 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4868 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4869 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4870 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4871 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4872 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4873 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4875 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4876 __builtin_ia32_wbinvd): New builtins.
4877 (SPECIAL_ARGS2): New.
4878 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4879 (SPECIAL_ARGS2): New.
4880 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4881 (ix86_valid_target_attribute_inner_p): Ditto.
4882 (ix86_init_mmx_sse_builtins): Add special_args2.
4883 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4884 TARGET_WBNOINVD_P): New.
4885 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4886 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4887 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4888 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4889 * config/i386/pconfigintrin.h: New file.
4890 * config/i386/wbnoinvdintrin.h: Ditto.
4891 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4892 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4894 2018-03-05 Richard Biener <rguenther@suse.de>
4896 PR tree-optimization/84670
4897 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4899 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4900 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4901 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4902 doesn't grow if all successors (recursively) were visited at least
4905 2018-03-05 Richard Biener <rguenther@suse.de>
4907 PR tree-optimization/84650
4908 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4909 if executed in the loop pipeline.
4911 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4913 * doc/configfiles.texi (Configuration Files): Move info about
4914 conditionalizing $target-protos.h to...
4915 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4916 differs from $target-protos.h.
4918 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4919 Chung-Ju Wu <jasonwucj@gmail.com>
4921 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4922 * config/nds32/nds32-multiple.md (setmemsi): Define.
4923 * config/nds32/nds32-memory-manipulation.c
4924 (nds32_gen_dup_4_byte_to_word_value): New.
4925 (emit_setmem_word_loop): New.
4926 (emit_setmem_byte_loop): New.
4927 (nds32_expand_setmem_loop): New.
4928 (nds32_expand_setmem_loop_v3m): New.
4929 (nds32_expand_setmem_unroll): New.
4930 (nds32_expand_setmem): New.
4932 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4933 Chung-Ju Wu <jasonwucj@gmail.com>
4935 * config/nds32/nds32-memory-manipulation.c
4936 (nds32_emit_load_store): New.
4937 (nds32_emit_post_inc_load_store): New.
4938 (nds32_emit_mem_move): New.
4939 (nds32_emit_mem_move_block): New.
4940 (nds32_expand_movmemsi_loop_unknown_size): New.
4941 (nds32_expand_movmemsi_loop_known_size): New.
4942 (nds32_expand_movmemsi_loop): New.
4943 (nds32_expand_movmemsi_unroll): New.
4944 (nds32_expand_movmemqi): Rename ...
4945 (nds32_expand_movmemsi): ... to this.
4946 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4947 (movmemsi): ... to this.
4948 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4949 (nds32_expand_movmemsi): ... to this.
4951 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4952 Monk Chiang <sh.chiang04@gmail.com>
4953 Chung-Ju Wu <jasonwucj@gmail.com>
4955 * config/nds32/nds32-protos.h
4956 (nds32_expand_load_multiple): New arguments.
4957 (nds32_expand_store_multiple): Ditto.
4958 (nds32_valid_multiple_load_store): Rename ...
4959 (nds32_valid_multiple_load_store_p): ... to this.
4960 * config/nds32/nds32-memory-manipulation.c
4961 (nds32_expand_load_multiple): Refine implementation.
4962 (nds32_expand_store_multiple): Ditto.
4963 * config/nds32/nds32-multiple.md
4964 (load_multiple): Update nds32_expand_load_multiple interface.
4965 (store_multiple): Update nds32_expand_store_multiple interface.
4966 * config/nds32/nds32-predicates.c
4967 (nds32_valid_multiple_load_store): Rename ...
4968 (nds32_valid_multiple_load_store_p): ... to this and refine
4970 * config/nds32/predicates.md
4971 (nds32_load_multiple_and_update_address_operation): New predicate.
4972 (nds32_store_multiple_and_update_address_operation): New predicate.
4974 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4975 Chung-Ju Wu <jasonwucj@gmail.com>
4977 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4978 (combo): New attribute.
4979 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4981 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4983 * config/nds32/nds32.opt: Change -mcmodel= default value.
4985 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4986 Monk Chiang <sh.chiang04@gmail.com>
4987 Chung-Ju Wu <jasonwucj@gmail.com>
4989 * config/nds32/constants.md (unspec_element): New enum.
4990 * config/nds32/constraints.md (Umw): New constraint.
4991 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4992 * config/nds32/nds32-intrinsic.md: Likewise.
4993 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4994 (nds32_valid_smw_lwm_base_p): New.
4995 (nds32_output_smw_single_word): New.
4996 (nds32_output_lmw_single_word): New.
4997 (nds32_expand_unaligned_load): New.
4998 (nds32_expand_unaligned_store): New.
4999 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5000 (nds32_output_smw_single_word): Declare.
5001 (nds32_output_lmw_single_word): Declare.
5002 (nds32_expand_unaligned_load): Declare.
5003 (nds32_expand_unaligned_store): Declare.
5004 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5005 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5006 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5007 NDS32_BUILTIN_UASTORE_DW.
5008 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5011 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5012 Kito Cheng <kito.cheng@gmail.com>
5013 Chung-Ju Wu <jasonwucj@gmail.com>
5015 * config/nds32/nds32-intrinsic.c
5016 (nds32_expand_builtin_null_ftype_reg): Delete.
5017 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5018 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5019 (nds32_read_argument): New.
5020 (nds32_legitimize_target): Ditto.
5021 (nds32_legitimize_argument): Ditto.
5022 (nds32_check_constant_argument): Ditto.
5023 (nds32_expand_unop_builtin): Ditto.
5024 (nds32_expand_unopimm_builtin): Ditto.
5025 (nds32_expand_binop_builtin): Ditto.
5026 (nds32_builtin_decl_impl): Ditto.
5027 (builtin_description): Ditto.
5028 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5029 (nds32_init_builtins_impl): Ditto.
5030 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5031 (nds32_builtin_decl): New.
5032 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5033 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5035 2018-03-02 Jeff Law <law@redhat.com>
5037 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5038 (redundant_insn, fill_simple_delay_slots): Likewise.
5039 (fill_slots_from_thread): Likewise.
5040 * resource.c (mark_referenced_resources): Likewise.
5041 (mark_set_resources, find_dead_or_set_registers): Likewise.
5043 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5045 * substring-locations.h (format_warning_va): Formatting fix for
5047 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5049 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5050 * substring-locations.c: Include intl.h.
5051 (format_warning_va): Turned into small wrapper around
5052 format_warning_n_va, renamed to ...
5053 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5054 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5056 (format_warning_at_substring_n): New function.
5057 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5058 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5059 format_warning_at_substring with just a shorter name instead of
5060 const function pointer.
5061 (fmtwarn_n): New function.
5062 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5063 appropriate, get rid of all the fmtstr temporaries, move conditionals
5064 with G_() wrapped string literals directly into fmtwarn arguments,
5065 cast dir.len to (int), formatting fixes.
5067 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5069 * doc/invoke.texi: Remove "Cilk Plus" references.
5071 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5072 Richard Biener <rguenther@suse.de>
5075 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5076 for error or warning attributes if CALL_FROM_THUNK_P is set.
5079 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5082 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5083 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5086 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5087 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5089 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5090 instead of -1U in last predictors element's probability member.
5092 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5095 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5096 arguments if they are comparable.
5098 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5100 PR tree-optimization/84634
5101 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5102 masks and masked_loop_p with a single loop_masks, making sure it's
5103 null for bb vectorization.
5105 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5107 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5108 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5109 loop->force_vectorize to check whether there is no alias.
5111 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5114 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5116 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5118 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5119 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5120 instead of a loop around prev_real_insn.
5121 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5125 * config/i386/i386.c (ix86_print_operand): Use conditional
5126 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5129 2018-03-02 Richard Biener <rguenther@suse.de>
5131 PR tree-optimization/84427
5132 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5133 (bitmap_set_subtract_values): Rewrite to handle multiple
5136 (prune_clobbered_mems): Likewise.
5137 (phi_translate): Take edge instead of pred/phiblock.
5138 (phi_translate_1): Likewise.
5139 (phi_translate_set): Likewise. Insert all translated
5140 exprs for a value into the set, keeping possibly multiple
5141 expressions per value.
5142 (compute_antic_aux): Adjust for phi_translate changes.
5143 When intersecting union the expressions and prune those
5144 not in the final value set, keeping possibly multiple
5145 expressions per value. Do not use value-insertion
5146 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5147 all expressions. Add verification that the value-sets
5148 only shrink during iteration.
5149 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5150 (do_pre_regular_insertion): Likewise.
5151 (do_pre_partial_partial_insertion): Likewise.
5153 2018-03-02 Richard Biener <rguenther@suse.de>
5156 * config/darwin.c (saved_debug_info_level): New static global.
5157 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5158 (darwin_asm_lto_end): Restore debug info generation settings.
5160 2018-03-01 Martin Liska <mliska@suse.cz>
5163 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5166 2018-03-01 Richard Biener <rguenther@suse.de>
5169 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5170 type annotation with LTO when debug was disabled at compile-time.
5172 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5174 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5176 (mips_final_postscan_insn): Likewise.
5178 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5180 PR rtl-optimization/84528
5181 * alias.c (init_alias_target): Add commentary.
5182 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5183 a unique base value if the frame pointer is not eliminated
5184 to the stack pointer.
5186 2018-03-01 Tom de Vries <tom@codesourcery.com>
5188 PR rtl-optimization/83327
5189 * lra-int.h (hard_regs_spilled_into): Declare.
5190 * lra.c (hard_regs_spilled_into): Define.
5191 (init_reg_info): Init hard_regs_spilled_into.
5192 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5193 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5194 (process_bb_lives): Handle hard_regs_spilled_into.
5195 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5196 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5198 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5200 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5201 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5202 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5203 * config/rs6000/aix72.h: New file.
5205 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5207 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5208 instead of warning_at with conditional singular and plural messages
5212 * stor-layout.c (update_alignment_for_field): For
5213 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5214 && !DECL_PACKED (field), do the alignment update, just use
5215 only desired_align instead of MAX (type_align, desired_align)
5217 (place_field): Don't do known_align < desired_align handling
5218 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5219 is non-NULL, instead do it after rli->prev_field handling and
5220 only if not within a bitfield word. For DECL_PACKED (field)
5221 use type_align of BITS_PER_UNIT.
5223 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5225 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5226 superfluous parentheses and trailing spaces.
5228 2018-02-28 Richard Biener <rguenther@suse.de>
5230 PR tree-optimization/84584
5231 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5232 SCoPs with fake exit edge.
5234 2018-02-28 Martin Liska <mliska@suse.cz>
5237 * timevar.c (timer::print): Fix format to properly print 100%
5240 2018-02-28 Richard Biener <rguenther@suse.de>
5243 * genmatch.c (capture_info::walk_match): Do not mark
5244 captured expressions without operands as expr_p given
5245 they act more like predicates and should be subject to
5246 "lost tail" side-effect preserving.
5248 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5250 PR rtl-optimization/81611
5251 * auto-inc-dec.c (attempt_change): Move dead note from
5252 mem_insn if it's the next use of regno
5253 (find_address): Take address use of reg holding
5254 non-incremented value. Add parm to limit search to the named
5256 (merge_in_block): Attempt to use a mem insn that is the next
5257 use of the original regno.
5259 2018-02-27 Martin Sebor <msebor@redhat.com>
5262 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5263 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5265 2018-02-27 Martin Sebor <msebor@redhat.com>
5267 PR translation/84207
5268 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5269 n argument to unsigned HOST_WIDE_INT.
5270 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5271 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5272 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5273 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5275 2018-02-27 Richard Biener <rguenther@suse.de>
5277 PR tree-optimization/84512
5278 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5279 Do not use the estimate returned from record_stmt_cost for
5280 the scalar iteration cost but sum properly using add_stmt_cost.
5282 2018-02-27 Richard Biener <rguenther@suse.de>
5284 PR tree-optimization/84466
5285 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5286 Adjust last change to less strictly validate use operands.
5288 2018-02-27 Martin Liska <mliska@suse.cz>
5290 PR gcov-profile/84548
5291 * gcov.c (process_file): Allow partial overlap and consider it
5292 also as group functions.
5293 (output_lines): Properly calculate range of lines for a group.
5295 2018-02-27 Martin Liska <mliska@suse.cz>
5297 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5298 'ggc' suffixes. Change first column width.
5299 (timer::print): Fix formatting of the column.
5301 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5303 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5304 preserve inline entry blocks for the sake of debug inline
5305 entry point markers alone.
5306 (remove_unused_locals): Suggest in comments a better place to
5307 force the preservation of inline entry blocks that are
5308 otherwise unused, but do not preserve them.
5310 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5312 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5314 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5317 * config/i386/constraints.md (Bs): Replace
5318 ix86_indirect_branch_register with
5319 TARGET_INDIRECT_BRANCH_REGISTER.
5321 * config/i386/i386.md (indirect_jump): Likewise.
5322 (tablejump): Likewise.
5323 (*sibcall_memory): Likewise.
5324 (*sibcall_value_memory): Likewise.
5325 Peepholes of indirect call and jump via memory: Likewise.
5326 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5327 (*sibcall_value_GOT_32): Likewise.
5328 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5329 (GOT_memory_operand): Likewise.
5330 (call_insn_operand): Likewise.
5331 (sibcall_insn_operand): Likewise.
5332 (GOT32_symbol_operand): Likewise.
5333 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5335 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5337 PR rtl-optimization/83496
5338 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5339 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5340 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5341 redundant insn, if any.
5342 (relax_delay_slots): Likewise.
5343 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5345 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5347 PR tree-optimization/83965
5348 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5349 that grouped statements are part of a reduction chain. Return
5350 true if the statement is not marked as a reduction itself but
5352 (vect_recog_dot_prod_pattern): Don't check whether the statement
5353 is part of a group here.
5354 (vect_recog_sad_pattern): Likewise.
5355 (vect_recog_widen_sum_pattern): Likewise.
5357 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5360 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5362 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5365 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5367 (ix86_output_indirect_function_return): New prototype.
5368 (ix86_split_simple_return_pop_internal): Likewise.
5369 * config/i386/i386.c (indirect_return_via_cx): New.
5370 (indirect_return_via_cx_bnd): Likewise.
5371 (indirect_thunk_name): Handle return va CX_REG.
5372 (output_indirect_thunk_function): Create alias for
5373 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5374 (ix86_output_indirect_jmp): Remove the bool argument.
5375 (ix86_output_indirect_function_return): New function.
5376 (ix86_split_simple_return_pop_internal): Likewise.
5377 * config/i386/i386.md (*indirect_jump): Don't pass false
5378 to ix86_output_indirect_jmp.
5379 (*tablejump_1): Likewise.
5380 (simple_return_pop_internal): Change it to define_insn_and_split.
5381 Call ix86_split_simple_return_pop_internal to split it for
5383 (simple_return_indirect_internal): Call
5384 ix86_output_indirect_function_return instead of
5385 ix86_output_indirect_jmp.
5387 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5390 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5391 memset and value initialization afterwards.
5393 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5395 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5397 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5400 * common/config/aarch64/aarch64-common.c
5401 (aarch_option_optimization_table[]): Switch
5402 off fomit-frame-pointer
5404 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5405 Chung-Ju Wu <jasonwucj@gmail.com>
5407 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5409 (store_multiple): Ditto.
5411 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5413 * config.gcc: Add --with-cpu support for nds32 target.
5414 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5415 * config/nds32/nds32.opt: Add -mcpu= option.
5417 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5419 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5420 isel=yes): Warn for these deprecated options.
5422 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5424 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5425 ISA_2_5_MASKS_EMBEDDED.
5427 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5429 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5430 p->max as pointers rather than using iterative_hash_expr.
5432 2018-02-23 Carl Love <cel@us.ibm.com>
5434 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5435 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5437 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5438 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5439 P8V_BUILTIN_VEC_VUNSIGNED2.
5441 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5444 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5445 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5447 * lra-constraints.c (curr_insn_transform): Set up
5448 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5450 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5451 * lra-eliminations.c (spill_pseudos): Ditto.
5452 (process_insn_for_elimination): Ditto.
5453 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5454 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5455 LRA_NON_CLOBBERED_ALT.
5457 2018-02-22 Martin Sebor <msebor@redhat.com>
5459 PR tree-optimization/84480
5460 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5461 to maybe_diag_stxncpy_trunc. Call it.
5462 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5463 from gimple_fold_builtin_strcpy. Print inlining stack.
5464 (handle_builtin_stxncpy): Print inlining stack.
5465 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5467 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5470 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5471 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5472 and -fcheck-pointer-bounds are used together.
5473 (indirect_thunk_prefix): New enum.
5474 (indirect_thunk_need_prefix): New function.
5475 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5476 "_nt" instead of "_bnd" for NOTRACK prefix.
5477 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5478 (output_indirect_thunk_function): Likewise.
5480 (ix86_code_end): Update output_indirect_thunk_function calls.
5481 (ix86_output_indirect_branch_via_reg): Replace
5482 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5483 (ix86_output_indirect_branch_via_push): Likewise.
5484 (ix86_output_function_return): Likewise.
5485 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5486 incompatible with -fcf-protection=branch and
5487 -fcheck-pointer-bounds.
5489 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5492 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5493 Change gcc_assert call to output_operand_lossage.
5495 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5497 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5499 2018-02-22 DJ Delorie <dj@redhat.com>
5500 Sebastian Perta <sebastian.perta@renesas.com>
5501 Oleg Endo <olegendo@gcc.gnu.org>
5503 * config/rx/rx.c (rx_rtx_costs): New function.
5504 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5506 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5508 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5510 2018-02-22 Martin Liska <mliska@suse.cz>
5513 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5514 Add "native" as a possible value.
5516 2018-02-22 Martin Liska <mliska@suse.cz>
5519 * config/i386/i386.c (ix86_option_override_internal):
5520 Add "native" as a possible value for -march and -mtune.
5522 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5525 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5526 to all type variants.
5528 PR tree-optimization/84503
5529 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5530 width as info->bitpos + info->bitsize - start.
5531 (merged_store_group::merge_overlapping): Simplify width computation.
5532 (check_no_overlap): New function.
5533 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5534 start + width and last_order of the group, fail if check_no_overlap
5536 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5537 to group if check_no_overlap fails.
5539 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5541 * config/rs6000/altivec.md: Delete contraint arguments to
5542 define_expand, define_split, and define_peephole2, and in
5543 define_insn_and_split if always unused.
5544 * config/rs6000/darwin.md: Ditto.
5545 * config/rs6000/dfp.md: Ditto.
5546 * config/rs6000/rs6000.md: Ditto.
5547 * config/rs6000/sync.md: Ditto.
5548 * config/rs6000/vector.md: Ditto.
5549 * config/rs6000/vsx.md: Ditto.
5551 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5553 * config/rs6000/altivec.md: Write output control strings as braced
5554 blocks instead of double-quoted strings.
5555 * config/rs6000/darwin.md: Ditto.
5556 * config/rs6000/rs6000.md: Ditto.
5557 * config/rs6000/vector.md: Ditto.
5558 * config/rs6000/vsx.md: Ditto.
5560 2018-02-21 Jason Merrill <jason@redhat.com>
5562 PR c++/84314 - ICE with templates and fastcall attribute.
5563 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5565 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5567 * ipa-cp.c (determine_versionability): Fix comment typos.
5569 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5572 * ipa-cp.c (determine_versionability): Do not version functions caling
5575 2018-02-21 Martin Liska <mliska@suse.cz>
5578 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5579 Add "native" as a possible value.
5580 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5581 the macro when native cpu detection is available.
5583 2018-02-21 Martin Liska <mliska@suse.cz>
5586 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5587 Add "native" as a possible value.
5588 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5589 when native cpu detection is available.
5591 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5592 Martin Sebor <msebor@redhat.com>
5594 PR tree-optimization/84478
5595 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5597 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5598 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5599 support which is conservatively correct, for 2 only stay conservative
5600 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5601 argument to the 2 argument get_range_strlen, adjust 6 arg
5602 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5604 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5605 (gimple_fold_builtin_strlen): Pass true as last argument to
5608 2018-02-20 Martin Sebor <msebor@redhat.com>
5611 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5612 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5613 (builtin_memref::builtin_memref): Factor out parts into
5614 set_base_and_offset and call it.
5616 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5619 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5620 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5621 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5622 search at the associated MODE_INT.
5624 2018-02-20 Jeff Law <law@redhat.com>
5627 PR tree-optimization/81592
5629 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5630 for range data rather than using global data.
5631 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5632 range data rather than using global data.
5633 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5634 pass it to children as needed.
5635 (struct directive::fmtresult): Similarly.
5636 (struct directive::set_width): Similarly.
5637 (struct directive::set_precision): Similarly.
5638 (format_integer, format_directive, parse_directive): Similarly.
5639 (format_none): Accept unnamed vr_values parameter.
5640 (format_percent, format_floating, format_character): Similarly.
5641 (format_string, format_plain): Similarly.
5642 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5643 the EVRP range analyzer for range data rather than using global data.
5644 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5645 gimple-ssa-evrp-analyze.h
5646 (class sprintf_dom_walker): Add after_dom_children member function.
5647 Add evrp_range_analyzer member.
5648 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5649 range analyzer as needed.
5650 (sprintf_dom_walker::after_dom_children): New member function.
5651 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5653 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5654 (evrp_range_analyzer::pop_to_marker): Likewise.
5656 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5658 PR tree-optimization/84419
5659 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5660 with the required type if its current type is compatible but
5663 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5666 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5667 after vectorization.
5669 2018-02-20 Martin Liska <mliska@suse.cz>
5672 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5673 possible values if we don't have a hint.
5675 2018-02-20 Martin Liska <mliska@suse.cz>
5679 * final.c (shorten_branches): Build align_tab array with one
5681 * opts.c (finish_options): Add alignment option limit check.
5682 (MAX_CODE_ALIGN): Likewise.
5683 (MAX_CODE_ALIGN_VALUE): Likewise.
5684 * doc/invoke.texi: Document maximum allowed option value for
5685 all -falign-* options.
5687 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5690 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5691 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5692 * var-tracking.c (emit_note_insn_var_location): Remove all references
5693 to NOTE_INSN_CALL_ARG_LOCATION.
5694 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5695 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5696 Use copy_rtx_if_shared.
5697 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5698 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5699 (dwarf2out_var_location): Remove handling of
5700 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5702 * final.c (final_scan_insn): Remove all references to
5703 NOTE_INSN_CALL_ARG_LOCATION.
5704 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5705 before dumping final insns.
5706 * except.c (emit_note_eh_region_end): Remove all references to
5707 NOTE_INSN_CALL_ARG_LOCATION.
5708 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5709 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5710 * config/arc/arc.c (hwloop_optimize): Likewise.
5711 * config/arm/arm.c (create_fix_barrier): Likewise.
5712 * config/s390/s390.c (s390_chunkify_start): Likewise.
5713 * config/sh/sh.c (find_barrier): Likewise.
5714 * config/i386/i386.c (rest_of_insert_endbranch,
5715 ix86_seh_fixup_eh_fallthru): Likewise.
5716 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5717 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5718 * config/frv/frv.c (frv_function_prologue): Likewise.
5719 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5721 (note_outside_basic_block_p): Remove all references to
5722 NOTE_INSN_CALL_ARG_LOCATION.
5723 * gengtype.c (adjust_field_rtx_def): Likewise.
5724 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5726 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5727 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5730 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5733 PR tree-optimization/84452
5734 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5735 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5738 2018-02-19 Martin Liska <mliska@suse.cz>
5741 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5743 2018-02-19 Martin Liska <mliska@suse.cz>
5744 Richard Sandiford <richard.sandiford@linaro.org>
5746 PR tree-optimization/82491
5747 * gimple-fold.c (get_base_constructor): Make earlier bail out
5750 2018-02-19 Carl Love <cel@us.ibm.com>
5752 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5753 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5755 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5756 P8V_BUILTIN_VEC_NEG.
5758 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5760 * config/rl78/rl78.md (movdf): New define expand.
5762 2018-02-19 Martin Liska <mliska@suse.cz>
5765 * doc/invoke.texi: Fix typo.
5766 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5768 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5770 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5771 handle rs6000_single_float and rs6000_double_float specially for
5774 2018-02-16 Jeff Law <law@redhat.com>
5776 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5779 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5782 * ipa-inline.c (inline_small_functions): Fix a typo.
5784 2018-02-16 Nathan Sidwell <nathan@acm.org>
5786 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5788 2018-02-16 Carl Love <cel@us.ibm.com>
5790 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5791 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5792 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5793 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5794 expansion to P8V_BUILTIN_VEC_FLOAT2.
5796 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5798 PR rtl-optimization/70023
5799 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5800 src_regno into account.
5802 2018-02-16 Carl Love <cel@us.ibm.com>
5804 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5805 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5806 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5807 * config/rs6000/rs6000.c: Remove case statements for
5808 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5809 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5810 and P9V_BUILTIN_VEC_VINSERT4B.
5811 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5812 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5813 * config/rs6000/vsx.md:
5814 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5817 2018-02-16 Carl Love <cel@us.ibm.com>
5819 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5821 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5823 * config/rs6000/rs6000-c.c: Add the definitions for
5824 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5825 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5826 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5827 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5828 definition for insert4b and define insn *insert3b_internal.
5829 * doc/extend.texi: Add documentation for vec_extract4b.
5831 2018-02-16 Nathan Sidwell <nathan@acm.org>
5833 * doc/extend.texi (Backwards Compatibility): Mention friend
5834 injection. Note for-scope is deprecated.
5835 * doc/invoke.texi (-ffriend-injection): Deprecate.
5837 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5839 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5840 that moved to I2, also allow destinations that are a paradoxical
5841 subreg (instead of a normal reg).
5843 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5846 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5849 2018-02-16 Richard Biener <rguenther@suse.de>
5851 PR tree-optimization/84037
5852 PR tree-optimization/84016
5854 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5855 Adjust vec_construct for the fact we need additional higher latency
5856 128bit inserts for AVX256 and AVX512 vector builds.
5857 (ix86_add_stmt_cost): Scale vector construction cost for
5860 2018-02-16 Richard Biener <rguenther@suse.de>
5862 PR tree-optimization/84417
5863 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5864 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5865 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5867 2018-02-16 Martin Liska <mliska@suse.cz>
5870 * internal-fn.def (ASAN_CHECK): Set proper flags.
5871 (ASAN_MARK): Likewise.
5873 2018-02-16 Julia Koval <julia.koval@intel.com>
5875 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5876 from PTA_CANNONLAKE.
5878 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5881 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5882 Use ++iter rather than iter++ for std::list iterators.
5883 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5884 defer deleting them until all nodes in the forest are processed. Do
5885 free even leaf nodes. Change to_process into auto_vec.
5888 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5889 * vec.h (vec_default_construct): Use memset instead of placement new
5890 if BROKEN_VALUE_INITIALIZATION is defined.
5891 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5892 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5895 PR rtl-optimization/83723
5896 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5897 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5898 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5900 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5902 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5904 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5906 PR rtl-optimization/81443
5907 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5908 from inner REGs to paradoxical SUBREGs.
5910 2018-02-16 Richard Biener <rguenther@suse.de>
5912 PR tree-optimization/84399
5913 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5914 For operands we can analyze at their definition make sure we can
5915 analyze them at each use as well.
5917 2018-02-16 Richard Biener <rguenther@suse.de>
5919 PR tree-optimization/84190
5920 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5921 volatile accesses if the decl isn't volatile.
5923 2018-02-15 Jason Merrill <jason@redhat.com>
5925 PR c++/84314 - ICE with templates and fastcall attribute.
5926 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5927 TYPE_CANONICAL on an existing type.
5929 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5931 PR tree-optimization/84383
5932 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5933 dstoff nor call operand_equal_p if dstbase is NULL.
5935 PR tree-optimization/84334
5936 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5937 also a CONSTANT_CLASS_P, punt.
5939 2018-02-14 Jim Wilson <jimw@sifive.com>
5941 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5942 first SMALL_OPERAND check. New local min_second_step. Move assert
5943 to where locals are set. Add TARGET_RVC support.
5944 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5946 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5948 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5950 2018-02-14 Martin Sebor <msebor@redhat.com>
5952 PR tree-optimization/83698
5953 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5954 arrays constrain the offset range to their bounds.
5955 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5956 (builtin_access::overlap): Avoid setting the size of overlap if it's
5958 (maybe_diag_overlap): Also consider arrays when deciding what values
5959 of offsets to include in diagnostics.
5961 2018-02-14 Martin Sebor <msebor@redhat.com>
5964 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5965 that correspond to the kind of a declaration.
5967 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5970 * config/pa/pa.md: Load address of PIC label using the linkage table
5971 if the label is nonlocal.
5973 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5975 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5976 warning message if user requests -maltivec=be.
5977 * doc/invoke.texi: Document deprecation of -maltivec=be.
5979 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5982 * config/rs6000/rs6000-c.c: Update definitions for
5983 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5984 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5986 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5989 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5990 add _get_ssp intrinsics. Remove argument from
5991 __builtin_ia32_rdssp[d|q].
5992 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5993 * config/i386/i386-builtin.def: Remove argument from
5994 __builtin_ia32_rdssp[d|q].
5995 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5996 ix86_expand_special_args_builtin for _rdssp[d|q].
5997 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5998 Clear register before usage.
5999 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6000 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6002 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6004 PR tree-optimization/84357
6005 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6006 operand 1 of an ARRAY_REF too.
6008 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6011 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6012 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6014 (set_of_reg): New struct.
6015 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6016 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6017 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6019 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6020 Split into bitclr, bitset, bitinvert patterns if appropriate.
6021 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6022 use rx_fuse_in_memory_bitop.
6023 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6024 to named insn, correct maximum insn length.
6026 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6029 * machmode.def: Define a complex mode for PARTIAL_INT.
6030 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6032 * doc/rtl.texi: Document CSPImode.
6033 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6035 (msp430_hard_regno_nregs_with_padding): Likewise.
6037 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6040 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6042 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6044 PR rtl-optimization/84169
6045 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6046 we generated a parallel as new i3 and we split that to new i2 and i3
6047 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6048 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6049 those to i2, not i1. Partially rewrite this scan code.
6051 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6054 * stor-layout.c (place_field): For variable length fields, adjust
6055 offset_align afterwards not just based on the field's alignment,
6056 but also on the size.
6059 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6060 of exps and logs in the use_exp2 case.
6062 2018-02-13 Jeff Law <law@redhat.com>
6064 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6067 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6070 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6074 * common.opt (gas-loc-support, gas-locview-support): New.
6075 (ginline-points, ginternal-reset-location-views): New.
6076 * doc/invoke.texi: Document them. Use @itemx where intended.
6077 (gvariable-location-views): Adjust.
6078 * target.def (reset_location_view): New.
6079 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6080 (TARGET_RESET_LOCATION_VIEW): New.
6081 * doc/tm.texi: Rebuilt.
6082 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6083 (dwarf2out_default_as_locview_support): New.
6084 (output_asm_line_debug_info): Use option variables.
6085 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6086 (output_loc_list): Likewise.
6087 (add_high_low_attributes): Check option variables.
6088 Don't output entry view attribute in strict mode.
6089 (gen_inlined_subroutine_die): Check option variables.
6090 (dwarf2out_inline_entry): Likewise.
6091 (init_sections_and_labels): Likewise.
6092 (dwarf2out_early_finish): Likewise.
6093 (maybe_reset_location_view): New, from...
6094 (dwarf2out_var_location): ... here. Call it.
6095 * debug.h (dwarf2out_default_as_loc_support): Declare.
6096 (dwarf2out_default_as_locview_support): Declare.
6097 * hooks.c (hook_int_rtx_insn_0): New.
6098 * hooks.h (hook_int_rtx_insn_0): Declare.
6099 * toplev.c (process_options): Take -gas-loc-support and
6100 -gas-locview-support from dwarf2out. Enable
6101 -gvariable-location-views by default only with locview
6102 assembler support. Enable -ginternal-reset-location-views by
6103 default only if the target defines the corresponding hook.
6104 Enable -ginline-points by default if location views are
6105 enabled; force it disabled if statement frontiers are
6107 * tree-inline.c (expand_call_inline): Check option variables.
6108 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6110 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6112 PR tree-optimization/84321
6113 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6114 handling. Also check whether the anti-range contains any values
6115 that satisfy the mask; switch to a VR_RANGE if not.
6117 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6120 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6122 2018-02-13 Martin Jambor <mjambor@suse.cz>
6125 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6126 of call statements, also set location of a load to a temporary.
6128 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6130 * config/rl78/rl78.c (add_vector_labels): New function.
6131 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6132 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6133 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6134 which checks that no arguments are passed.
6135 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6136 * doc/extend.texi: Documentation for the new attribute.
6138 2018-02-13 Andreas Schwab <schwab@suse.de>
6140 * config/riscv/linux.h (CPP_SPEC): Define.
6142 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6145 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6146 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6147 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6148 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6149 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6150 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6151 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6152 temporarily for AES and PCLMUL builtins.
6154 PR tree-optimization/84339
6155 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6156 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6160 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6161 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6162 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6164 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6166 * omp-simd-clone.h: New file.
6167 * omp-simd-clone.c: Include omp-simd-clone.h.
6168 (expand_simd_clones): No longer static.
6169 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6170 cgraph.h and omp-simd-clone.h.
6171 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6172 (vect_recog_widen_shift_pattern): Formatting fix.
6173 (vect_pattern_recog_1): Don't check optab for calls.
6176 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6177 operands[2] into a REG before using gen_lowpart on it.
6179 2018-02-12 Jeff Law <law@redhat.com>
6182 * config/sh/sh.c (find_barrier): Consider a sibling call
6185 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6186 successfully back substituting a reg.
6188 2018-02-12 Richard Biener <rguenther@suse.de>
6190 PR tree-optimization/84037
6191 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6192 parameter, move visited init to caller.
6193 (vect_slp_analyze_operations): Separate cost from validity
6194 check, initialize visited once for all instances.
6195 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6197 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6199 (vect_model_promotion_demotion_cost): Likewise.
6200 (vectorizable_bswap): Guard cost modeling with !slp_node
6201 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6203 (vectorizable_call): Likewise.
6204 (vectorizable_conversion): Likewise.
6205 (vectorizable_assignment): Likewise.
6206 (vectorizable_shift): Likewise.
6207 (vectorizable_operation): Likewise.
6208 (vectorizable_store): Likewise.
6209 (vectorizable_load): Likewise.
6210 (vectorizable_condition): Likewise.
6211 (vectorizable_comparison): Likewise.
6213 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6216 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6217 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6218 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6220 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6223 * params.def (inline-min-speedup): Increase from 8 to 15.
6224 (max-inline-insns-auto): Decrease from 40 to 30.
6225 * ipa-split.c (consider_split): Add some buffer for function to
6226 be considered inlining candidate.
6227 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6230 2018-02-12 Richard Biener <rguenther@suse.de>
6232 PR tree-optimization/84037
6233 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6234 matched stmts if we cannot swap the non-matched ones.
6236 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6238 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6239 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6240 _mm_maskz_scalef_round_ss): New intrinsics.
6241 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6242 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6243 __builtin_ia32_scalefss_round): Remove.
6244 (__builtin_ia32_scalefsd_mask_round,
6245 __builtin_ia32_scalefss_mask_round): New intrinsics.
6246 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6247 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6248 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6249 "<round_constraint>")): Changed to ...
6250 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6251 "<round_scalar_constraint>")): ... this.
6252 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6253 %0, %1, %2<round_op3>}"): Changed to ...
6254 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6255 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6256 %2<round_scalar_mask_op3>}"): ... this.
6257 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6259 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6261 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6262 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6263 (_mm_maskz_sqrt_round_ss): New intrinsics.
6264 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6265 (__builtin_ia32_sqrtsd_mask_round)
6266 (__builtin_ia32_sqrtss_mask_round): New builtins.
6267 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6268 (__builtin_ia32_sqrtss_round): Remove.
6269 (__builtin_ia32_sqrtsd_mask_round)
6270 (__builtin_ia32_sqrtss_mask_round): New builtins.
6271 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6272 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6273 ((match_operand:VF_128 1 "vector_operand"
6274 "xBm,<round_constraint>")): Changed to ...
6275 ((match_operand:VF_128 1 "vector_operand"
6276 "xBm,<round_scalar_constraint>")): ... this.
6277 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6278 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6279 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6280 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6281 %<iptr>1<round_scalar_mask_op3>}): ... this.
6282 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6283 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6285 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6288 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6289 Cast vec_cmpeq result to correct type.
6290 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6291 Cast vec_cmpgt result to correct type.
6293 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6295 * final.c (final_scan_insn_1): Renamed from...
6296 (final_scan_insn): ... this. New wrapper, to recover
6297 seen from the outermost call in recursive ones.
6298 * config/sparc/sparc.c (output_return): Drop seen from call.
6299 (output_sibcall): Likewise.
6300 * config/visium/visium.c (output_branch): Likewise.
6302 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6304 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6307 2018-02-10 Alan Modra <amodra@gmail.com>
6310 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6311 Specify LR as an input.
6313 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6316 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6317 remove_member_access_dummy_vars): New functions.
6318 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6319 lower_omp_1, execute_lower_omp): Use them.
6321 PR rtl-optimization/84308
6322 * shrink-wrap.c (spread_components): Release todo vector.
6324 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6326 PR rtl-optimization/57193
6327 * ira-color.c (struct allocno_color_data): Add member
6328 conflict_allocno_hard_prefs.
6329 (update_conflict_allocno_hard_prefs): New.
6330 (bucket_allocno_compare_func): Add a preference based on
6331 conflict_allocno_hard_prefs.
6332 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6333 (color_allocnos): Remove a dead code. Initiate
6334 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6336 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6339 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6340 constraint from =wa to wa. Avoid a subreg on the output operand,
6341 instead use a pseudo and subreg it in a move.
6342 (p9_xxbrd_<mode>): Changed to ...
6343 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6344 (p9_xxbrd_v2df): New expander.
6345 (p9_xxbrw_<mode>): Changed to ...
6346 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6347 (p9_xxbrw_v4sf): New expander.
6349 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6351 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6353 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6356 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6357 multiply in 32-bit mode.
6358 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6359 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6362 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6364 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6365 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6366 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6367 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6369 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6372 * dwarf2out.c (is_trivial_indirect_ref): New function.
6373 (dwarf2out_late_global_decl): Do not generate a location
6374 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6375 and that are not defined in the current unit.
6377 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6379 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6380 instead of a libcall for UNORDERED.
6382 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6385 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6386 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6388 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6391 * config/s390/s390.c (s390_set_current_function): Invoke
6392 s390_indirect_branch_settings also if fndecl didn't change.
6394 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6396 * config/rs6000/rs6000.md (blockage): Set length to zero.
6398 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6400 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6402 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6405 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6406 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6410 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6411 PARALLEL incoming that failed vt_get_decl_and_offset check.
6414 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6415 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6417 (get_variable_section): For decls in named .bss* sections pass true as
6418 second argument to bss_initializer_p.
6420 2018-02-09 Marek Polacek <polacek@redhat.com>
6421 Jakub Jelinek <jakub@redhat.com>
6424 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6425 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6426 Sync some changes from cxx_fold_indirect_ref.
6428 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6430 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6432 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6433 (BLOCK_INLINE_ENTRY_LABEL): New.
6434 (dwarf2out_var_location): Disregard inline entry markers.
6435 (inline_entry_data): New struct.
6436 (inline_entry_data_hasher): New hashtable type.
6437 (inline_entry_data_hasher::hash): New.
6438 (inline_entry_data_hasher::equal): New.
6439 (inline_entry_data_table): New variable.
6440 (add_high_low_attributes): Add DW_AT_entry_pc and
6441 DW_AT_GNU_entry_view attributes if a pending entry is found
6442 in inline_entry_data_table. Add old entry_pc attribute only
6443 if debug nonbinding markers are disabled.
6444 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6445 markers are enabled.
6446 (block_within_block_p, dwarf2out_inline_entry): New.
6447 (dwarf2out_finish): Check that no entries remained in
6448 inline_entry_data_table.
6449 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6450 (final_scan_insn, notice_source_line): Likewise.
6451 (rest_of_clean_state): Skip inline entry markers.
6452 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6454 * gimple.c (gimple_build_debug_inline_entry): New.
6455 * gimple.h (enum gimple_debug_subcode): Add
6456 GIMPLE_DEBUG_INLINE_ENTRY.
6457 (gimple_build_debug_inline_entry): Declare.
6458 (gimple_debug_inline_entry_p): New.
6459 (gimple_debug_nonbind_marker_p): Adjust.
6460 * insn-notes.def (INLINE_ENTRY): New.
6461 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6462 inline entry marker notes.
6463 (print_insn): Likewise.
6464 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6465 (INSN_DEBUG_MARKER_KIND): Likewise.
6466 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6467 * tree-inline.c (expand_call_inline): Build and insert
6468 debug_inline_entry stmt.
6469 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6470 inline entry blocks early, if nonbind markers are enabled.
6471 (dump_scope_block): Dump fragment info.
6472 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6473 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6474 (gimple_build_debug_inline_entry): New.
6475 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6476 Enable/disable inline entry points too.
6477 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6478 (DEBUG_INSN): Describe inline entry markers.
6480 * common.opt (gvariable-location-views): New.
6481 (gvariable-location-views=incompat5): New.
6482 * config.in: Rebuilt.
6483 * configure: Rebuilt.
6484 * configure.ac: Test assembler for view support.
6485 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6486 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6487 * dwarf2out.c (var_loc_view): New typedef.
6488 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6489 (dwarf2out_locviews_in_attribute): New.
6490 (dwarf2out_locviews_in_loclist): New.
6491 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6492 (enum dw_line_info_opcode): Add LI_adv_address.
6493 (struct dw_line_info_table): Add view.
6494 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6495 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6496 (zero_view_p): New variable.
6497 (ZERO_VIEW_P): New macro.
6498 (output_asm_line_debug_info): New.
6499 (struct var_loc_node): Add view.
6500 (add_AT_view_list, AT_loc_list): New.
6501 (add_var_loc_to_decl): Add view param. Test it against last.
6502 (new_loc_list): Add view params. Record them.
6503 (AT_loc_list_ptr): Handle loc and view lists.
6504 (view_list_to_loc_list_val_node): New.
6505 (print_dw_val): Handle dw_val_class_view_list.
6506 (size_of_die): Likewise.
6507 (value_format): Likewise.
6508 (loc_list_has_views): New.
6509 (gen_llsym): Set vl_symbol too.
6510 (maybe_gen_llsym, skip_loc_list_entry): New.
6511 (dwarf2out_maybe_output_loclist_view_pair): New.
6512 (output_loc_list): Output view list or entries too.
6513 (output_view_list_offset): New.
6514 (output_die): Handle dw_val_class_view_list.
6515 (output_dwarf_version): New.
6516 (output_compilation_unit_header): Use it.
6517 (output_skeleton_debug_sections): Likewise.
6518 (output_rnglists, output_line_info): Likewise.
6519 (output_pubnames, output_aranges): Update version comments.
6520 (output_one_line_info_table): Output view numbers in asm comments.
6521 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6522 Call maybe_gen_llsym.
6523 (loc_list_from_tree_1): Adjust.
6524 (add_AT_location_description): Create view list attribute if
6525 needed, check it's absent otherwise.
6526 (convert_cfa_to_fb_loc_list): Adjust.
6527 (maybe_emit_file): Call output_asm_line_debug_info for test.
6528 (dwarf2out_var_location): Reset views as needed. Precompute
6529 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6530 attribute. Set view.
6531 (new_line_info_table): Reset next view.
6532 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6533 (dwarf2out_source_line): Likewise. Output view resets and labels to
6534 the assembler, or select appropriate line info opcodes.
6535 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6536 (optimize_string_length): Catch it. Adjust.
6537 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6538 dw_val_class_view_list, and remove it if no longer needed.
6539 (hash_loc_list): Hash view numbers.
6540 (loc_list_hasher::equal): Compare them.
6541 (optimize_location_lists): Check whether a view list symbol is
6542 needed, and whether the locview attribute is present, and
6543 whether they match. Remove the locview attribute if no longer
6545 (index_location_lists): Call skip_loc_list_entry for test.
6546 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6547 Use output_dwarf_version.
6548 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6549 (struct dw_val_node): Add val_view_list.
6550 * final.c (SEEN_NEXT_VIEW): New.
6551 (set_next_view_needed): New.
6552 (clear_next_view_needed): New.
6553 (maybe_output_next_view): New.
6554 (final_start_function): Rename to...
6555 (final_start_function_1): ... this. Take pointer to FIRST,
6556 add SEEN parameter. Emit param bindings in the initial view.
6557 (final_start_function): Reintroduce SEEN-less interface.
6558 (final): Rename to...
6559 (final_1): ... this. Take SEEN parameter. Output final pending
6560 next view at the end.
6561 (final): Reintroduce seen-less interface.
6562 (final_scan_insn): Output pending next view before switching
6563 sections or ending a block. Mark the next view as needed when
6564 outputting variable locations. Notify debug backend of section
6565 changes, and of location view changes.
6566 (rest_of_handle_final): Adjust.
6567 * toplev.c (process_options): Autodetect value for debug variable
6568 location views option. Warn on incompat5 without -gdwarf-5.
6569 * doc/invoke.texi (gvariable-location-views): New.
6570 (gvariable-location-views=incompat5): New.
6571 (gno-variable-location-views): New.
6573 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6575 PR tree-optimization/84136
6576 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6577 that the result of find_edge is non-NULL.
6579 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6582 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6583 storing integer register in SImode. Fix cost of 256 and 512
6584 byte aligned SSE register store.
6586 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6588 * config/i386/i386.c (ix86_multiplication_cost): Fix
6589 multiplication cost for TARGET_AVX512DQ.
6591 2018-02-08 Marek Polacek <polacek@redhat.com>
6593 PR tree-optimization/84238
6594 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6597 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6599 PR tree-optimization/84265
6600 * tree-vect-stmts.c (vectorizable_store): Don't treat
6601 VMAT_CONTIGUOUS accesses as grouped.
6602 (vectorizable_load): Likewise.
6604 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6606 PR tree-optimization/81635
6607 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6608 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6609 (test_round_for_mask): New functions.
6610 (wide_int_cc_tests): Call test_round_for_mask.
6611 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6612 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6613 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6614 range returned by get_range_info.
6616 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6619 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6620 * symtab.c: Include builtins.h
6621 (symtab_node::output_to_lto_symbol_table_p): Move here
6622 from lto-streamer-out.c:output_symbol_p.
6623 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6624 (output_symbol_p): Move all logic to symtab.c
6625 (produce_symtab): Update.
6627 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6629 * config/s390/s390-opts.h (enum indirect_branch): Define.
6630 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6631 (s390_indirect_branch_via_thunk)
6632 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6633 (enum s390_indirect_branch_type): Define.
6634 * config/s390/s390.c (struct s390_frame_layout, struct
6635 machine_function): Remove.
6636 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6637 (indirect_branch_table_label_no, indirect_branch_table_name):
6639 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6640 (enum s390_indirect_branch_option): Define.
6641 (s390_return_addr_from_memory): New function.
6642 (s390_handle_string_attribute): New function.
6643 (s390_attribute_table): Add new attribute handler.
6644 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6645 (s390_indirect_branch_via_thunk): New function.
6646 (s390_indirect_branch_via_inline_thunk): New function.
6647 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6648 sibling call optimization for non z10 compiles.
6649 (s390_emit_call): Force indirect branch target to be a single
6650 register. Add r1 clobber for non-z10 compiles.
6651 (s390_emit_epilogue): Emit return jump via return_use expander.
6652 (s390_reorg): Handle JUMP_INSNs as execute targets.
6653 (s390_option_override_internal): Perform validity checks for the
6654 new command line options.
6655 (s390_indirect_branch_attrvalue): New function.
6656 (s390_indirect_branch_settings): New function.
6657 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6658 (s390_output_indirect_thunk_function): New function.
6659 (s390_code_end): Implement target hook.
6660 (s390_case_values_threshold): Implement target hook.
6661 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6663 * config/s390/s390.h (struct s390_frame_layout)
6664 (struct machine_function): Move here from s390.c.
6665 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6666 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6667 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6668 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6669 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6670 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6671 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6672 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6673 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6674 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6675 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6676 (mnemonic attribute): Add values which aren't recognized
6678 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6679 pattern for branch conversion. Fix mnemonic attribute.
6680 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6681 indirect branch via thunk if requested.
6682 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6683 ("*indirect_jump"): Disable for branch conversion using out of
6685 ("indirect_jump_via_thunk<mode>_z10")
6686 ("indirect_jump_via_thunk<mode>")
6687 ("indirect_jump_via_inlinethunk<mode>_z10")
6688 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6689 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6690 ("casesi_jump_via_inlinethunk<mode>_z10")
6691 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6692 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6693 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6694 ("*indirect2_jump"): Disable for branch conversion.
6695 ("casesi_jump"): Turn into expander and expand patterns for branch
6697 ("return_use"): New expander.
6698 ("*return"): Emit return via thunk and rename it to ...
6699 ("*return<mode>"): ... this one.
6700 * config/s390/s390.opt: Add new options and and enum for the
6703 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6705 * lra-constraints.c (match_reload): Unconditionally use
6706 gen_lowpart_SUBREG, rather than selecting between that
6707 and equivalent gen_rtx_SUBREG code.
6709 2018-02-08 Richard Biener <rguenther@suse.de>
6711 PR tree-optimization/84233
6712 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6713 changed flag instead of boguously re-using phi_inserted.
6715 2018-02-08 Martin Jambor <mjambor@suse.cz>
6717 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6718 static local variables.
6720 2018-02-08 Richard Biener <rguenther@suse.de>
6722 PR tree-optimization/84278
6723 * tree-vect-stmts.c (vectorizable_store): When looking for
6724 smaller vector types to perform grouped strided loads/stores
6725 make sure the mode is supported by the target.
6726 (vectorizable_load): Likewise.
6728 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6730 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6731 Increase LDP/STP opportunities by adding adjacent callee-saves.
6733 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6735 PR rtl-optimization/84068
6736 PR rtl-optimization/83459
6737 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6739 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6741 PR tree-optimization/84224
6742 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6743 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6746 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6749 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6750 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6751 count, remove one USE.
6753 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6755 * doc/install.texi (Configuration): Document the
6756 --with-long-double-format={ibm,ieee} PowerPC configuration
6760 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6761 Convert from define_expand to be define_insn_and_split. Rework
6762 float/double/_Float128 conversions to QI/HI/SImode to work with
6763 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6764 conversions to QI/HImode types did a store and then a load to
6765 truncate the value. For conversions to VSX registers, don't split
6766 the insn, instead emit the code directly. Use the code iterator
6767 any_fix to combine signed and unsigned conversions.
6768 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6769 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6770 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6771 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6772 (fix_<mode>di2_hw): Likewise.
6773 (fixuns_<mode>di2_hw): Likewise.
6774 (fix_<mode>si2_hw): Likewise.
6775 (fixuns_<mode>si2_hw): Likewise.
6776 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6777 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6778 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6779 fix<uns>_trunc<SFDF:mode>si2_p8.
6780 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6782 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6783 (fix<uns>_<mode>_mem): Likewise.
6784 (fctiw<u>z_<mode>_mem): Likewise.
6785 (fix<uns>_<mode>_mem): Likewise.
6786 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6787 the register allocator from doing a direct move to the GPRs to do
6788 a store, and instead use the ISA 3.0 store byte/half-word from
6789 vector register instruction. For IEEE 128-bit floating point,
6790 also optimize stores of 32-bit ints.
6791 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6793 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6795 * genextract.c (push_pathstr_operand): New function to support
6797 (walk_rtx): Call push_pathstr_operand.
6798 (print_path): Support [a-zA-Z].
6800 2018-02-07 Richard Biener <rguenther@suse.de>
6802 PR tree-optimization/84037
6803 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6804 (cse_and_gimplify_to_preheader): Declare.
6805 (vect_get_place_in_interleaving_chain): Likewise.
6806 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6808 (_loop_vec_info::~_loop_vec_info): Delete it.
6809 (cse_and_gimplify_to_preheader): New function.
6810 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6811 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6812 (vectorizable_load): Likewise. For grouped stores always base
6813 the IV on the first element.
6814 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6815 condition before gimplifying.
6817 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6819 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6820 *DIV_EXPR and *MOD_EXPR.
6822 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6825 * config/i386/i386.c (ix86_option_override_internal): Mask out
6826 the CF_SET bit when checking -fcf-protection.
6828 2018-02-07 Tom de Vries <tom@codesourcery.com>
6831 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6834 2018-02-07 Richard Biener <rguenther@suse.de>
6836 PR tree-optimization/84204
6837 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6840 PR tree-optimization/84205
6841 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6842 special-case isl_ast_op_zdiv_r.
6844 PR tree-optimization/84223
6845 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6846 Only add conditions from within the region.
6847 (gather_bbs::after_dom_children): Adjust.
6849 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6852 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6853 * config/avr/avr.md: Only post-reload split REG-REG moves if
6854 either register is GENERAL_REG_P.
6856 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6858 PR tree-optimization/84235
6859 * tree-ssa-scopedtables.c
6860 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6861 if the subtraction is performed in floating point type where NaNs are
6862 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6863 build 1. Formatting fix.
6865 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6868 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6869 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6870 and skip it regardless of bb boundaries. Use CALL_P macro,
6871 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6872 unnecessarily, formatting fix.
6874 2018-02-06 Michael Collison <michael.collison@arm.com>
6876 * config/arm/thumb2.md:
6877 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6878 (*thumb_mov_notscc): Ditto.
6880 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6883 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6884 unsigned_fix, not "s".
6886 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6888 * configure.ac (gcc_fn_eh_frame_ro): New function.
6889 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6890 correct .eh_frame permissions.
6891 * configure: Regenerate.
6893 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6895 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6898 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6900 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6901 Display warning message for -mno-speculate-indirect-jumps.
6903 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6905 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6907 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6909 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6911 PR tree-optimization/84225
6912 * tree-eh.c (find_trapping_overflow): Only call
6913 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6915 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6918 * config/i386/i386.c: Reimplement the check of possible options
6919 -mibt/-mshstk conbination. Change error messages.
6920 * doc/invoke.texi: Fix a typo: remove extra '='.
6922 2018-02-06 Marek Polacek <polacek@redhat.com>
6924 PR tree-optimization/84228
6925 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6927 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6930 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6931 emitted arch directives.
6932 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6933 __ARM_FEATURE_COPROC before changing architectures.
6935 2018-02-06 Richard Biener <rguenther@suse.de>
6937 * config/i386/i386.c (print_reg): Fix typo.
6938 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6940 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6942 * configure: Regenerate.
6944 2018-02-05 Martin Sebor <msebor@redhat.com>
6946 PR tree-optimization/83369
6947 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6950 2018-02-05 Martin Liska <mliska@suse.cz>
6952 * doc/invoke.texi: Cherry-pick upstream r323995.
6954 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6956 * ira.c (ira_init_register_move_cost): Adjust comment.
6958 2018-02-05 Martin Liska <mliska@suse.cz>
6960 PR gcov-profile/84137
6961 * doc/gcov.texi: Fix typo in documentation.
6963 2018-02-05 Martin Liska <mliska@suse.cz>
6965 PR gcov-profile/83879
6966 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6967 using dlopen functionality.
6969 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6971 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6972 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6973 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6974 _mm_maskz_range_round_ss): New intrinsics.
6975 (__builtin_ia32_rangesd128_round)
6976 (__builtin_ia32_rangess128_round): Remove.
6977 (__builtin_ia32_rangesd128_mask_round,
6978 __builtin_ia32_rangess128_mask_round): New builtins.
6979 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6980 __builtin_ia32_rangess128_round): Remove.
6981 (__builtin_ia32_rangesd128_mask_round,
6982 __builtin_ia32_rangess128_mask_round): New builtins.
6983 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6984 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6985 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6986 "<round_saeonly_constraint>")): Changed to ...
6987 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6988 "<round_saeonly_scalar_constraint>")): ... this.
6989 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6990 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6991 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6992 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6993 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6995 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6997 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6999 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7000 Remove all values except native, 8540 and 8548.
7002 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7004 * config/i386/i386.c (ix86_output_function_return): Pass
7005 INVALID_REGNUM, instead of -1, as invalid register number to
7006 indirect_thunk_name and output_indirect_thunk.
7008 2018-02-02 Julia Koval <julia.koval@intel.com>
7010 * config.gcc: Add -march=icelake.
7011 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7012 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7013 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7014 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7015 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7016 (processor_target_table): Add icelake.
7017 (ix86_option_override_internal): Handle new PTAs.
7018 (get_builtin_code_for_version): Handle icelake.
7019 (M_INTEL_COREI7_ICELAKE): New.
7020 (fold_builtin_cpu): Handle icelake.
7021 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7022 * doc/invoke.texi: Add -march=icelake.
7024 2018-02-02 Julia Koval <julia.koval@intel.com>
7026 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7027 to wide_int_bitmask.
7028 * wide-int-bitmask.h: New.
7030 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7033 * config/i386/i386.md: Replace Pmode with word_mode in
7034 builtin_setjmp_setup and builtin_longjmp to support x32.
7036 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7040 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7042 (rs6000_supported_cpu_names): New static variable.
7043 (linux_cpu_translation_table): Likewise.
7044 (elf_platform) <cpu>: Define new static variable and use it.
7045 Translate kernel AT_PLATFORM name to canonical name if needed.
7046 Error if platform name is unknown.
7048 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7051 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7053 2018-02-01 Jeff Law <law@redhat.com>
7056 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7057 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7058 the scratch if RELEASE_VIA_POP is false.
7059 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7060 If we have to save a temporary register, decrement SIZE appropriately.
7061 Pass new arguments to release_scratch_register_on_entry.
7062 (ix86_adjust_stack_and_probe): Likewise.
7063 (ix86_emit_probe_stack_range): Pass new arguments to
7064 release_scratch_register_on_entry.
7066 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7068 PR rtl-optimization/84157
7069 * combine.c (change_zero_ext): Use REG_P predicate in
7070 front of HARD_REGISTER_P predicate.
7072 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7074 * config/avr/avr.c (avr_option_override): Move disabling of
7075 -fdelete-null-pointer-checks to...
7076 * common/config/avr/avr-common.c (avr_option_optimization_table):
7079 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7081 PR tree-optimization/81635
7082 * tree-data-ref.c (split_constant_offset_1): For types that
7083 wrap on overflow, try to use range info to prove that wrapping
7086 2018-02-01 Renlin Li <renlin.li@arm.com>
7089 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7091 (aarch64_register_move_cost): Likewise.
7092 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7094 (REG_CLASS_NAMES): Likewise.
7095 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7096 TAILCALL_ADDR_REGS. Remove IP registers.
7097 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7099 2018-02-01 Richard Biener <rguenther@suse.de>
7101 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7102 for specifying RPO order and allow NULL for that.
7103 * domwalk.c (dom_walker::dom_walker): Likewise.
7104 (dom_walker::walk): Handle NULL RPO order.
7105 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7107 (rewrite_update_dom_walker): Likewise.
7108 (mark_def_dom_walker): Likewise.
7110 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7112 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7113 (aarch64_maybe_expand_sve_subreg_move): Declare.
7114 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7115 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7117 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7118 that are semantically a reverse operation.
7119 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7120 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7121 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7123 (aarch64_can_change_mode_class): For big-endian, forbid changes
7124 between two SVE modes if they have different element sizes.
7126 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7128 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7129 the TImode handling for big-endian targets.
7131 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7133 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7134 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7136 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7137 Remove BSWAP handing for big-endian targets and use the form of
7138 LD1RQ appropariate for the mode.
7140 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7142 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7143 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7146 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7149 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7150 check for operands that need to go through aarch64_sve_reload_be.
7152 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7154 PR tree-optimization/81661
7155 PR tree-optimization/84117
7156 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7157 * tree-eh.c: Include gimplify.h.
7158 (find_trapping_overflow, replace_trapping_overflow,
7159 rewrite_to_non_trapping_overflow): New functions.
7160 * tree-vect-loop.c: Include tree-eh.h.
7161 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7162 * tree-data-ref.c: Include tree-eh.h.
7163 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7165 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7167 PR rtl-optimization/84123
7168 * combine.c (change_zero_ext): Check if hard register satisfies
7169 can_change_dest_mode before calling gen_lowpart_SUBREG.
7171 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7174 * ira.c (ira_init_register_move_cost): Remove assert.
7176 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7178 PR rtl-optimization/84071
7179 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7180 * doc/tm.texi: Regenerate.
7182 2018-01-31 Richard Biener <rguenther@suse.de>
7184 PR tree-optimization/84132
7185 * tree-data-ref.c (analyze_miv_subscript): Properly
7186 check whether evolution_function_is_affine_multivariate_p
7187 before calling gcd_of_steps_may_divide_p.
7189 2018-01-31 Julia Koval <julia.koval@intel.com>
7192 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7193 * config/i386/i386.md (rdpid_rex64) New.
7194 (rdpid): Make 32bit only.
7196 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7199 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7200 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7202 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7205 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7207 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7209 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7211 PR rtl-optimization/84071
7212 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7213 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7215 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7217 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7218 (arc_attribute_table): Add 'aux' attribute.
7219 (arc_in_small_data_p): Consider aux like variables.
7220 (arc_is_aux_reg_p): New function.
7221 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7222 (arc_get_aux_arg): New function.
7223 (prepare_move_operands): Handle aux-register access.
7224 (arc_handle_aux_attribute): New function.
7225 * doc/extend.texi (ARC Variable attributes): Add subsection.
7227 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7229 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7230 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7231 (arc_attribute_table): Add 'uncached' attribute.
7232 (arc_print_operand): Print '.di' flag for uncached memory
7234 (arc_in_small_data_p): Do not consider for small data the uncached
7236 (arc_is_uncached_mem_p): New function.
7237 * config/arc/predicates.md (compact_store_memory_operand): Check
7238 for uncached memory accesses.
7239 (nonvol_nonimm_operand): Likewise.
7240 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7242 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7245 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7246 falign-loops=): Add Optimization flag.
7248 2018-01-30 Jeff Law <law@redhat.com>
7251 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7252 INT_REGISTERS_SAVED. Check it prior to calling
7253 get_scratch_register_on_entry.
7254 (ix86_adjust_stack_and_probe): Similarly.
7255 (ix86_emit_probe_stack_range): Similarly.
7256 (ix86_expand_prologue): Corresponding changes.
7258 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7261 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7262 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7264 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7267 * lra-constraints.c (curr_insn_transform): Process AND in the
7270 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7272 PR rtl-optimization/83986
7273 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7274 dependence against last_pending_memory_flush in addition to
7277 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7279 PR tree-optimization/81611
7280 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7283 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7286 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7289 2018-01-30 Richard Biener <rguenther@suse.de>
7290 Jakub Jelinek <jakub@redhat.com>
7292 PR tree-optimization/84111
7293 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7294 inner loops added during recursion, as they don't have up-to-date
7297 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7300 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7301 (can_inline_edge_by_limits_p): ... here.
7302 (can_early_inline_edge_p, check_callers,
7303 update_caller_keys, update_callee_keys, recursive_inlining,
7304 add_new_edges_to_heap, speculation_useful_p,
7305 inline_small_functions,
7306 inline_small_functions, flatten_function,
7307 inline_to_all_callers_1): Update.
7309 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7311 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7314 2018-01-30 Richard Biener <rguenther@suse.de>
7316 PR tree-optimization/83008
7317 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7318 invariant and constant vector uses in stmts when they need
7321 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7324 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7325 * configure: Regenerate.
7327 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7329 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7331 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7332 Use gen_rtx_REG rather than gen_lowpart.
7334 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7336 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7337 rather than 0 when creating partial subregs.
7339 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7341 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7344 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7347 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7348 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7349 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7350 flags. This restores the settings used before the 2017-07-24.
7351 Turning off pre increment/decrement/modify allows IVOPTS to
7352 optimize DF/SF loops where the index is an int.
7354 2018-01-29 Richard Biener <rguenther@suse.de>
7355 Kelvin Nilsen <kelvin@gcc.gnu.org>
7358 * tree-vect-stmts.c (vectorizable_call): Don't call
7359 targetm.vectorize_builtin_md_vectorized_function if callee is
7362 2018-01-22 Carl Love <cel@us.ibm.com>
7364 * doc/extend.tex: Fix typo in second arg in
7365 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7367 2018-01-29 Richard Biener <rguenther@suse.de>
7369 PR tree-optimization/84086
7370 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7371 (flush_ssaname_freelist): When SSA names were released reset
7372 the SCEV hash table.
7374 2018-01-29 Richard Biener <rguenther@suse.de>
7376 PR tree-optimization/84057
7377 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7378 removed paths when removing edges.
7380 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7382 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7383 -mfunction-return=@var{choice}.
7385 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7388 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7390 (layout::print_source_line): Likewise.
7391 (test_get_line_width_without_trailing_whitespace): Add test cases.
7393 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7396 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7399 2018-01-26 Jim Wilson <jimw@sifive.com>
7401 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7403 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7406 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7408 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7409 and CMP + SUB-immediate -> SUBS.
7411 2018-01-26 Martin Sebor <msebor@redhat.com>
7413 PR tree-optimization/83896
7414 * tree-ssa-strlen.c (get_string_len): Rename...
7415 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7416 Avoid assuming length is constant.
7417 (handle_char_store): Use HOST_WIDE_INT for string length.
7419 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7422 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7423 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7425 2018-01-26 Richard Biener <rguenther@suse.de>
7427 PR rtl-optimization/84003
7428 * dse.c (record_store): Only record redundant stores when
7429 the earlier store aliases at least all accesses the later one does.
7431 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7433 PR rtl-optimization/83985
7434 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7435 REG_CFA_RESTORE insns.
7436 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7437 REG_CFA_RESTORE insns here.
7440 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7441 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7443 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7445 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7446 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7447 (arc_init): Likewise.
7448 (arc_override_options): Likewise.
7449 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7451 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7453 * config/arc/arc.h (TARGET_DBNZ): Define.
7454 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7455 properly set the tune attribute.
7456 (dbnz): Use TARGET_DBNZ guard.
7457 * config/arc/arc.opt (mtune): Add core3 option.
7459 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7461 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7462 recognize new pic like addresses.
7463 (arc_delegitimize_address): Clean up.
7465 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7467 * config/arc/arc-arches.def: Option mrf16 valid for all
7469 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7470 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7471 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7472 * config/arc/arc-tables.opt: Regenerate.
7473 * config/arc/arc.c (arc_conditional_register_usage): Handle
7474 reduced register file case.
7475 (arc_file_start): Set must have build attributes.
7476 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7478 * config/arc/arc.opt (mrf16): Add new option.
7479 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7480 * config/arc/genmultilib.awk: Handle new mrf16 option.
7481 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7482 * config/arc/t-multilib: Regenerate.
7483 * doc/invoke.texi (ARC Options): Document mrf16 option.
7485 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7487 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7488 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7489 (arc_attribute_table): Add 'secure_call' attribute.
7490 (arc_print_operand): Print secure call operand.
7491 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7493 (arc_is_secure_call_p): New function. * config/arc/arc.md
7494 (call_i): Add support for sjli instruction.
7495 (call_value_i): Likewise.
7496 * config/arc/constraints.md (Csc): New constraint.
7498 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7499 John Eric Martin <John.Martin@emmicro-us.com>
7501 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7502 * config/arc/arc.c (_arc_jli_section): New struct.
7503 (arc_jli_section): New type.
7504 (rc_jli_sections): New static variable.
7505 (arc_handle_jli_attribute): New function.
7506 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7507 (arc_file_end): New function.
7508 (TARGET_ASM_FILE_END): Define.
7509 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7510 (arc_add_jli_section): New function.
7511 (jli_call_scan): Likewise.
7512 (arc_reorg): Call jli_call_scan.
7513 (arc_output_addsi): Remove 'S' from printing asm operand.
7514 (arc_is_jli_call_p): New function.
7515 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7517 (movhi_insn): Likewise.
7518 (movsi_insn): Likewise.
7519 (movsi_set_cc_insn): Likewise.
7520 (loadqi_update): Likewise.
7521 (load_zeroextendqisi_update): Likewise.
7522 (load_signextendqisi_update): Likewise.
7523 (loadhi_update): Likewise.
7524 (load_zeroextendhisi_update): Likewise.
7525 (load_signextendhisi_update): Likewise.
7526 (loadsi_update): Likewise.
7527 (loadsf_update): Likewise.
7528 (movsicc_insn): Likewise.
7529 (bset_insn): Likewise.
7530 (bxor_insn): Likewise.
7531 (bclr_insn): Likewise.
7532 (bmsk_insn): Likewise.
7533 (bicsi3_insn): Likewise.
7534 (cmpsi_cc_c_insn): Likewise.
7535 (movsi_ne): Likewise.
7536 (movsi_cond_exec): Likewise.
7537 (clrsbsi2): Likewise.
7548 (call_i): Remove 'S' asm letter, add jli instruction.
7549 (call_value_i): Likewise.
7550 * config/arc/arc.op (mjli-always): New option.
7551 * config/arc/constraints.md (Cji): New constraint.
7552 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7554 (subsf3_fpx): Likewise.
7555 (mulsf3_fpx): Likewise.
7556 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7558 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7560 * doc/invoke.texi (ARC): Document mjli-always option.
7562 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7564 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7565 avoid addition with 0 and use incw and decw where possible.
7567 2018-01-26 Richard Biener <rguenther@suse.de>
7569 PR tree-optimization/81082
7570 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7571 association if it requires casting to unsigned.
7572 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7573 from fold_plusminus_mult_expr to catch important cases late when
7574 range info is available.
7576 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7578 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7579 * configure.ac (hidden_linkonce): New test.
7580 * configure: Regenerate.
7581 * config.in: Regenerate.
7583 2018-01-26 Julia Koval <julia.koval@intel.com>
7585 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7586 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7587 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7588 _mm_mask_bitshuffle_epi64_mask): Fix type.
7589 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7590 USI_FTYPE_V4DI_V4DI_USI): Remove.
7591 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7592 __builtin_ia32_vpshufbitqmb256_mask,
7593 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7594 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7595 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7597 2018-01-26 Alan Modra <amodra@gmail.com>
7600 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7601 UNSPEC_VBPERMQ. Sort other unspecs.
7603 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7605 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7607 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7610 * predict.c (drop_profile): Do not push/pop cfun; update also
7612 (handle_missing_profiles): Fix logic looking for zero profiles.
7614 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7617 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7618 on functions with #pragma omp declare simd or functions with simd
7620 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7621 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7622 Remove trailing \n from warning_at calls.
7624 2018-01-25 Tom de Vries <tom@codesourcery.com>
7627 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7628 for neutered workers.
7630 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7633 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7634 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7636 2018-01-24 Jeff Law <law@redhat.com>
7639 * i386.c (get_probe_interval): Move to earlier point.
7640 (ix86_compute_frame_layout): If -fstack-clash-protection and
7641 the frame is larger than the probe interval, then use pushes
7642 to save registers rather than reg->mem moves.
7643 (ix86_expand_prologue): Remove conditional for int_registers_saved
7646 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7649 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7650 min/max for never referenced object.
7652 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7655 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7657 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7658 attributes from DECL_ATTRIBUTES (decl) without affecting
7659 DECL_ATTRIBUTES (current_function_decl).
7660 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7661 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7663 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7665 PR tree-optimization/83979
7666 * fold-const.c (fold_comparison): Use constant_boolean_node
7667 instead of boolean_{true,false}_node.
7669 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7671 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7674 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7676 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7677 Simplify the clause that sets the length attribute.
7678 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7679 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7680 clause that sets the length attribute.
7681 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7683 2018-01-24 Tom de Vries <tom@codesourcery.com>
7686 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7687 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7688 Add strict parameter.
7689 (prevent_branch_around_nothing): Insert dummy insn between branch to
7690 label and label with no ptx insn inbetween.
7691 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7693 2018-01-24 Tom de Vries <tom@codesourcery.com>
7696 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7697 for neutered threads in warp.
7698 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7700 2018-01-24 Richard Biener <rguenther@suse.de>
7702 PR tree-optimization/83176
7703 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7706 2018-01-24 Richard Biener <rguenther@suse.de>
7708 PR tree-optimization/82819
7709 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7710 code generating pluses that are no-ops in the target precision.
7712 2018-01-24 Richard Biener <rguenther@suse.de>
7715 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7717 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7719 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7720 to merge probabilities.
7721 * predict.c (probably_never_executed): Also mark as cold functions
7722 with global 0 profile and guessed local profile.
7723 * profile-count.c (profile_probability::combine_with_count): New
7725 * profile-count.h (profile_probability::operator*,
7726 profile_probability::operator*=, profile_probability::operator/,
7727 profile_probability::operator/=): Reduce precision to adjusted
7728 and set value to guessed on contradictory divisions.
7729 (profile_probability::combine_with_freq): Remove.
7730 (profile_probability::combine_wiht_count): Declare.
7731 (profile_count::force_nonzero):: Set to adjusted.
7732 (profile_count::probability_in):: Set quality to adjusted.
7733 * tree-ssa-tail-merge.c (replace_block_by): Use
7736 2018-01-23 Andrew Waterman <andrew@sifive.com>
7737 Jim Wilson <jimw@sifive.com>
7739 * config/riscv/riscv.c (riscv_stack_boundary): New.
7740 (riscv_option_override): Set riscv_stack_boundary. Handle
7741 riscv_preferred_stack_boundary_arg.
7742 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7743 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7744 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7745 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7746 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7747 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7749 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7752 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7753 of struct ix86_frame.
7754 (ix86_expand_epilogue): Likewise. Add a local variable for
7755 the reg_save_offset field in struct ix86_frame.
7757 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7759 PR tree-optimization/82604
7760 * tree-loop-distribution.c (enum partition_kind): New enum item
7761 PKIND_PARTIAL_MEMSET.
7762 (partition_builtin_p): Support above new enum item.
7763 (generate_code_for_partition): Ditto.
7764 (compute_access_range): Differentiate cases that equality can be
7765 proven at all loops, the innermost loops or no loops.
7766 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7767 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7768 (finalize_partitions, distribute_loop): Don't fuse partition of
7769 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7770 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7773 2018-01-23 Martin Liska <mliska@suse.cz>
7775 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7776 order to ignore the predictor.
7777 (PRED_POLYMORPHIC_CALL): Likewise.
7778 (PRED_RECURSIVE_CALL): Likewise.
7780 2018-01-23 Martin Liska <mliska@suse.cz>
7782 * tree-profile.c (tree_profiling): Print function header to
7783 aware reader which function we are working on.
7784 * value-prof.c (gimple_find_values_to_profile): Do not print
7785 not interesting value histograms.
7787 2018-01-23 Martin Liska <mliska@suse.cz>
7789 * profile-count.h (enum profile_quality): Add
7790 profile_uninitialized as the first value. Do not number values
7791 as they are zero based.
7792 (profile_count::verify): Update sanity check.
7793 (profile_probability::verify): Likewise.
7795 2018-01-23 Nathan Sidwell <nathan@acm.org>
7797 * doc/invoke.texi (ffor-scope): Deprecate.
7799 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7801 PR tree-optimization/83510
7802 * domwalk.c (set_all_edges_as_executable): New function.
7803 (dom_walker::dom_walker): Convert bool param
7804 "skip_unreachable_blocks" to enum reachability. Move setup of
7805 edge flags to set_all_edges_as_executable and only do it when
7806 reachability is REACHABLE_BLOCKS.
7807 * domwalk.h (enum dom_walker::reachability): New enum.
7808 (dom_walker::dom_walker): Convert bool param
7809 "skip_unreachable_blocks" to enum reachability.
7810 (set_all_edges_as_executable): New decl.
7811 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7812 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7814 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7815 but converting true to REACHABLE_BLOCKS.
7816 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7818 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7819 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7820 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7822 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7823 if check_all_array_refs will be called.
7825 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7827 * tree.c (selftest::test_location_wrappers): Add more test
7830 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7832 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7833 (selftest::test_bit_in_range): Likewise.
7835 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7838 * doc/sourcebuild.texi (vect_float): Say that the selector
7839 only describes the situation when -funsafe-math-optimizations is on.
7840 (vect_float_strict): Document.
7842 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7844 PR tree-optimization/83965
7845 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7846 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7847 instead of checking only for a reduction.
7848 (vect_recog_widen_sum_pattern): Likewise.
7850 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7852 * predict.c (probably_never_executed): Only use precise profile info.
7853 (compute_function_frequency): Skip after inlining hack since we now
7854 have quality checking.
7856 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7858 * profile-count.h (profile_probability::very_unlikely,
7859 profile_probability::unlikely, profile_probability::even): Set
7860 precision to guessed.
7862 2018-01-23 Richard Biener <rguenther@suse.de>
7864 PR tree-optimization/83963
7865 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7866 Properly terminate dominator walk when crossing the exit edge not
7867 when visiting its source block.
7869 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7872 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7873 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7875 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7877 PR tree-optimization/83957
7878 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7879 semicolon after for body surrounded by braces.
7881 PR tree-optimization/83081
7882 * profile-count.h (profile_probability::split): New method.
7883 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7884 Use profile_probability::split.
7885 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7886 when splitting a single conditional jump into 2.
7888 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7890 PR tree-optimization/69452
7891 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7894 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7896 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7897 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7898 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7900 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7902 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7903 * config/rl78/rl78.md (movdi): New define_expand.
7904 * config/rl78/rl78.c (rl78_split_movdi): New function.
7906 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7909 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7911 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7912 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7913 128-bit to produce an UNSPEC move to get the double word with the
7914 signbit and then a shift directly to do signbit.
7915 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7916 implementation with a new version that just does either a direct
7917 move or a regular move. Move memory interface to separate insns.
7918 Move insns so they are next to the expander.
7919 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7920 with signbit move. Split big and little endian case.
7921 (signbit<mode>2_dm_mem_le): Likewise.
7922 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7923 (signbit<mode>2_dm2): Likewise.
7925 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7927 * config/rl78/rl78.md (anddi3): New define_expand.
7929 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7931 * config/rl78/rl78.md (umindi3): New define_expand.
7933 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7935 * config/rl78/rl78.md (smindi3): New define_expand.
7937 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7939 * config/rl78/rl78.md (smaxdi3): New define_expand.
7941 2018-01-22 Carl Love <cel@us.ibm.com>
7943 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7944 LVX_V1TI): Add macro expansion.
7945 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7946 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7947 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7948 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7949 Change check to determine if the instruction is a byte reversing
7950 entry. Fix typo in comment.
7951 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7952 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7953 Add def_builtin calls for new builtins.
7954 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7955 Add define_insn expansion.
7957 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7959 * config/rl78/rl78.md (umaxdi3): New define_expand.
7961 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7963 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7964 for non-QImode registers.
7966 2018-01-22 Richard Biener <rguenther@suse.de>
7968 PR tree-optimization/83963
7969 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7970 including the loop exit block.
7971 (scop_detection::merge_sese): Likewise.
7972 (scop_detection::add_scop): Do it here instead.
7974 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7976 * doc/sourcebuild.texi (arm_softfloat): Document.
7978 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7981 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7982 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7983 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7985 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7986 David Edelsohn <dje.gcc@gmail.com>
7989 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7990 Change "crset eq" to "crset 2".
7991 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7992 (*call_indirect_aix<mode>_nospec): Likewise.
7993 (*call_value_indirect_aix<mode>_nospec): Likewise.
7994 (*call_indirect_elfv2<mode>_nospec): Likewise.
7995 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7996 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7997 change assembly output from . to $.
7998 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7999 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8000 (*tablejump<mode>_internal1_nospec): Likewise.
8002 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8005 * config/sh/sh_optimize_sett_clrt.cc:
8006 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8008 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8010 PR tree-optimization/83940
8011 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8012 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8013 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8014 use it to pass back the definition type.
8015 (vect_check_store_rhs): Likewise rhs_dt_out.
8016 (vect_build_gather_load_calls): Add a mask_dt argument and use
8017 it instead of a call to vect_is_simple_use.
8018 (vectorizable_store): Update calls to vect_check_load_store_mask
8019 and vect_check_store_rhs. Use the dt returned by the latter instead
8020 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8021 instead of calls to vect_is_simple_use. Pass the scalar rather
8022 than the vector operand to vect_is_simple_use when handling
8023 second and subsequent copies of an rhs value.
8024 (vectorizable_load): Update calls to vect_check_load_store_mask
8025 and vect_build_gather_load_calls. Use the cached mask_dt and
8026 gs_info.offset_dt instead of calls to vect_is_simple_use.
8028 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8031 * tree-emutls.c: Include gimplify.h.
8032 (lower_emutls_2): New function.
8033 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8034 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8035 it before further processing.
8038 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8039 UINTVAL (trueop1) instead of INTVAL (op1).
8041 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8045 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8046 INCOMING_FRAME_SP_OFFSET if not defined.
8047 (scan_trace): Add ENTRY argument. If true and
8048 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8049 emit a note to adjust the CFA offset.
8050 (create_cfi_notes): Adjust scan_trace callers.
8051 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8052 INCOMING_FRAME_SP_OFFSET in the CIE.
8053 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8054 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8056 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8057 * doc/tm.texi: Regenerated.
8059 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8061 PR rtl-optimization/83147
8062 * lra-constraints.c (remove_inheritance_pseudos): Use
8063 lra_substitute_pseudo_within_insn.
8065 2018-01-19 Tom de Vries <tom@codesourcery.com>
8066 Cesar Philippidis <cesar@codesourcery.com>
8069 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8071 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8074 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8075 spaces for function labels.
8077 2018-01-19 Martin Liska <mliska@suse.cz>
8079 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8080 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8081 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8082 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8083 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8084 (PRED_CONST_RETURN): Change from 69 to 65.
8085 (PRED_NULL_RETURN): Change from 91 to 71.
8086 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8087 (PRED_LOOP_GUARD): Change from 66 to 73.
8089 2018-01-19 Martin Liska <mliska@suse.cz>
8091 * predict.c (predict_insn_def): Add new assert.
8092 (struct branch_predictor): Change type to signed integer.
8093 (test_prediction_value_range): Amend test to cover
8095 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8096 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8097 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8098 (PRED_LOOP_IV_COMPARE): Likewise.
8099 * predict.h (PROB_UNINITIALIZED): Define new constant.
8101 2018-01-19 Martin Liska <mliska@suse.cz>
8103 * predict.c (dump_prediction): Add new format for
8104 analyze_brprob.py script which is enabled with -details
8106 * profile-count.h (precise_p): New function.
8108 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8110 PR tree-optimization/83922
8111 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8112 there are no statements that need masking.
8113 (vect_active_double_reduction_p): New function.
8114 (vect_analyze_loop_operations): Use it when handling phis that
8115 are not in the loop header.
8117 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8119 PR tree-optimization/83914
8120 * tree-vect-loop.c (vectorizable_induction): Don't convert
8121 init_expr or apply the peeling adjustment for inductions
8122 that are nested within the vectorized loop.
8124 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8126 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8129 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8133 * function.h (gimplify_parameters): Add gimple_seq * argument.
8134 * function.c: Include gimple.h and options.h.
8135 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8136 for the added local temporaries if needed.
8137 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8138 if there are any parameter cleanups, wrap whole body into a
8139 try/finally with the cleanups.
8141 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8144 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8145 Use GET_MODE_CLASS for scalar floating point.
8147 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8151 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8152 Fix call of call_cgraph_insertion_hooks.
8154 2018-01-18 Martin Sebor <msebor@redhat.com>
8156 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8158 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8161 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8164 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8167 * common.opt: (-ffile-prefix-map): New option.
8168 * opts.c (common_handle_option): Defer it.
8169 * opts-global.c (handle_common_deferred_options): Handle it.
8170 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8171 * file-prefix-map.h: New file.
8172 (remap_debug_filename, add_debug_prefix_map): ...here.
8173 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8174 * final.c (debug_prefix_map, add_debug_prefix_map
8175 remap_debug_filename): Move to...
8176 * file-prefix-map.c: New file.
8177 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8178 generalize, get rid of alloca(), use strrchr() instead of strchr().
8179 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8180 Implement in terms of add_prefix_map().
8181 (remap_macro_filename, remap_debug_filename): Implement in term of
8183 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8184 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8185 * dbxout.c: Include file-prefix-map.h.
8186 * varasm.c: Likewise.
8187 * vmsdbgout.c: Likewise.
8188 * xcoffout.c: Likewise.
8189 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8190 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8191 * doc/invoke.texi (-ffile-prefix-map): Document.
8192 (-fdebug-prefix-map): Update description.
8194 2018-01-18 Martin Liska <mliska@suse.cz>
8196 * config/i386/i386.c (indirect_thunk_name): Document that also
8198 (output_indirect_thunk): Document why both instructions
8199 (pause and lfence) are generated.
8201 2018-01-18 Richard Biener <rguenther@suse.de>
8203 PR tree-optimization/83887
8204 * graphite-scop-detection.c
8205 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8206 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8207 (scop_detection::merge_sese): Re-implement with a flood-fill
8208 algorithm that properly finds a SESE region if it exists.
8210 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8213 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8214 pointer_diff optimizations use view_convert instead of convert.
8216 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8218 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8219 Generate different code for -mno-speculate-indirect-jumps.
8220 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8221 (*call_indirect_aix<mode>): Disable for
8222 -mno-speculate-indirect-jumps.
8223 (*call_indirect_aix<mode>_nospec): New define_insn.
8224 (*call_value_indirect_aix<mode>): Disable for
8225 -mno-speculate-indirect-jumps.
8226 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8227 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8228 -mno-speculate-indirect-jumps.
8229 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8231 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8233 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8234 long double type, set the flags for noting the default long double
8235 type, even if we don't pass or return a long double type.
8237 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8240 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8243 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8245 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8246 support for merge[hl].
8247 (fold_mergehl_helper): New helper function.
8248 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8249 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8250 (altivec_vmrglw_direct): Add xxmrglw insn.
8252 2018-01-17 Andrew Waterman <andrew@sifive.com>
8254 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8255 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8257 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8260 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8261 call the lto_location_cache before reading the
8262 DECL_SOURCE_LOCATION of the types.
8264 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8265 Richard Sandiford <richard.sandiford@linaro.org>
8267 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8268 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8269 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8270 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8271 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8273 * config/aarch64/constraints.md (aarch64_movti_operand):
8275 * config/aarch64/predicates.md (Uti): Add new constraint.
8277 2018-01-17 Carl Love <cel@us.ibm.com>
8279 * config/rs6000/vsx.md (define_expand xl_len_r,
8280 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8281 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8283 (define_expand, define_insn): Move the shift left from the
8284 define_insn to the define_expand for lxvl and stxvl instructions.
8285 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8286 and XL_LEN_R definitions to PURE.
8288 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8290 * config/i386/i386.c (indirect_thunk_name): Declare regno
8291 as unsigned int. Compare regno with INVALID_REGNUM.
8292 (output_indirect_thunk): Ditto.
8293 (output_indirect_thunk_function): Ditto.
8294 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8295 in the call to output_indirect_thunk_function.
8297 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8300 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8301 rather than the size of inner_type to determine the stack slot size
8302 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8304 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8307 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8310 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8312 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8313 endian Linux systems to optionally enable multilibs for selecting
8314 the long double type if the user configured an explicit type.
8315 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8316 have no long double multilibs if not defined.
8317 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8318 warn if the user used -mabi={ieee,ibm}longdouble and we built
8319 multilibs for long double.
8320 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8321 appropriate multilib option.
8322 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8324 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8325 for building long double multilibs.
8326 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8328 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8330 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8333 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8335 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8338 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8341 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8344 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8346 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8347 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8349 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8351 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8352 different rtl trees depending on TARGET_64BIT.
8353 (rs6000_gen_lvx): Likewise.
8355 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8357 * config/visium/visium.md (nop): Tweak comment.
8358 (hazard_nop): Likewise.
8360 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8362 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8363 -mspeculate-indirect-jumps.
8364 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8365 for -mno-speculate-indirect-jumps.
8366 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8367 (*call_value_indirect_elfv2<mode>): Disable for
8368 -mno-speculate-indirect-jumps.
8369 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8370 (indirect_jump): Emit different RTL for
8371 -mno-speculate-indirect-jumps.
8372 (*indirect_jump<mode>): Disable for
8373 -mno-speculate-indirect-jumps.
8374 (*indirect_jump<mode>_nospec): New define_insn.
8375 (tablejump): Emit different RTL for
8376 -mno-speculate-indirect-jumps.
8377 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8378 (tablejumpsi_nospec): New define_expand.
8379 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8380 (tablejumpdi_nospec): New define_expand.
8381 (*tablejump<mode>_internal1): Disable for
8382 -mno-speculate-indirect-jumps.
8383 (*tablejump<mode>_internal1_nospec): New define_insn.
8384 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8387 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8389 * caller-save.c (insert_save): Drop unnecessary parameter. All
8392 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8393 Richard Biener <rguenth@suse.de>
8396 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8397 return early, inline manually is_gimple_sizepos. Make sure if we
8398 call gimplify_expr we don't end up with a gimple constant.
8399 * tree.c (variably_modified_type_p): Don't return true for
8400 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8401 * gimplify.h (is_gimple_sizepos): Remove.
8403 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8405 PR tree-optimization/83857
8406 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8407 vectorizable_live_operation for pure SLP statements.
8408 (vectorizable_live_operation): Handle PHIs.
8410 2018-01-16 Richard Biener <rguenther@suse.de>
8412 PR tree-optimization/83867
8413 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8414 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8416 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8419 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8420 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8421 If off is not INTEGER_CST, issue a may not be aligned warning
8422 rather than isn't aligned. Use isn%'t rather than isn't.
8423 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8425 <case MULT_EXPR>: Improve the case when bottom and one of the
8426 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8427 operand, in that case check if the other operand is multiple of
8428 bottom divided by the INTEGER_CST operand.
8430 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8433 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8434 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8435 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8436 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8437 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8438 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8439 (pa_function_arg_size): New function.
8441 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8443 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8444 in a separate statement.
8446 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8448 PR tree-optimization/83847
8449 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8450 group gathers and scatters.
8452 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8454 PR rtl-optimization/86620
8455 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8457 PR rtl-optimization/83213
8458 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8459 to last if both are JUMP_INSNs.
8461 PR tree-optimization/83843
8462 * gimple-ssa-store-merging.c
8463 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8464 store_immediate_info for bswap/nop orig_stores.
8466 2018-01-15 Andrew Waterman <andrew@sifive.com>
8468 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8470 <UDIV>: Increase cost if !TARGET_DIV.
8472 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8474 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8475 (define_attr "cr_logical_3op"): New.
8476 (cceq_ior_compare): Adjust.
8477 (cceq_ior_compare_complement): Adjust.
8478 (*cceq_rev_compare): Adjust.
8479 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8480 (is_cracked_insn): Adjust.
8481 (insn_must_be_first_in_group): Adjust.
8482 * config/rs6000/40x.md: Adjust.
8483 * config/rs6000/440.md: Adjust.
8484 * config/rs6000/476.md: Adjust.
8485 * config/rs6000/601.md: Adjust.
8486 * config/rs6000/603.md: Adjust.
8487 * config/rs6000/6xx.md: Adjust.
8488 * config/rs6000/7450.md: Adjust.
8489 * config/rs6000/7xx.md: Adjust.
8490 * config/rs6000/8540.md: Adjust.
8491 * config/rs6000/cell.md: Adjust.
8492 * config/rs6000/e300c2c3.md: Adjust.
8493 * config/rs6000/e500mc.md: Adjust.
8494 * config/rs6000/e500mc64.md: Adjust.
8495 * config/rs6000/e5500.md: Adjust.
8496 * config/rs6000/e6500.md: Adjust.
8497 * config/rs6000/mpc.md: Adjust.
8498 * config/rs6000/power4.md: Adjust.
8499 * config/rs6000/power5.md: Adjust.
8500 * config/rs6000/power6.md: Adjust.
8501 * config/rs6000/power7.md: Adjust.
8502 * config/rs6000/power8.md: Adjust.
8503 * config/rs6000/power9.md: Adjust.
8504 * config/rs6000/rs64.md: Adjust.
8505 * config/rs6000/titan.md: Adjust.
8507 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8509 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8510 ix86_indirect_branch_register logic.
8512 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8514 * config/i386/constraints.md (Bs): Update
8515 ix86_indirect_branch_register check. Don't check
8516 ix86_indirect_branch_register with GOT_memory_operand.
8518 * config/i386/predicates.md (GOT_memory_operand): Don't check
8519 ix86_indirect_branch_register here.
8520 (GOT32_symbol_operand): Likewise.
8522 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8524 * config/i386/predicates.md (constant_call_address_operand):
8525 Rewrite ix86_indirect_branch_register logic.
8526 (sibcall_insn_operand): Likewise.
8528 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8530 * config/i386/constraints.md (Bs): Replace
8531 ix86_indirect_branch_thunk_register with
8532 ix86_indirect_branch_register.
8534 * config/i386/i386.md (indirect_jump): Likewise.
8535 (tablejump): Likewise.
8536 (*sibcall_memory): Likewise.
8537 (*sibcall_value_memory): Likewise.
8538 Peepholes of indirect call and jump via memory: Likewise.
8539 * config/i386/i386.opt: Likewise.
8540 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8541 (GOT_memory_operand): Likewise.
8542 (call_insn_operand): Likewise.
8543 (sibcall_insn_operand): Likewise.
8544 (GOT32_symbol_operand): Likewise.
8546 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8549 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8550 type rather than type addr's type points to.
8551 (expand_omp_atomic_mutex): Likewise.
8552 (expand_omp_atomic): Likewise.
8554 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8557 * config/i386/i386.c (output_indirect_thunk_function): Use
8558 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8559 for __x86_return_thunk.
8561 2018-01-15 Richard Biener <rguenther@suse.de>
8564 * expmed.c (extract_bit_field_1): Fix typo.
8566 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8569 * config/arm/iterators.md (VF): New mode iterator.
8570 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8571 Remove integer-related logic from pattern.
8572 (neon_vabd<mode>_3): Likewise.
8574 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8577 * common.opt (fstrict-overflow): No longer an alias.
8578 (fwrapv-pointer): New option.
8579 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8580 also for pointer types based on flag_wrapv_pointer.
8581 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8582 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8583 opts->x_flag_wrapv got set.
8584 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8585 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8586 POINTER_TYPE_OVERFLOW_UNDEFINED.
8587 * match.pd: Likewise in address comparison pattern.
8588 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8590 2018-01-15 Richard Biener <rguenther@suse.de>
8593 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8594 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8595 Reset type names to their identifier if their TYPE_DECL doesn't
8596 have linkage (and thus is used for ODR and devirt).
8597 (save_debug_info_for_decl): Remove.
8598 (save_debug_info_for_type): Likewise.
8599 (add_tree_to_fld_list): Adjust.
8600 * tree-pretty-print.c (dump_generic_node): Make dumping of
8601 type names more robust.
8603 2018-01-15 Richard Biener <rguenther@suse.de>
8605 * BASE-VER: Bump to 8.0.1.
8607 2018-01-14 Martin Sebor <msebor@redhat.com>
8610 * builtins.c (check_access): Avoid warning when the no-warning bit
8613 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8615 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8616 * ira-color (allocno_hard_regs_compare): Likewise.
8618 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8621 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8622 Use .pushsection/.popsection.
8624 2018-01-14 Martin Sebor <msebor@redhat.com>
8627 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8629 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8631 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8632 entry from extra_headers.
8633 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8634 extra_headers, make the list bitwise identical to the i?86-*-* one.
8636 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8638 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8639 -mcmodel=large with -mindirect-branch=thunk,
8640 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8641 -mfunction-return=thunk-extern.
8642 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8643 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8644 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8646 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8648 * config/i386/i386.c (print_reg): Print the name of the full
8649 integer register without '%'.
8650 (ix86_print_operand): Handle 'V'.
8651 * doc/extend.texi: Document 'V' modifier.
8653 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8655 * config/i386/constraints.md (Bs): Disallow memory operand for
8656 -mindirect-branch-register.
8658 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8659 (GOT_memory_operand): Likewise.
8660 (call_insn_operand): Likewise.
8661 (sibcall_insn_operand): Likewise.
8662 (GOT32_symbol_operand): Likewise.
8663 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8664 for -mindirect-branch-register.
8665 (tablejump): Likewise.
8666 (*sibcall_memory): Likewise.
8667 (*sibcall_value_memory): Likewise.
8668 Disallow peepholes of indirect call and jump via memory for
8669 -mindirect-branch-register.
8670 (*call_pop): Replace m with Bw.
8671 (*call_value_pop): Likewise.
8672 (*sibcall_pop_memory): Replace m with Bs.
8673 * config/i386/i386.opt (mindirect-branch-register): New option.
8674 * doc/invoke.texi: Document -mindirect-branch-register option.
8676 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8678 * config/i386/i386-protos.h (ix86_output_function_return): New.
8679 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8680 set function_return_type.
8681 (indirect_thunk_name): Add ret_p to indicate thunk for function
8683 (output_indirect_thunk_function): Pass false to
8684 indirect_thunk_name.
8685 (ix86_output_indirect_branch_via_reg): Likewise.
8686 (ix86_output_indirect_branch_via_push): Likewise.
8687 (output_indirect_thunk_function): Create alias for function
8688 return thunk if regno < 0.
8689 (ix86_output_function_return): New function.
8690 (ix86_handle_fndecl_attribute): Handle function_return.
8691 (ix86_attribute_table): Add function_return.
8692 * config/i386/i386.h (machine_function): Add
8693 function_return_type.
8694 * config/i386/i386.md (simple_return_internal): Use
8695 ix86_output_function_return.
8696 (simple_return_internal_long): Likewise.
8697 * config/i386/i386.opt (mfunction-return=): New option.
8698 (indirect_branch): Mention -mfunction-return=.
8699 * doc/extend.texi: Document function_return function attribute.
8700 * doc/invoke.texi: Document -mfunction-return= option.
8702 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8704 * config/i386/i386-opts.h (indirect_branch): New.
8705 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8706 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8707 with local indirect jump when converting indirect call and jump.
8708 (ix86_set_indirect_branch_type): New.
8709 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8710 (indirectlabelno): New.
8711 (indirect_thunk_needed): Likewise.
8712 (indirect_thunk_bnd_needed): Likewise.
8713 (indirect_thunks_used): Likewise.
8714 (indirect_thunks_bnd_used): Likewise.
8715 (INDIRECT_LABEL): Likewise.
8716 (indirect_thunk_name): Likewise.
8717 (output_indirect_thunk): Likewise.
8718 (output_indirect_thunk_function): Likewise.
8719 (ix86_output_indirect_branch_via_reg): Likewise.
8720 (ix86_output_indirect_branch_via_push): Likewise.
8721 (ix86_output_indirect_branch): Likewise.
8722 (ix86_output_indirect_jmp): Likewise.
8723 (ix86_code_end): Call output_indirect_thunk_function if needed.
8724 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8726 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8727 (ix86_attribute_table): Add indirect_branch.
8728 * config/i386/i386.h (machine_function): Add indirect_branch_type
8729 and has_local_indirect_jump.
8730 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8732 (tablejump): Likewise.
8733 (*indirect_jump): Use ix86_output_indirect_jmp.
8734 (*tablejump_1): Likewise.
8735 (simple_return_indirect_internal): Likewise.
8736 * config/i386/i386.opt (mindirect-branch=): New option.
8737 (indirect_branch): New.
8740 (thunk-inline): Likewise.
8741 (thunk-extern): Likewise.
8742 * doc/extend.texi: Document indirect_branch function attribute.
8743 * doc/invoke.texi: Document -mindirect-branch= option.
8745 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8748 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8750 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8752 * ipa-inline.c (want_inline_small_function_p): Return false if
8753 inlining has already failed with CIF_FINAL_ERROR.
8754 (update_caller_keys): Call want_inline_small_function_p before
8756 (update_callee_keys): Likewise.
8758 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8760 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8762 (rs6000_quadword_masked_address_p): Likewise.
8763 (quad_aligned_load_p): Likewise.
8764 (quad_aligned_store_p): Likewise.
8765 (const_load_sequence_p): Add comment to describe the outer-most loop.
8766 (mimic_memory_attributes_and_flags): New function.
8767 (rs6000_gen_stvx): Likewise.
8768 (replace_swapped_aligned_store): Likewise.
8769 (rs6000_gen_lvx): Likewise.
8770 (replace_swapped_aligned_load): Likewise.
8771 (replace_swapped_load_constant): Capitalize argument name in
8772 comment describing this function.
8773 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8774 and stores that access quad-word aligned addresses and replace
8775 with stvx or lvx instructions when appropriate.
8776 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8777 New function prototype.
8778 (rs6000_quadword_masked_address_p): Likewise.
8779 (rs6000_gen_lvx): Likewise.
8780 (rs6000_gen_stvx): Likewise.
8781 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8782 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8783 when memory address is aligned.
8784 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8785 this split to select lvx instruction when memory address is aligned.
8786 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8787 instruction when memory address is aligned.
8788 (*vsx_le_perm_load_v16qi): Likewise.
8789 (four unnamed splitters): Modify to select the stvx instruction
8790 when memory is aligned.
8792 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8794 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8795 which appears in the queue multiple times.
8797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8798 Alan Hayward <alan.hayward@arm.com>
8799 David Sherwood <david.sherwood@arm.com>
8801 * tree-vectorizer.h (vec_lower_bound): New structure.
8802 (_loop_vec_info): Add check_nonzero and lower_bounds.
8803 (LOOP_VINFO_CHECK_NONZERO): New macro.
8804 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8805 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8806 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8807 fields. Make seg_len the distance travelled, not including the
8809 (dr_direction_indicator): Declare.
8810 (dr_zero_step_indicator): Likewise.
8811 (dr_known_forward_stride_p): Likewise.
8812 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8814 (runtime_alias_check_p): Allow runtime alias checks with
8816 (operator ==): Compare access_size and align.
8817 (prune_runtime_alias_test_list): Rework for new distinction between
8818 the access_size and seg_len.
8819 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8821 (get_segment_min_max): New function.
8822 (create_intersect_range_checks): Use it.
8823 (dr_step_indicator): New function.
8824 (dr_direction_indicator): Likewise.
8825 (dr_zero_step_indicator): Likewise.
8826 (dr_known_forward_stride_p): Likewise.
8827 * tree-loop-distribution.c (data_ref_segment_size): Return
8828 DR_STEP * (niters - 1).
8829 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8831 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8832 (vect_preserves_scalar_order_p): New function, split out from...
8833 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8834 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8835 (vect_vfa_access_size): New function.
8836 (vect_vfa_align): Likewise.
8837 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8838 (dump_lower_bound): New function.
8839 (vect_check_lower_bound): Likewise.
8840 (vect_small_gap_p): Likewise.
8841 (vectorizable_with_step_bound_p): Likewise.
8842 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8843 depencies if the vectorization factor is 1. Convert the checks
8844 for nonzero steps into checks on the bounds of DR_STEP. Try using
8845 a bunds check for variable steps if the minimum required step is
8846 relatively small. Update calls to the dr_with_seg_len
8847 constructor and to vect_compile_time_alias.
8848 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8850 (vect_loop_versioning): Call it.
8851 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8853 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8856 Alan Hayward <alan.hayward@arm.com>
8857 David Sherwood <david.sherwood@arm.com>
8859 * doc/sourcebuild.texi (vect_scatter_store): Document.
8860 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8862 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8864 * genopinit.c (main): Add supports_vec_scatter_store and
8865 supports_vec_scatter_store_cached to target_optabs.
8866 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8867 IFN_MASK_SCATTER_STORE.
8868 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8870 * internal-fn.h (internal_store_fn_p): Declare.
8871 (internal_fn_stored_value_index): Likewise.
8872 * internal-fn.c (scatter_store_direct): New macro.
8873 (expand_scatter_store_optab_fn): New function.
8874 (direct_scatter_store_optab_supported_p): New macro.
8875 (internal_store_fn_p): New function.
8876 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8877 IFN_MASK_SCATTER_STORE.
8878 (internal_fn_mask_index): Likewise.
8879 (internal_fn_stored_value_index): New function.
8880 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8882 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8883 * optabs-query.c (supports_vec_scatter_store_p): New function.
8884 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8885 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8886 true for scatter stores.
8887 (vect_gather_scatter_fn_p): Handle scatter stores too.
8888 (vect_check_gather_scatter): Consider using scatter stores if
8889 supports_vec_scatter_store_p.
8890 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8892 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8893 internal_fn_stored_value_index.
8894 (check_load_store_masking): Handle scatter stores too.
8895 (vect_get_store_rhs): Make public.
8896 (vectorizable_call): Use internal_store_fn_p.
8897 (vectorizable_store): Handle scatter store internal functions.
8898 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8899 when deciding whether the end of the group has been reached.
8900 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8901 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8902 (mask_scatter_store<mode>): New insns.
8904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8905 Alan Hayward <alan.hayward@arm.com>
8906 David Sherwood <david.sherwood@arm.com>
8908 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8909 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8910 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8912 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8913 Use vect_truncate_gather_scatter_offset if we can't treat the
8914 operation as a normal gather load or scatter store.
8915 (get_group_load_store_type): Take the gather_scatter_info
8916 as argument. Try using a gather load or scatter store for
8917 single-element groups.
8918 (get_load_store_type): Update calls to get_group_load_store_type
8919 and vect_use_strided_gather_scatters_p.
8921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8922 Alan Hayward <alan.hayward@arm.com>
8923 David Sherwood <david.sherwood@arm.com>
8925 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8926 optional tree argument.
8927 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8929 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8930 but continue to use the current value as a fallback.
8931 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8932 to compare the updates.
8933 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8934 (get_load_store_type): Use it when handling a strided access.
8935 (vect_get_strided_load_store_ops): New function.
8936 (vect_get_data_ptr_increment): Likewise.
8937 (vectorizable_load): Handle strided gather loads. Always pass
8938 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8941 Alan Hayward <alan.hayward@arm.com>
8942 David Sherwood <david.sherwood@arm.com>
8944 * doc/md.texi (gather_load@var{m}): Document.
8945 (mask_gather_load@var{m}): Likewise.
8946 * genopinit.c (main): Add supports_vec_gather_load and
8947 supports_vec_gather_load_cached to target_optabs.
8948 * optabs-tree.c (init_tree_optimization_optabs): Use
8949 ggc_cleared_alloc to allocate target_optabs.
8950 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8951 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8953 * internal-fn.h (internal_load_fn_p): Declare.
8954 (internal_gather_scatter_fn_p): Likewise.
8955 (internal_fn_mask_index): Likewise.
8956 (internal_gather_scatter_fn_supported_p): Likewise.
8957 * internal-fn.c (gather_load_direct): New macro.
8958 (expand_gather_load_optab_fn): New function.
8959 (direct_gather_load_optab_supported_p): New macro.
8960 (direct_internal_fn_optab): New function.
8961 (internal_load_fn_p): Likewise.
8962 (internal_gather_scatter_fn_p): Likewise.
8963 (internal_fn_mask_index): Likewise.
8964 (internal_gather_scatter_fn_supported_p): Likewise.
8965 * optabs-query.c (supports_at_least_one_mode_p): New function.
8966 (supports_vec_gather_load_p): Likewise.
8967 * optabs-query.h (supports_vec_gather_load_p): Declare.
8968 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8969 and memory_type field.
8970 (NUM_PATTERNS): Bump to 15.
8971 * tree-vect-data-refs.c: Include internal-fn.h.
8972 (vect_gather_scatter_fn_p): New function.
8973 (vect_describe_gather_scatter_call): Likewise.
8974 (vect_check_gather_scatter): Try using internal functions for
8975 gather loads. Recognize existing calls to a gather load function.
8976 (vect_analyze_data_refs): Consider using gather loads if
8977 supports_vec_gather_load_p.
8978 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8979 (vect_get_gather_scatter_offset_type): Likewise.
8980 (vect_convert_mask_for_vectype): Likewise.
8981 (vect_add_conversion_to_patterm): Likewise.
8982 (vect_try_gather_scatter_pattern): Likewise.
8983 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8984 (vect_vect_recog_func_ptrs): Add it.
8985 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8986 internal_fn_mask_index and internal_gather_scatter_fn_p.
8987 (check_load_store_masking): Take the gather_scatter_info as an
8988 argument and handle gather loads.
8989 (vect_get_gather_scatter_ops): New function.
8990 (vectorizable_call): Check internal_load_fn_p.
8991 (vectorizable_load): Likewise. Handle gather load internal
8993 (vectorizable_store): Update call to check_load_store_masking.
8994 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8995 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8996 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8997 (aarch64_gather_scale_operand_d): New predicates.
8998 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8999 (mask_gather_load<mode>): New insns.
9001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9002 Alan Hayward <alan.hayward@arm.com>
9003 David Sherwood <david.sherwood@arm.com>
9005 * optabs.def (fold_left_plus_optab): New optab.
9006 * doc/md.texi (fold_left_plus_@var{m}): Document.
9007 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9008 * internal-fn.c (fold_left_direct): Define.
9009 (expand_fold_left_optab_fn): Likewise.
9010 (direct_fold_left_optab_supported_p): Likewise.
9011 * fold-const-call.c (fold_const_fold_left): New function.
9012 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9013 * tree-parloops.c (valid_reduction_p): New function.
9014 (gather_scalar_reductions): Use it.
9015 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9016 (vect_finish_replace_stmt): Declare.
9017 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9018 (needs_fold_left_reduction_p): New function, split out from...
9019 (vect_is_simple_reduction): ...here. Accept reductions that
9020 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9021 (vect_force_simple_reduction): Also store the reduction type in
9022 the assignment's STMT_VINFO_REDUC_TYPE.
9023 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9024 (merge_with_identity): New function.
9025 (vect_expand_fold_left): Likewise.
9026 (vectorize_fold_left_reduction): Likewise.
9027 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9028 scalar phi in place for it. Check for target support and reject
9029 cases that would reassociate the operation. Defer the transform
9030 phase to vectorize_fold_left_reduction.
9031 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9032 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9033 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9037 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9038 call to ifc_temp_var.
9040 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9041 Alan Hayward <alan.hayward@arm.com>
9042 David Sherwood <david.sherwood@arm.com>
9044 * target.def (legitimize_address_displacement): Take the original
9045 offset as a poly_int.
9046 * targhooks.h (default_legitimize_address_displacement): Update
9048 * targhooks.c (default_legitimize_address_displacement): Likewise.
9049 * doc/tm.texi: Regenerate.
9050 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9051 as an argument, moving assert of ad->disp == ad->disp_term to...
9052 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9053 Try calling targetm.legitimize_address_displacement before expanding
9054 the address rather than afterwards, and adjust for the new interface.
9055 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9056 Match the new hook interface. Handle SVE addresses.
9057 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9062 * Makefile.in (OBJS): Add early-remat.o.
9063 * target.def (select_early_remat_modes): New hook.
9064 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9065 * doc/tm.texi: Regenerate.
9066 * targhooks.h (default_select_early_remat_modes): Declare.
9067 * targhooks.c (default_select_early_remat_modes): New function.
9068 * timevar.def (TV_EARLY_REMAT): New timevar.
9069 * passes.def (pass_early_remat): New pass.
9070 * tree-pass.h (make_pass_early_remat): Declare.
9071 * early-remat.c: New file.
9072 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9074 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9076 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9077 Alan Hayward <alan.hayward@arm.com>
9078 David Sherwood <david.sherwood@arm.com>
9080 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9081 vfm1 with a bound_epilog parameter.
9082 (vect_do_peeling): Update calls accordingly, and move the prologue
9083 call earlier in the function. Treat the base bound_epilog as 0 for
9084 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9085 this base when peeling for gaps.
9086 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9087 with fully-masked loops.
9088 (vect_estimate_min_profitable_iters): Handle the single peeled
9089 iteration in that case.
9091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9092 Alan Hayward <alan.hayward@arm.com>
9093 David Sherwood <david.sherwood@arm.com>
9095 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9096 single-element interleaving even if the size is not a power of 2.
9097 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9098 accesses for single-element interleaving if the group size is
9101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9102 Alan Hayward <alan.hayward@arm.com>
9103 David Sherwood <david.sherwood@arm.com>
9105 * doc/md.texi (fold_extract_last_@var{m}): Document.
9106 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9107 * optabs.def (fold_extract_last_optab): New optab.
9108 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9109 * internal-fn.c (fold_extract_direct): New macro.
9110 (expand_fold_extract_optab_fn): Likewise.
9111 (direct_fold_extract_optab_supported_p): Likewise.
9112 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9113 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9114 EXTRACT_LAST_REDUCTION.
9115 (get_initial_def_for_reduction): Do not create an initial vector
9116 for EXTRACT_LAST_REDUCTION reductions.
9117 (vectorizable_reduction): Leave the scalar phi in place for
9118 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9119 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9120 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9121 transform phase to vectorizable_condition.
9122 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9124 (vect_finish_stmt_generation): ...here.
9125 (vect_finish_replace_stmt): New function.
9126 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9127 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9129 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9132 Alan Hayward <alan.hayward@arm.com>
9133 David Sherwood <david.sherwood@arm.com>
9135 * doc/md.texi (extract_last_@var{m}): Document.
9136 * optabs.def (extract_last_optab): New optab.
9137 * internal-fn.def (EXTRACT_LAST): New internal function.
9138 * internal-fn.c (cond_unary_direct): New macro.
9139 (expand_cond_unary_optab_fn): Likewise.
9140 (direct_cond_unary_optab_supported_p): Likewise.
9141 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9142 loops using EXTRACT_LAST.
9143 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9144 (extract_last_<mode>): ...this optab.
9145 (vec_extract<mode><Vel>): Update accordingly.
9147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9148 Alan Hayward <alan.hayward@arm.com>
9149 David Sherwood <david.sherwood@arm.com>
9151 * target.def (empty_mask_is_expensive): New hook.
9152 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9153 * doc/tm.texi: Regenerate.
9154 * targhooks.h (default_empty_mask_is_expensive): Declare.
9155 * targhooks.c (default_empty_mask_is_expensive): New function.
9156 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9157 if the target says that empty masks are expensive.
9158 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9160 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9163 Alan Hayward <alan.hayward@arm.com>
9164 David Sherwood <david.sherwood@arm.com>
9166 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9167 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9168 (vect_use_loop_mask_for_alignment_p): New function.
9169 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9170 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9171 niters_skip argument. Make sure that the first niters_skip elements
9172 of the first iteration are inactive.
9173 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9174 Update call to vect_set_loop_masks_directly.
9175 (get_misalign_in_elems): New function, split out from...
9176 (vect_gen_prolog_loop_niters): ...here.
9177 (vect_update_init_of_dr): Take a code argument that specifies whether
9178 the adjustment should be added or subtracted.
9179 (vect_update_init_of_drs): Likewise.
9180 (vect_prepare_for_masked_peels): New function.
9181 (vect_do_peeling): Skip prologue peeling if we're using a mask
9182 instead. Update call to vect_update_inits_of_drs.
9183 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9185 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9186 alignment. Do not include the number of peeled iterations in
9187 the minimum threshold in that case.
9188 (vectorizable_induction): Adjust the start value down by
9189 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9190 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9191 Take the number of skipped iterations into account when calculating
9193 * tree-vect-stmts.c (vect_gen_while_not): New function.
9195 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9196 Alan Hayward <alan.hayward@arm.com>
9197 David Sherwood <david.sherwood@arm.com>
9199 * doc/sourcebuild.texi (vect_fully_masked): Document.
9200 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9202 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9204 (vect_analyze_loop_2): ...here. Don't check the vectorization
9205 factor against the number of loop iterations if the loop is
9208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9209 Alan Hayward <alan.hayward@arm.com>
9210 David Sherwood <david.sherwood@arm.com>
9212 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9213 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9214 (dump_groups): Update accordingly.
9215 (iv_use::mem_type): New member variable.
9216 (address_p): New function.
9217 (record_use): Add a mem_type argument and initialize the new
9219 (record_group_use): Add a mem_type argument. Use address_p.
9220 Remove obsolete null checks of base_object. Update call to record_use.
9221 (find_interesting_uses_op): Update call to record_group_use.
9222 (find_interesting_uses_cond): Likewise.
9223 (find_interesting_uses_address): Likewise.
9224 (get_mem_type_for_internal_fn): New function.
9225 (find_address_like_use): Likewise.
9226 (find_interesting_uses_stmt): Try find_address_like_use before
9227 calling find_interesting_uses_op.
9228 (addr_offset_valid_p): Use the iv mem_type field as the type
9229 of the addressed memory.
9230 (add_autoinc_candidates): Likewise.
9231 (get_address_cost): Likewise.
9232 (split_small_address_groups_p): Use address_p.
9233 (split_address_groups): Likewise.
9234 (add_iv_candidate_for_use): Likewise.
9235 (autoinc_possible_for_pair): Likewise.
9236 (rewrite_groups): Likewise.
9237 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9238 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9239 (get_alias_ptr_type_for_ptr_address): New function.
9240 (rewrite_use_address): Rewrite address uses in calls that were
9241 identified by find_address_like_use.
9243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9244 Alan Hayward <alan.hayward@arm.com>
9245 David Sherwood <david.sherwood@arm.com>
9247 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9249 * gimple-expr.h (is_gimple_addressable: Likewise.
9250 * gimple-expr.c (is_gimple_address): Likewise.
9251 * internal-fn.c (expand_call_mem_ref): New function.
9252 (expand_mask_load_optab_fn): Use it.
9253 (expand_mask_store_optab_fn): Likewise.
9255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9256 Alan Hayward <alan.hayward@arm.com>
9257 David Sherwood <david.sherwood@arm.com>
9259 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9260 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9261 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9262 (cond_umax@var{mode}): Document.
9263 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9264 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9265 (cond_umin_optab, cond_umax_optab): New optabs.
9266 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9267 (COND_IOR, COND_XOR): New internal functions.
9268 * internal-fn.h (get_conditional_internal_fn): Declare.
9269 * internal-fn.c (cond_binary_direct): New macro.
9270 (expand_cond_binary_optab_fn): Likewise.
9271 (direct_cond_binary_optab_supported_p): Likewise.
9272 (get_conditional_internal_fn): New function.
9273 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9274 Cope with reduction statements that are vectorized as calls rather
9276 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9277 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9278 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9279 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9280 (UNSPEC_COND_EOR): New unspecs.
9281 (optab): Add mappings for them.
9282 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9283 (sve_int_op, sve_fp_op): New int attributes.
9285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9286 Alan Hayward <alan.hayward@arm.com>
9287 David Sherwood <david.sherwood@arm.com>
9289 * optabs.def (while_ult_optab): New optab.
9290 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9291 * internal-fn.def (WHILE_ULT): New internal function.
9292 * internal-fn.h (direct_internal_fn_supported_p): New override
9293 that takes two types as argument.
9294 * internal-fn.c (while_direct): New macro.
9295 (expand_while_optab_fn): New function.
9296 (convert_optab_supported_p): Likewise.
9297 (direct_while_optab_supported_p): New macro.
9298 * wide-int.h (wi::udiv_ceil): New function.
9299 * tree-vectorizer.h (rgroup_masks): New structure.
9300 (vec_loop_masks): New typedef.
9301 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9303 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9304 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9305 (vect_max_vf): New function.
9306 (slpeel_make_loop_iterate_ntimes): Delete.
9307 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9308 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9309 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9310 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9311 internal-fn.h, stor-layout.h and optabs-query.h.
9312 (vect_set_loop_mask): New function.
9313 (add_preheader_seq): Likewise.
9314 (add_header_seq): Likewise.
9315 (interleave_supported_p): Likewise.
9316 (vect_maybe_permute_loop_masks): Likewise.
9317 (vect_set_loop_masks_directly): Likewise.
9318 (vect_set_loop_condition_masked): Likewise.
9319 (vect_set_loop_condition_unmasked): New function, split out from
9320 slpeel_make_loop_iterate_ntimes.
9321 (slpeel_make_loop_iterate_ntimes): Rename to..
9322 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9323 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9324 (vect_do_peeling): Update call accordingly.
9325 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9327 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9328 mask_compare_type, can_fully_mask_p and fully_masked_p.
9329 (release_vec_loop_masks): New function.
9330 (_loop_vec_info): Use it to free the loop masks.
9331 (can_produce_all_loop_masks_p): New function.
9332 (vect_get_max_nscalars_per_iter): Likewise.
9333 (vect_verify_full_masking): Likewise.
9334 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9335 retries, and free the mask rgroups before retrying. Check loop-wide
9336 reasons for disallowing fully-masked loops. Make the final decision
9337 about whether use a fully-masked loop or not.
9338 (vect_estimate_min_profitable_iters): Do not assume that peeling
9339 for the number of iterations will be needed for fully-masked loops.
9340 (vectorizable_reduction): Disable fully-masked loops.
9341 (vectorizable_live_operation): Likewise.
9342 (vect_halve_mask_nunits): New function.
9343 (vect_double_mask_nunits): Likewise.
9344 (vect_record_loop_mask): Likewise.
9345 (vect_get_loop_mask): Likewise.
9346 (vect_transform_loop): Handle the case in which the final loop
9347 iteration might handle a partial vector. Call vect_set_loop_condition
9348 instead of slpeel_make_loop_iterate_ntimes.
9349 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9350 (check_load_store_masking): New function.
9351 (prepare_load_store_mask): Likewise.
9352 (vectorizable_store): Handle fully-masked loops.
9353 (vectorizable_load): Likewise.
9354 (supportable_widening_operation): Use vect_halve_mask_nunits for
9356 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9357 (vect_gen_while): New function.
9358 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9359 (aarch64_uqdec<mode>): New insn.
9361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9362 Alan Hayward <alan.hayward@arm.com>
9363 David Sherwood <david.sherwood@arm.com>
9365 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9366 (reduc_xor_scal_optab): New optabs.
9367 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9368 (reduc_xor_scal_@var{m}): Document.
9369 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9370 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9372 * fold-const-call.c (fold_const_call): Handle them.
9373 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9374 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9375 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9376 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9377 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9378 (UNSPEC_XORV): New unspecs.
9379 (optab): Add entries for them.
9380 (BITWISEV): New int iterator.
9381 (bit_reduc_op): New int attributes.
9383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9384 Alan Hayward <alan.hayward@arm.com>
9385 David Sherwood <david.sherwood@arm.com>
9387 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9388 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9389 * optabs.def (vec_shl_insert_optab): New optab.
9390 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9391 (duplicate_and_interleave): Likewise.
9392 * tree-vect-loop.c: Include internal-fn.h.
9393 (neutral_op_for_slp_reduction): New function, split out from
9394 get_initial_defs_for_reduction.
9395 (get_initial_def_for_reduction): Handle option 2 for variable-length
9396 vectors by loading the neutral value into a vector and then shifting
9397 the initial value into element 0.
9398 (get_initial_defs_for_reduction): Replace the code argument with
9399 the neutral value calculated by neutral_op_for_slp_reduction.
9400 Use gimple_build_vector for constant-length vectors.
9401 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9402 but the first group_size elements have a neutral value.
9403 Use duplicate_and_interleave otherwise.
9404 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9405 Update call to get_initial_defs_for_reduction. Handle SLP
9406 reductions for variable-length vectors by creating one vector
9407 result for each scalar result, with the elements associated
9408 with other scalar results stubbed out with the neutral value.
9409 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9410 Require IFN_VEC_SHL_INSERT for double reductions on
9411 variable-length vectors, or SLP reductions that have
9412 a neutral value. Require can_duplicate_and_interleave_p
9413 support for variable-length unchained SLP reductions if there
9414 is no neutral value, such as for MIN/MAX reductions. Also require
9415 the number of vector elements to be a multiple of the number of
9416 SLP statements when doing variable-length unchained SLP reductions.
9417 Update call to vect_create_epilog_for_reduction.
9418 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9419 and remove initial values.
9420 (duplicate_and_interleave): Make public.
9421 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9422 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9424 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9425 Alan Hayward <alan.hayward@arm.com>
9426 David Sherwood <david.sherwood@arm.com>
9428 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9429 (can_duplicate_and_interleave_p): New function.
9430 (vect_get_and_check_slp_defs): Take the vector of statements
9431 rather than just the current one. Remove excess parentheses.
9432 Restriction rejectinon of vect_constant_def and vect_external_def
9433 for variable-length vectors to boolean types, or types for which
9434 can_duplicate_and_interleave_p is false.
9435 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9436 (duplicate_and_interleave): New function.
9437 (vect_get_constant_vectors): Use gimple_build_vector for
9438 constant-length vectors and suitable variable-length constant
9439 vectors. Use duplicate_and_interleave for other variable-length
9440 vectors. Don't defer the update when inserting new statements.
9442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9443 Alan Hayward <alan.hayward@arm.com>
9444 David Sherwood <david.sherwood@arm.com>
9446 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9447 min_profitable_iters doesn't go negative.
9449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9450 Alan Hayward <alan.hayward@arm.com>
9451 David Sherwood <david.sherwood@arm.com>
9453 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9454 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9455 * optabs.def (vec_mask_load_lanes_optab): New optab.
9456 (vec_mask_store_lanes_optab): Likewise.
9457 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9458 (MASK_STORE_LANES): Likewise.
9459 * internal-fn.c (mask_load_lanes_direct): New macro.
9460 (mask_store_lanes_direct): Likewise.
9461 (expand_mask_load_optab_fn): Handle masked operations.
9462 (expand_mask_load_lanes_optab_fn): New macro.
9463 (expand_mask_store_optab_fn): Handle masked operations.
9464 (expand_mask_store_lanes_optab_fn): New macro.
9465 (direct_mask_load_lanes_optab_supported_p): Likewise.
9466 (direct_mask_store_lanes_optab_supported_p): Likewise.
9467 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9469 (vect_load_lanes_supported): Likewise.
9470 * tree-vect-data-refs.c (strip_conversion): New function.
9471 (can_group_stmts_p): Likewise.
9472 (vect_analyze_data_ref_accesses): Use it instead of checking
9473 for a pair of assignments.
9474 (vect_store_lanes_supported): Take a masked_p parameter.
9475 (vect_load_lanes_supported): Likewise.
9476 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9477 vect_store_lanes_supported and vect_load_lanes_supported.
9478 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9479 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9480 parameter. Don't allow gaps for masked accesses.
9481 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9482 and vect_load_lanes_supported.
9483 (get_load_store_type): Take a masked_p parameter and update
9484 call to get_group_load_store_type.
9485 (vectorizable_store): Update call to get_load_store_type.
9486 Handle IFN_MASK_STORE_LANES.
9487 (vectorizable_load): Update call to get_load_store_type.
9488 Handle IFN_MASK_LOAD_LANES.
9490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9491 Alan Hayward <alan.hayward@arm.com>
9492 David Sherwood <david.sherwood@arm.com>
9494 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9496 * config/aarch64/aarch64-protos.h
9497 (aarch64_sve_struct_memory_operand_p): Declare.
9498 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9499 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9500 (VPRED, vpred): Handle SVE structure modes.
9501 * config/aarch64/constraints.md (Utx): New constraint.
9502 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9503 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9504 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9505 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9506 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9507 structure modes. Split into pieces after RA.
9508 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9509 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9511 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9512 SVE structure modes.
9513 (aarch64_classify_address): Likewise.
9514 (sizetochar): Move earlier in file.
9515 (aarch64_print_operand): Handle SVE register lists.
9516 (aarch64_array_mode): New function.
9517 (aarch64_sve_struct_memory_operand_p): Likewise.
9518 (TARGET_ARRAY_MODE): Redefine.
9520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9521 Alan Hayward <alan.hayward@arm.com>
9522 David Sherwood <david.sherwood@arm.com>
9524 * target.def (array_mode): New target hook.
9525 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9526 * doc/tm.texi: Regenerate.
9527 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9528 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9529 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9531 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9535 Alan Hayward <alan.hayward@arm.com>
9536 David Sherwood <david.sherwood@arm.com>
9538 * fold-const.c (fold_binary_loc): Check the argument types
9539 rather than the result type when testing for a vector operation.
9541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9543 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9544 * doc/tm.texi: Regenerate.
9546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9547 Alan Hayward <alan.hayward@arm.com>
9548 David Sherwood <david.sherwood@arm.com>
9550 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9551 (sve): Document new AArch64 extension.
9552 * doc/md.texi (w): Extend the description of the AArch64
9553 constraint to include SVE vectors.
9554 (Upl, Upa): Document new AArch64 predicate constraints.
9555 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9557 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9558 (msve-vector-bits=): New option.
9559 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9560 SVE when these are disabled.
9561 (sve): New extension.
9562 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9563 modes. Adjust their number of units based on aarch64_sve_vg.
9564 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9565 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9566 aarch64_addr_query_type.
9567 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9568 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9569 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9570 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9571 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9572 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9573 (aarch64_simd_imm_zero_p): Delete.
9574 (aarch64_check_zero_based_sve_index_immediate): Declare.
9575 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9576 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9577 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9578 (aarch64_sve_float_mul_immediate_p): Likewise.
9579 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9581 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9582 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9583 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9584 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9585 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9586 (aarch64_regmode_natural_size): Likewise.
9587 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9588 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9590 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9591 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9592 for VG and the SVE predicate registers.
9593 (V_ALIASES): Add a "z"-prefixed alias.
9594 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9595 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9596 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9597 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9598 (REG_CLASS_NAMES): Add entries for them.
9599 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9600 and the predicate registers.
9601 (aarch64_sve_vg): Declare.
9602 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9603 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9604 (REGMODE_NATURAL_SIZE): Define.
9605 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9607 * config/aarch64/aarch64.c: Include cfgrtl.h.
9608 (simd_immediate_info): Add a constructor for series vectors,
9609 and an associated step field.
9610 (aarch64_sve_vg): New variable.
9611 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9612 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9613 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9614 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9615 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9616 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9617 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9618 (aarch64_get_mask_mode): New functions.
9619 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9620 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9621 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9622 predicate modes and predicate registers. Explicitly restrict
9623 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9624 to store a vector mode if it is recognized by
9625 aarch64_classify_vector_mode.
9626 (aarch64_regmode_natural_size): New function.
9627 (aarch64_hard_regno_caller_save_mode): Return the original mode
9629 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9630 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9631 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9632 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9634 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9635 does not overlap dest if the function is frame-related. Handle
9637 (aarch64_split_add_offset): New function.
9638 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9639 them aarch64_add_offset.
9640 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9641 and update call to aarch64_sub_sp.
9642 (aarch64_add_cfa_expression): New function.
9643 (aarch64_expand_prologue): Pass extra temporary registers to the
9644 functions above. Handle the case in which we need to emit new
9645 DW_CFA_expressions for registers that were originally saved
9646 relative to the stack pointer, but now have to be expressed
9647 relative to the frame pointer.
9648 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9650 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9651 IP0 and IP1 values for SVE frames.
9652 (aarch64_expand_vec_series): New function.
9653 (aarch64_expand_sve_widened_duplicate): Likewise.
9654 (aarch64_expand_sve_const_vector): Likewise.
9655 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9656 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9657 into the register, rather than emitting a SET directly.
9658 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9659 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9660 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9661 (offset_9bit_signed_scaled_p): New functions.
9662 (aarch64_replicate_bitmask_imm): New function.
9663 (aarch64_bitmask_imm): Use it.
9664 (aarch64_cannot_force_const_mem): Reject expressions involving
9665 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9666 (aarch64_classify_index): Handle SVE indices, by requiring
9667 a plain register index with a scale that matches the element size.
9668 (aarch64_classify_address): Handle SVE addresses. Assert that
9669 the mode of the address is VOIDmode or an integer mode.
9670 Update call to aarch64_classify_symbol.
9671 (aarch64_classify_symbolic_expression): Update call to
9672 aarch64_classify_symbol.
9673 (aarch64_const_vec_all_in_range_p): New function.
9674 (aarch64_print_vector_float_operand): Likewise.
9675 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9676 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9677 and the FP immediates 1.0 and 0.5.
9678 (aarch64_print_address_internal): Handle SVE addresses.
9679 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9680 (aarch64_regno_regclass): Handle predicate registers.
9681 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9683 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9684 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9685 (aarch64_convert_sve_vector_bits): New function.
9686 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9687 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9689 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9690 Handle SVE vector and predicate modes. Accept VL-based constants
9691 that need only one temporary register, and VL offsets that require
9692 no temporary registers.
9693 (aarch64_conditional_register_usage): Mark the predicate registers
9694 as fixed if SVE isn't available.
9695 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9696 Return true for SVE vector and predicate modes.
9697 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9698 rather than an unsigned int. Handle SVE modes.
9699 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9701 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9703 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9704 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9705 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9706 (aarch64_sve_float_mul_immediate_p): New functions.
9707 (aarch64_sve_valid_immediate): New function.
9708 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9709 Explicitly reject structure modes. Check for INDEX constants.
9710 Handle PTRUE and PFALSE constants.
9711 (aarch64_check_zero_based_sve_index_immediate): New function.
9712 (aarch64_simd_imm_zero_p): Delete.
9713 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9714 vector modes. Accept constants in the range of CNT[BHWD].
9715 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9716 ask for an Advanced SIMD mode.
9717 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9718 (aarch64_simd_vector_alignment): Handle SVE predicates.
9719 (aarch64_vectorize_preferred_vector_alignment): New function.
9720 (aarch64_simd_vector_alignment_reachable): Use it instead of
9722 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9723 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9725 (MAX_VECT_LEN): Delete.
9726 (expand_vec_perm_d): Add a vec_flags field.
9727 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9728 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9729 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9731 (aarch64_evpc_rev): Rename to...
9732 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9733 (aarch64_evpc_rev_global): New function.
9734 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9735 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9737 (aarch64_evpc_sve_tbl): New function.
9738 (aarch64_expand_vec_perm_const_1): Update after rename of
9739 aarch64_evpc_rev. Handle SVE permutes too, trying
9740 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9741 than aarch64_evpc_tbl.
9742 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9743 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9744 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9745 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9746 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9747 (aarch64_expand_sve_vcond): New functions.
9748 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9749 of aarch64_vector_mode_p.
9750 (aarch64_dwarf_poly_indeterminate_value): New function.
9751 (aarch64_compute_pressure_classes): Likewise.
9752 (aarch64_can_change_mode_class): Likewise.
9753 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9754 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9755 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9756 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9757 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9758 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9759 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9760 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9762 (Dn, Dl, Dr): Accept const as well as const_vector.
9763 (Dz): Likewise. Compare against CONST0_RTX.
9764 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9765 of "vector" where appropriate.
9766 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9767 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9768 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9769 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9770 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9771 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9772 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9773 (v_int_equiv): Extend to SVE modes.
9774 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9776 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9777 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9778 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9779 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9780 (SVE_COND_FP_CMP): New int iterators.
9781 (perm_hilo): Handle the new unpack unspecs.
9782 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9784 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9785 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9786 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9787 (aarch64_equality_operator, aarch64_constant_vector_operand)
9788 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9789 (aarch64_sve_nonimmediate_operand): Likewise.
9790 (aarch64_sve_general_operand): Likewise.
9791 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9792 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9793 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9794 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9795 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9796 (aarch64_sve_float_arith_immediate): Likewise.
9797 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9798 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9799 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9800 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9801 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9802 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9803 (aarch64_sve_float_arith_operand): Likewise.
9804 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9805 (aarch64_sve_float_mul_operand): Likewise.
9806 (aarch64_sve_vec_perm_operand): Likewise.
9807 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9808 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9809 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9810 as well as const_vector.
9811 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9812 in file. Use CONST0_RTX and CONSTM1_RTX.
9813 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9814 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9815 Use aarch64_simd_imm_zero.
9816 * config/aarch64/aarch64-sve.md: New file.
9817 * config/aarch64/aarch64.md: Include it.
9818 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9819 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9820 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9821 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9822 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9823 (sve): New attribute.
9824 (enabled): Disable instructions with the sve attribute unless
9826 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9827 aarch64_expand_mov_immediate.
9828 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9829 CNT[BHSD] immediates.
9830 (movti): Split CONST_POLY_INT moves into two halves.
9831 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9832 Split additions that need a temporary here if the destination
9833 is the stack pointer.
9834 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9835 (*add<mode>3_poly_1): New instruction.
9836 (set_clobber_cc): New expander.
9838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9840 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9841 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9842 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9843 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9844 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9845 Change innermode from fixed_mode_size to machine_mode.
9846 (simplify_subreg): Update call accordingly. Handle a constant-sized
9847 subreg of a variable-length CONST_VECTOR.
9849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9850 Alan Hayward <alan.hayward@arm.com>
9851 David Sherwood <david.sherwood@arm.com>
9853 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9854 (add_offset_to_base): New function, split out from...
9855 (create_mem_ref): ...here. When handling a scale other than 1,
9856 check first whether the address is valid without the offset.
9857 Add it into the base if so, leaving the index and scale as-is.
9859 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9862 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9863 fold_for_warn before checking if arg2 is INTEGER_CST.
9865 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9867 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9868 (store_multiple_operation): Delete.
9869 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9870 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9871 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9872 guarded by TARGET_STRING.
9873 (rs6000_output_load_multiple): Delete.
9874 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9875 OPTION_MASK_STRING / TARGET_STRING handling.
9876 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9877 (const rs6000_opt_masks) <"string">: Change mask to 0.
9878 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9879 (MASK_STRING): Delete.
9880 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9882 (load_multiple): Delete.
9889 (store_multiple): Delete.
9896 (movmemsi_8reg): Delete.
9897 (corresponding unnamed define_insn): Delete.
9898 (movmemsi_6reg): Delete.
9899 (corresponding unnamed define_insn): Delete.
9900 (movmemsi_4reg): Delete.
9901 (corresponding unnamed define_insn): Delete.
9902 (movmemsi_2reg): Delete.
9903 (corresponding unnamed define_insn): Delete.
9904 (movmemsi_1reg): Delete.
9905 (corresponding unnamed define_insn): Delete.
9906 * config/rs6000/rs6000.opt (mno-string): New.
9907 (mstring): Replace by deprecation warning stub.
9908 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9910 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9912 * regrename.c (regrename_do_replace): If replacing the same
9913 reg multiple times, try to reuse last created gen_raw_REG.
9916 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9917 main to workaround a bug in GDB.
9919 2018-01-12 Tom de Vries <tom@codesourcery.com>
9922 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9924 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9926 PR rtl-optimization/80481
9927 * ira-color.c (get_cap_member): New function.
9928 (allocnos_conflict_by_live_ranges_p): Use it.
9929 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9930 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9932 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9935 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9936 (*saddl_se_1): Ditto.
9938 (*ssubl_se_1): Ditto.
9940 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9942 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9943 rather than wi::to_widest for DR_INITs.
9944 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9945 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9946 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9948 (vect_analyze_group_access_1): Note that here.
9950 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9952 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9953 polynomial type sizes.
9955 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9957 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9958 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9959 (gimple_add_tmp_var): Likewise.
9961 2018-01-12 Martin Liska <mliska@suse.cz>
9963 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9964 (gimple_alloc_sizes): Likewise.
9965 (dump_gimple_statistics): Use PRIu64 in printf format.
9966 * gimple.h: Change uint64_t to int.
9968 2018-01-12 Martin Liska <mliska@suse.cz>
9970 * tree-core.h: Use uint64_t instead of int.
9971 * tree.c (tree_node_counts): Likewise.
9972 (tree_node_sizes): Likewise.
9973 (dump_tree_statistics): Use PRIu64 in printf format.
9975 2018-01-12 Martin Liska <mliska@suse.cz>
9977 * Makefile.in: As qsort_chk is implemented in vec.c, add
9978 vec.o to linkage of gencfn-macros.
9979 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9980 passing the info to record_node_allocation_statistics.
9981 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9983 * ggc-common.c (struct ggc_usage): Add operator== and use
9984 it in operator< and compare function.
9985 * mem-stats.h (struct mem_usage): Likewise.
9986 * vec.c (struct vec_usage): Remove operator< and compare
9987 function. Can be simply inherited.
9989 2018-01-12 Martin Jambor <mjambor@suse.cz>
9992 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9993 * tree-ssa-math-opts.c: Include domwalk.h.
9994 (convert_mult_to_fma_1): New function.
9995 (fma_transformation_info): New type.
9996 (fma_deferring_state): Likewise.
9997 (cancel_fma_deferring): New function.
9998 (result_of_phi): Likewise.
9999 (last_fma_candidate_feeds_initial_phi): Likewise.
10000 (convert_mult_to_fma): Added deferring logic, split actual
10001 transformation to convert_mult_to_fma_1.
10002 (math_opts_dom_walker): New type.
10003 (math_opts_dom_walker::after_dom_children): New method, body moved
10004 here from pass_optimize_widening_mul::execute, added deferring logic
10006 (pass_optimize_widening_mul::execute): Moved most of code to
10007 math_opts_dom_walker::after_dom_children.
10008 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10009 * config/i386/i386.c (ix86_option_override_internal): Added
10010 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10012 2018-01-12 Richard Biener <rguenther@suse.de>
10015 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10016 inline instance vars.
10018 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10021 * config/rx/rx.c (rx_is_restricted_memory_address):
10022 Handle SUBREG case.
10024 2018-01-12 Richard Biener <rguenther@suse.de>
10026 PR tree-optimization/80846
10027 * target.def (split_reduction): New target hook.
10028 * targhooks.c (default_split_reduction): New function.
10029 * targhooks.h (default_split_reduction): Declare.
10030 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10031 target requests first reduce vectors by combining low and high
10033 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10034 (get_vectype_for_scalar_type_and_size): Export.
10035 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10036 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10037 * doc/tm.texi: Regenerate.
10038 * config/i386/i386.c (ix86_split_reduction): Implement
10039 TARGET_VECTORIZE_SPLIT_REDUCTION.
10041 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10044 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10045 in PIC mode except for TARGET_VXWORKS_RTP.
10046 * config/sparc/sparc.c: Include cfgrtl.h.
10047 (TARGET_INIT_PIC_REG): Define.
10048 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10049 (sparc_pic_register_p): New predicate.
10050 (sparc_legitimate_address_p): Use it.
10051 (sparc_legitimize_pic_address): Likewise.
10052 (sparc_delegitimize_address): Likewise.
10053 (sparc_mode_dependent_address_p): Likewise.
10054 (gen_load_pcrel_sym): Remove 4th parameter.
10055 (load_got_register): Adjust call to above. Remove obsolete stuff.
10056 (sparc_expand_prologue): Do not call load_got_register here.
10057 (sparc_flat_expand_prologue): Likewise.
10058 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10059 (sparc_use_pseudo_pic_reg): New function.
10060 (sparc_init_pic_reg): Likewise.
10061 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10062 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10064 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10066 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10067 Add item for branch_cost.
10069 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10071 PR rtl-optimization/83565
10072 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10073 not extend the result to a larger mode for rotate operations.
10074 (num_sign_bit_copies1): Likewise.
10076 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10079 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10081 Use values-Xc.o for -pedantic.
10082 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10084 2018-01-12 Martin Liska <mliska@suse.cz>
10087 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10089 (possible_polymorphic_call_targets): Use it.
10090 (ipa_devirt): Likewise.
10092 2018-01-12 Martin Liska <mliska@suse.cz>
10094 * profile-count.h (enum profile_quality): Use 0 as invalid
10095 enum value of profile_quality.
10097 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10099 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10100 -mext-string options.
10102 2018-01-12 Richard Biener <rguenther@suse.de>
10104 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10105 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10106 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10108 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10110 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10112 * configure.ac (--with-long-double-format): Add support for the
10113 configuration option to change the default long double format on
10115 * config.gcc (powerpc*-linux*-*): Likewise.
10116 * configure: Regenerate.
10117 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10118 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10119 used without modification.
10121 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10123 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10124 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10125 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10126 MISC_BUILTIN_SPEC_BARRIER.
10127 (rs6000_init_builtins): Likewise.
10128 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10130 (speculation_barrier): New define_insn.
10131 * doc/extend.texi: Document __builtin_speculation_barrier.
10133 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10136 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10137 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10138 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10140 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10141 integral modes instead of "ss" and "sd".
10142 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10143 vectors with 32-bit and 64-bit elements.
10144 (vecdupssescalarmodesuffix): New mode attribute.
10145 (vec_dup<mode>): Use it.
10147 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10150 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10151 frame if argument is passed on stack.
10153 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10156 * ree.c (combine_reaching_defs): Optimize also
10157 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10158 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10160 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10162 PR middle-end/83189
10163 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10165 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10167 PR middle-end/83718
10168 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10169 after they are computed.
10171 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10173 PR tree-optimization/83695
10174 * gimple-loop-linterchange.cc
10175 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10176 reset cached scev information after interchange.
10177 (pass_linterchange::execute): Remove call to scev_reset_htab.
10179 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10181 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10182 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10183 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10184 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10185 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10186 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10187 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10188 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10189 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10190 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10191 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10192 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10193 (V_lane_reg): Likewise.
10194 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10196 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10197 (vfmal_lane_low<mode>_intrinsic,
10198 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10199 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10200 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10201 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10202 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10203 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10205 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10207 * config/arm/arm-cpus.in (fp16fml): New feature.
10208 (ALL_SIMD): Add fp16fml.
10209 (armv8.2-a): Add fp16fml as an option.
10210 (armv8.3-a): Likewise.
10211 (armv8.4-a): Add fp16fml as part of fp16.
10212 * config/arm/arm.h (TARGET_FP16FML): Define.
10213 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10215 * config/arm/arm-modes.def (V2HF): Define.
10216 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10217 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10218 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10219 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10220 vfmsl_low, vfmsl_high): New set of builtins.
10221 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10222 (vfml_op): New code attribute.
10223 (VFMLHALVES): New int iterator.
10224 (VFML, VFMLSEL): New mode attributes.
10225 (V_reg): Define mapping for V2HF.
10226 (V_hi, V_lo): New mode attributes.
10227 (VF_constraint): Likewise.
10228 (vfml_half, vfml_half_selector): New int attributes.
10229 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10231 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10232 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10234 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10235 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10236 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10237 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10239 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10240 Document new effective target and option set.
10242 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10244 * config/arm/arm-cpus.in (armv8_4): New feature.
10245 (ARMv8_4a): New fgroup.
10246 (armv8.4-a): New arch.
10247 * config/arm/arm-tables.opt: Regenerate.
10248 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10249 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10250 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10251 Add matching rules for -march=armv8.4-a and extensions.
10252 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10254 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10257 * config/rx/rx.md (BW): New mode attribute.
10258 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10260 2018-01-11 Richard Biener <rguenther@suse.de>
10262 PR tree-optimization/83435
10263 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10264 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10265 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10267 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10268 Alan Hayward <alan.hayward@arm.com>
10269 David Sherwood <david.sherwood@arm.com>
10271 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10273 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10274 (aarch64_print_address_internal): Use it to check for a zero offset.
10276 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10277 Alan Hayward <alan.hayward@arm.com>
10278 David Sherwood <david.sherwood@arm.com>
10280 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10281 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10282 Return a poly_int64 rather than a HOST_WIDE_INT.
10283 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10284 rather than a HOST_WIDE_INT.
10285 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10286 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10287 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10288 final_offset from HOST_WIDE_INT to poly_int64.
10289 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10290 to_constant when getting the number of units in an Advanced SIMD
10292 (aarch64_builtin_vectorized_function): Check for a constant number
10294 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10296 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10297 attribute instead of GET_MODE_NUNITS.
10298 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10299 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10300 GET_MODE_SIZE for fixed-size registers.
10301 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10302 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10303 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10304 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10305 (aarch64_print_operand, aarch64_print_address_internal)
10306 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10307 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10308 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10309 Handle polynomial GET_MODE_SIZE.
10310 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10311 wider than SImode without modification.
10312 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10313 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10314 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10315 passing and returning SVE modes.
10316 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10317 rather than GEN_INT.
10318 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10319 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10320 (aarch64_allocate_and_probe_stack_space): Likewise.
10321 (aarch64_layout_frame): Cope with polynomial offsets.
10322 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10323 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10324 polynomial offsets.
10325 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10326 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10327 poly_int64 rather than a HOST_WIDE_INT.
10328 (aarch64_get_separate_components, aarch64_process_components)
10329 (aarch64_expand_prologue, aarch64_expand_epilogue)
10330 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10331 (aarch64_anchor_offset): New function, split out from...
10332 (aarch64_legitimize_address): ...here.
10333 (aarch64_builtin_vectorization_cost): Handle polynomial
10334 TYPE_VECTOR_SUBPARTS.
10335 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10337 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10338 number of elements from the PARALLEL rather than the mode.
10339 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10340 rather than GET_MODE_BITSIZE.
10341 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10342 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10343 (aarch64_expand_vec_perm_const_1): Handle polynomial
10344 d->perm.length () and d->perm elements.
10345 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10346 Apply to_constant to d->perm elements.
10347 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10348 polynomial CONST_VECTOR_NUNITS.
10349 (aarch64_move_pointer): Take amount as a poly_int64 rather
10351 (aarch64_progress_pointer): Avoid temporary variable.
10352 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10353 the mode attribute instead of GET_MODE.
10355 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10356 Alan Hayward <alan.hayward@arm.com>
10357 David Sherwood <david.sherwood@arm.com>
10359 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10360 x exists before using it.
10361 (aarch64_add_constant_internal): Rename to...
10362 (aarch64_add_offset_1): ...this. Replace regnum with separate
10363 src and dest rtxes. Handle the case in which they're different,
10364 including when the offset is zero. Replace scratchreg with an rtx.
10365 Use 2 additions if there is no spare register into which we can
10366 move a 16-bit constant.
10367 (aarch64_add_constant): Delete.
10368 (aarch64_add_offset): Replace reg with separate src and dest
10369 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10370 Use aarch64_add_offset_1.
10371 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10372 an rtx rather than an int. Take the delta as a poly_int64
10373 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10374 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10375 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10376 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10377 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10378 and aarch64_add_sp.
10379 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10380 aarch64_add_constant.
10382 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10384 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10385 Use scalar_float_mode.
10387 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10389 * config/aarch64/aarch64-simd.md
10390 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10391 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10392 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10393 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10394 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10395 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10396 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10397 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10398 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10399 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10401 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10404 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10405 targ_options->x_arm_arch_string is non NULL.
10407 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10409 * config/aarch64/aarch64.h
10410 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10412 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10415 * expmed.c (emit_store_flag_force): Swap if const op0
10416 and change VOIDmode to mode of op0.
10418 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10420 PR rtl-optimization/83761
10421 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10422 than bytes to mode_for_size.
10424 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10426 PR middle-end/83189
10427 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10428 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10431 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10433 PR middle-end/83575
10434 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10435 when in layout mode.
10436 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10437 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10440 2018-01-10 Michael Collison <michael.collison@arm.com>
10442 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10443 * config/aarch64/aarch64-option-extension.def: Add
10444 AARCH64_OPT_EXTENSION of 'fp16fml'.
10445 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10446 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10447 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10448 * config/aarch64/constraints.md (Ui7): New constraint.
10449 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10450 (VFMLA_SEL_W): Ditto.
10453 (VFMLA16_LOW): New int iterator.
10454 (VFMLA16_HIGH): Ditto.
10455 (UNSPEC_FMLAL): New unspec.
10456 (UNSPEC_FMLSL): Ditto.
10457 (UNSPEC_FMLAL2): Ditto.
10458 (UNSPEC_FMLSL2): Ditto.
10459 (f16mac): New code attribute.
10460 * config/aarch64/aarch64-simd-builtins.def
10461 (aarch64_fmlal_lowv2sf): Ditto.
10462 (aarch64_fmlsl_lowv2sf): Ditto.
10463 (aarch64_fmlalq_lowv4sf): Ditto.
10464 (aarch64_fmlslq_lowv4sf): Ditto.
10465 (aarch64_fmlal_highv2sf): Ditto.
10466 (aarch64_fmlsl_highv2sf): Ditto.
10467 (aarch64_fmlalq_highv4sf): Ditto.
10468 (aarch64_fmlslq_highv4sf): Ditto.
10469 (aarch64_fmlal_lane_lowv2sf): Ditto.
10470 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10471 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10472 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10473 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10474 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10475 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10476 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10477 (aarch64_fmlal_lane_highv2sf): Ditto.
10478 (aarch64_fmlsl_lane_highv2sf): Ditto.
10479 (aarch64_fmlal_laneq_highv2sf): Ditto.
10480 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10481 (aarch64_fmlalq_lane_highv4sf): Ditto.
10482 (aarch64_fmlsl_lane_highv4sf): Ditto.
10483 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10484 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10485 * config/aarch64/aarch64-simd.md:
10486 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10487 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10488 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10489 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10490 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10491 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10492 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10493 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10494 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10495 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10496 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10497 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10498 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10499 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10500 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10501 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10502 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10503 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10504 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10505 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10506 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10507 (vfmlsl_low_u32): Ditto.
10508 (vfmlalq_low_u32): Ditto.
10509 (vfmlslq_low_u32): Ditto.
10510 (vfmlal_high_u32): Ditto.
10511 (vfmlsl_high_u32): Ditto.
10512 (vfmlalq_high_u32): Ditto.
10513 (vfmlslq_high_u32): Ditto.
10514 (vfmlal_lane_low_u32): Ditto.
10515 (vfmlsl_lane_low_u32): Ditto.
10516 (vfmlal_laneq_low_u32): Ditto.
10517 (vfmlsl_laneq_low_u32): Ditto.
10518 (vfmlalq_lane_low_u32): Ditto.
10519 (vfmlslq_lane_low_u32): Ditto.
10520 (vfmlalq_laneq_low_u32): Ditto.
10521 (vfmlslq_laneq_low_u32): Ditto.
10522 (vfmlal_lane_high_u32): Ditto.
10523 (vfmlsl_lane_high_u32): Ditto.
10524 (vfmlal_laneq_high_u32): Ditto.
10525 (vfmlsl_laneq_high_u32): Ditto.
10526 (vfmlalq_lane_high_u32): Ditto.
10527 (vfmlslq_lane_high_u32): Ditto.
10528 (vfmlalq_laneq_high_u32): Ditto.
10529 (vfmlslq_laneq_high_u32): Ditto.
10530 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10531 (AARCH64_FL_FOR_ARCH8_4): New.
10532 (AARCH64_ISA_F16FML): New ISA flag.
10533 (TARGET_F16FML): New feature flag for fp16fml.
10534 (doc/invoke.texi): Document new fp16fml option.
10536 2018-01-10 Michael Collison <michael.collison@arm.com>
10538 * config/aarch64/aarch64-builtins.c:
10539 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10540 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10541 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10542 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10543 (AARCH64_ISA_SHA3): New ISA flag.
10544 (TARGET_SHA3): New feature flag for sha3.
10545 * config/aarch64/iterators.md (sha512_op): New int attribute.
10546 (CRYPTO_SHA512): New int iterator.
10547 (UNSPEC_SHA512H): New unspec.
10548 (UNSPEC_SHA512H2): Ditto.
10549 (UNSPEC_SHA512SU0): Ditto.
10550 (UNSPEC_SHA512SU1): Ditto.
10551 * config/aarch64/aarch64-simd-builtins.def
10552 (aarch64_crypto_sha512hqv2di): New builtin.
10553 (aarch64_crypto_sha512h2qv2di): Ditto.
10554 (aarch64_crypto_sha512su0qv2di): Ditto.
10555 (aarch64_crypto_sha512su1qv2di): Ditto.
10556 (aarch64_eor3qv8hi): Ditto.
10557 (aarch64_rax1qv2di): Ditto.
10558 (aarch64_xarqv2di): Ditto.
10559 (aarch64_bcaxqv8hi): Ditto.
10560 * config/aarch64/aarch64-simd.md:
10561 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10562 (aarch64_crypto_sha512su0qv2di): Ditto.
10563 (aarch64_crypto_sha512su1qv2di): Ditto.
10564 (aarch64_eor3qv8hi): Ditto.
10565 (aarch64_rax1qv2di): Ditto.
10566 (aarch64_xarqv2di): Ditto.
10567 (aarch64_bcaxqv8hi): Ditto.
10568 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10569 (vsha512h2q_u64): Ditto.
10570 (vsha512su0q_u64): Ditto.
10571 (vsha512su1q_u64): Ditto.
10572 (veor3q_u16): Ditto.
10573 (vrax1q_u64): Ditto.
10574 (vxarq_u64): Ditto.
10575 (vbcaxq_u16): Ditto.
10576 * config/arm/types.md (crypto_sha512): New type attribute.
10577 (crypto_sha3): Ditto.
10578 (doc/invoke.texi): Document new sha3 option.
10580 2018-01-10 Michael Collison <michael.collison@arm.com>
10582 * config/aarch64/aarch64-builtins.c:
10583 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10584 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10585 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10586 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10587 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10588 (AARCH64_ISA_SM4): New ISA flag.
10589 (TARGET_SM4): New feature flag for sm4.
10590 * config/aarch64/aarch64-simd-builtins.def
10591 (aarch64_sm3ss1qv4si): Ditto.
10592 (aarch64_sm3tt1aq4si): Ditto.
10593 (aarch64_sm3tt1bq4si): Ditto.
10594 (aarch64_sm3tt2aq4si): Ditto.
10595 (aarch64_sm3tt2bq4si): Ditto.
10596 (aarch64_sm3partw1qv4si): Ditto.
10597 (aarch64_sm3partw2qv4si): Ditto.
10598 (aarch64_sm4eqv4si): Ditto.
10599 (aarch64_sm4ekeyqv4si): Ditto.
10600 * config/aarch64/aarch64-simd.md:
10601 (aarch64_sm3ss1qv4si): Ditto.
10602 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10603 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10604 (aarch64_sm4eqv4si): Ditto.
10605 (aarch64_sm4ekeyqv4si): Ditto.
10606 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10607 (sm3part_op): Ditto.
10608 (CRYPTO_SM3TT): Ditto.
10609 (CRYPTO_SM3PART): Ditto.
10610 (UNSPEC_SM3SS1): New unspec.
10611 (UNSPEC_SM3TT1A): Ditto.
10612 (UNSPEC_SM3TT1B): Ditto.
10613 (UNSPEC_SM3TT2A): Ditto.
10614 (UNSPEC_SM3TT2B): Ditto.
10615 (UNSPEC_SM3PARTW1): Ditto.
10616 (UNSPEC_SM3PARTW2): Ditto.
10617 (UNSPEC_SM4E): Ditto.
10618 (UNSPEC_SM4EKEY): Ditto.
10619 * config/aarch64/constraints.md (Ui2): New constraint.
10620 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10621 * config/arm/types.md (crypto_sm3): New type attribute.
10622 (crypto_sm4): Ditto.
10623 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10624 (vsm3tt1aq_u32): Ditto.
10625 (vsm3tt1bq_u32): Ditto.
10626 (vsm3tt2aq_u32): Ditto.
10627 (vsm3tt2bq_u32): Ditto.
10628 (vsm3partw1q_u32): Ditto.
10629 (vsm3partw2q_u32): Ditto.
10630 (vsm4eq_u32): Ditto.
10631 (vsm4ekeyq_u32): Ditto.
10632 (doc/invoke.texi): Document new sm4 option.
10634 2018-01-10 Michael Collison <michael.collison@arm.com>
10636 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10637 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10638 (AARCH64_FL_FOR_ARCH8_4): New.
10639 (AARCH64_FL_V8_4): New flag.
10640 (doc/invoke.texi): Document new armv8.4-a option.
10642 2018-01-10 Michael Collison <michael.collison@arm.com>
10644 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10645 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10646 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10647 * config/aarch64/aarch64-option-extension.def: Add
10648 AARCH64_OPT_EXTENSION of 'sha2'.
10649 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10650 (crypto): Disable sha2 and aes if crypto disabled.
10651 (crypto): Enable aes and sha2 if enabled.
10652 (simd): Disable sha2 and aes if simd disabled.
10653 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10655 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10656 (TARGET_SHA2): New feature flag for sha2.
10657 (TARGET_AES): New feature flag for aes.
10658 * config/aarch64/aarch64-simd.md:
10659 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10660 conditional on TARGET_AES.
10661 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10662 (aarch64_crypto_sha1hsi): Make pattern conditional
10664 (aarch64_crypto_sha1hv4si): Ditto.
10665 (aarch64_be_crypto_sha1hv4si): Ditto.
10666 (aarch64_crypto_sha1su1v4si): Ditto.
10667 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10668 (aarch64_crypto_sha1su0v4si): Ditto.
10669 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10670 (aarch64_crypto_sha256su0v4si): Ditto.
10671 (aarch64_crypto_sha256su1v4si): Ditto.
10672 (doc/invoke.texi): Document new aes and sha2 options.
10674 2018-01-10 Martin Sebor <msebor@redhat.com>
10676 PR tree-optimization/83781
10677 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10680 2018-01-11 Martin Sebor <msebor@gmail.com>
10681 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10683 PR tree-optimization/83501
10684 PR tree-optimization/81703
10686 * tree-ssa-strlen.c (get_string_cst): Rename...
10687 (get_string_len): ...to this. Handle global constants.
10688 (handle_char_store): Adjust.
10690 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10691 Jim Wilson <jimw@sifive.com>
10693 * config/riscv/riscv-protos.h (riscv_output_return): New.
10694 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10695 (riscv_attribute_table, riscv_output_return),
10696 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10697 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10698 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10700 (riscv_expand_prologue): Add early return for naked function.
10701 (riscv_expand_epilogue): Likewise.
10702 (riscv_function_ok_for_sibcall): Return false for naked function.
10703 (riscv_set_current_function): New.
10704 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10705 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10706 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10707 * doc/extend.texi (RISC-V Function Attributes): New.
10709 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10711 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10712 check for 128-bit long double before checking TCmode.
10713 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10714 128-bit long doubles before checking TFmode or TCmode.
10715 (FLOAT128_IBM_P): Likewise.
10717 2018-01-10 Martin Sebor <msebor@redhat.com>
10719 PR tree-optimization/83671
10720 * builtins.c (c_strlen): Unconditionally return zero for the empty
10722 Use -Warray-bounds for warnings.
10723 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10724 for non-constant array indices with COMPONENT_REF, arrays of
10725 arrays, and pointers to arrays.
10726 (gimple_fold_builtin_strlen): Determine and set length range for
10727 non-constant character arrays.
10729 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10731 PR middle-end/81897
10732 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10735 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10737 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10739 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10742 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10743 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10744 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10745 indexed_or_indirect_operand predicate.
10746 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10747 (*vsx_le_perm_load_v8hi): Likewise.
10748 (*vsx_le_perm_load_v16qi): Likewise.
10749 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10750 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10751 (*vsx_le_perm_store_v8hi): Likewise.
10752 (*vsx_le_perm_store_v16qi): Likewise.
10753 (eight unnamed splitters): Likewise.
10755 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10757 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10758 * config/rs6000/emmintrin.h: Likewise.
10759 * config/rs6000/mmintrin.h: Likewise.
10760 * config/rs6000/xmmintrin.h: Likewise.
10762 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10765 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10767 * tree.c (tree_nop_conversion): Return true for location wrapper
10769 (maybe_wrap_with_location): New function.
10770 (selftest::check_strip_nops): New function.
10771 (selftest::test_location_wrappers): New function.
10772 (selftest::tree_c_tests): Call it.
10773 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10774 (maybe_wrap_with_location): New decl.
10775 (EXPR_LOCATION_WRAPPER_P): New macro.
10776 (location_wrapper_p): New inline function.
10777 (tree_strip_any_location_wrapper): New inline function.
10779 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10782 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10783 stack_realign_offset for the largest alignment of stack slot
10785 (ix86_find_max_used_stack_alignment): New function.
10786 (ix86_finalize_stack_frame_flags): Use it. Set
10787 max_used_stack_alignment if we don't realign stack.
10788 * config/i386/i386.h (machine_function): Add
10789 max_used_stack_alignment.
10791 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10793 * config/arm/arm.opt (-mbranch-cost): New option.
10794 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10797 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10800 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10801 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10803 2018-01-10 Richard Biener <rguenther@suse.de>
10806 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10807 early out so it also covers the case where we have a non-NULL
10810 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10812 PR tree-optimization/83753
10813 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10814 for non-strided grouped accesses if the number of elements is 1.
10816 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10819 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10820 * i386.h (TARGET_USE_GATHER): Define.
10821 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10823 2018-01-10 Martin Liska <mliska@suse.cz>
10826 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10827 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10829 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10830 CLEANUP_NO_PARTITIONING is not set.
10832 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10834 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10835 for vectors, as a partial revert of r254296.
10836 * rtl.h (const_vec_p): Delete.
10837 (const_vec_duplicate_p): Don't test for vector CONSTs.
10838 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10839 * expmed.c (make_tree): Likewise.
10842 * common.md (E, F): Use CONSTANT_P instead of checking for
10844 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10845 checking for CONST_VECTOR.
10847 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10849 PR middle-end/83575
10850 * predict.c (force_edge_cold): Handle in more sane way edges
10851 with no prediction.
10853 2018-01-09 Carl Love <cel@us.ibm.com>
10855 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10857 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10858 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10859 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10860 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10861 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10862 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10863 * config/rs6000/rs6000-protos.h: Add extern defition for
10864 rs6000_generate_float2_double_code.
10865 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10867 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10868 (float2_v2df): Add define_expand.
10870 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10873 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10874 op_mode in the force_to_mode call.
10876 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10878 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10879 instead of checking each element individually.
10880 (aarch64_evpc_uzp): Likewise.
10881 (aarch64_evpc_zip): Likewise.
10882 (aarch64_evpc_ext): Likewise.
10883 (aarch64_evpc_rev): Likewise.
10884 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10885 instead of checking each element individually. Return true without
10887 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10888 whether all selected elements come from the same input, instead of
10889 checking each element individually. Remove calls to gen_rtx_REG,
10890 start_sequence and end_sequence and instead assert that no rtl is
10893 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10895 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10896 order of HIGH and CONST checks.
10898 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10900 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10901 if the destination isn't an SSA_NAME.
10903 2018-01-09 Richard Biener <rguenther@suse.de>
10905 PR tree-optimization/83668
10906 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10908 (canonicalize_loop_form): ... here, renamed from ...
10909 (canonicalize_loop_closed_ssa_form): ... this and amended to
10910 swap successor edges for loop exit blocks to make us use
10911 the RPO order we need for initial schedule generation.
10913 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10915 PR tree-optimization/64811
10916 * match.pd: When optimizing comparisons with Inf, avoid
10917 introducing or losing exceptions from comparisons with NaN.
10919 2018-01-09 Martin Liska <mliska@suse.cz>
10922 * asan.c (shadow_mem_size): Add gcc_assert.
10924 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10926 Don't save registers in main().
10929 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10930 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10931 * config/avr/avr.c (avr_set_current_function): Don't error if
10932 naked, OS_task or OS_main are specified at the same time.
10933 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10935 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10937 * common/config/avr/avr-common.c (avr_option_optimization_table):
10938 Switch on -mmain-is-OS_task for optimizing compilations.
10940 2018-01-09 Richard Biener <rguenther@suse.de>
10942 PR tree-optimization/83572
10943 * graphite.c: Include cfganal.h.
10944 (graphite_transform_loops): Connect infinite loops to exit
10945 and remove fake edges at the end.
10947 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10949 * ipa-inline.c (edge_badness): Revert accidental checkin.
10951 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10954 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10955 symbols; not inline clones.
10957 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10960 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10961 hard registers. Formatting fixes.
10963 PR preprocessor/83722
10964 * gcc.c (try_generate_repro): Pass
10965 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10966 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10969 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10970 Kito Cheng <kito.cheng@gmail.com>
10972 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10973 (riscv_leaf_function_p): Delete.
10974 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10976 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10978 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10980 (do_ifelse): New function.
10981 (do_isel): New function.
10982 (do_sub3): New function.
10983 (do_add3): New function.
10984 (do_load_mask_compare): New function.
10985 (do_overlap_load_compare): New function.
10986 (expand_compare_loop): New function.
10987 (expand_block_compare): Call expand_compare_loop() when appropriate.
10988 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10989 option description.
10990 (-mblock-compare-inline-loop-limit): New option.
10992 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10995 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10996 Reverse order of second and third operands in first alternative.
10997 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10998 of first and second elements in UNSPEC_VPERMR vector.
10999 (altivec_expand_vec_perm_le): Likewise.
11001 2018-01-08 Jeff Law <law@redhat.com>
11003 PR rtl-optimizatin/81308
11004 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11005 (process_switch): If group_case_labels makes a change, then set
11007 (pass_convert_switch::execute): If a switch is converted, then
11008 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11010 PR rtl-optimization/81308
11011 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11014 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11016 PR target/83663 - Revert r255946
11017 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11018 generation for cases where splatting a value is not useful.
11019 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11020 across a vec_duplicate and a paradoxical subreg forming a vector
11021 mode to a vec_concat.
11023 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11025 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11026 -march=armv8.3-a variants.
11027 * config/arm/t-multilib: Likewise.
11028 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11030 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11032 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11034 (cceq_ior_compare_complement): Give it a name so I can use it, and
11035 change boolean_or_operator predicate to boolean_operator so it can
11036 be used to generate a crand.
11037 (eqne): New code iterator.
11038 (bd/bd_neg): New code_attrs.
11039 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11040 a single define_insn.
11041 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11042 decrement (bdnzt/bdnzf/bdzt/bdzf).
11043 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11044 with the new names of the branch decrement patterns, and added the
11045 names of the branch decrement conditional patterns.
11047 2018-01-08 Richard Biener <rguenther@suse.de>
11049 PR tree-optimization/83563
11050 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11053 2018-01-08 Richard Biener <rguenther@suse.de>
11055 PR middle-end/83713
11056 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11058 2018-01-08 Richard Biener <rguenther@suse.de>
11060 PR tree-optimization/83685
11061 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11062 references to abnormals.
11064 2018-01-08 Richard Biener <rguenther@suse.de>
11067 * dwarf2out.c (output_indirect_strings): Handle empty
11068 skeleton_debug_str_hash.
11069 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11071 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11073 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11074 (emit_store_direct): Likewise.
11075 (arc_trampoline_adjust_address): Likewise.
11076 (arc_asm_trampoline_template): New function.
11077 (arc_initialize_trampoline): Use asm_trampoline_template.
11078 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11079 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11080 * config/arc/arc.md (flush_icache): Delete pattern.
11082 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11084 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11085 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11088 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11091 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11092 by not USED_FOR_TARGET.
11093 (make_pass_resolve_sw_modes): Likewise.
11095 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11097 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11100 2018-01-08 Richard Biener <rguenther@suse.de>
11102 PR middle-end/83580
11103 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11105 2018-01-08 Richard Biener <rguenther@suse.de>
11107 PR middle-end/83517
11108 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11110 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11112 PR middle-end/81897
11113 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11114 basic blocks with a small number of successors.
11115 (convert_control_dep_chain_into_preds): Improve handling of
11117 (dump_predicates): Split apart into...
11118 (dump_pred_chain): ...here...
11119 (dump_pred_info): ...and here.
11120 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11121 (can_chain_union_be_invalidated_p): Improve check for invalidation
11123 (uninit_uses_cannot_happen): Avoid unnecessary if
11124 convert_control_dep_chain_into_preds yielded nothing.
11126 2018-01-06 Martin Sebor <msebor@redhat.com>
11128 PR tree-optimization/83640
11129 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11130 subtracting negative offset from size.
11131 (builtin_access::overlap): Adjust offset bounds of the access to fall
11132 within the size of the object if possible.
11134 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11136 PR rtl-optimization/83699
11137 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11138 extract_bit_field_as_subreg to cases in which the extracted
11139 value is also a vector.
11141 * lra-constraints.c (process_alt_operands): Test for the equivalence
11142 substitutions when detecting a possible reload cycle.
11144 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11147 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11148 by default if flag_selective_schedling{,2}. Formatting fixes.
11150 PR rtl-optimization/83682
11151 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11152 if it has non-VECTOR_MODE element mode.
11153 (vec_duplicate_p): Likewise.
11155 PR middle-end/83694
11156 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11157 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11159 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11162 * config/i386/i386-builtin.def
11163 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11164 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11165 Require also OPTION_MASK_ISA_AVX512F in addition to
11166 OPTION_MASK_ISA_GFNI.
11167 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11168 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11169 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11170 to OPTION_MASK_ISA_GFNI.
11171 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11172 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11173 OPTION_MASK_ISA_AVX512BW.
11174 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11175 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11176 addition to OPTION_MASK_ISA_GFNI.
11177 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11178 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11179 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11180 to OPTION_MASK_ISA_GFNI.
11181 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11182 a requirement for all ISAs rather than any of them with a few
11184 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11186 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11187 bitmasks to be enabled with 3 exceptions, instead of requiring any
11188 enabled ISA with lots of exceptions.
11189 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11190 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11191 Change avx512bw in isa attribute to avx512f.
11192 * config/i386/sgxintrin.h: Add license boilerplate.
11193 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11194 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11195 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11196 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11198 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11199 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11200 temporarily sse2 rather than sse if not enabled already.
11203 * config/i386/sse.md (VI248_VLBW): Rename to ...
11204 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11205 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11206 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11207 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11208 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11209 mode iterator instead of VI248_VLBW.
11211 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11213 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11214 (record_modified): Skip clobbers; add debug output.
11215 (param_change_prob): Use sreal frequencies.
11217 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11219 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11220 punt for user-aligned variables.
11222 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11224 * tree-chrec.c (chrec_contains_symbols): Return true for
11227 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11230 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11231 of (x|y) == x for BICS pattern.
11233 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11235 PR tree-optimization/83605
11236 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11237 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11240 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11242 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11243 * config/epiphany/rtems.h: New file.
11245 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11246 Uros Bizjak <ubizjak@gmail.com>
11249 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11250 QIreg_operand instead of register_operand predicate.
11251 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11252 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11253 comments instead of -fmitigate[-_]rop.
11255 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11258 * cgraphunit.c (symbol_table::compile): Switch to text_section
11259 before calling assembly_start debug hook.
11260 * run-rtl-passes.c (run_rtl_passes): Likewise.
11263 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11265 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11266 range_int_cst_p rather than !symbolic_range_p before calling
11267 extract_range_from_multiplicative_op_1.
11269 2018-01-04 Jeff Law <law@redhat.com>
11271 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11272 redundant test in assertion.
11274 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11276 * doc/rtl.texi: Document machine_mode wrapper classes.
11278 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11280 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11281 using tree_to_uhwi.
11283 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11285 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11286 the VEC_PERM_EXPR fold to fail.
11288 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11291 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11292 to switched_sections.
11294 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11297 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11298 test for d.testing.
11300 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11303 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11304 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11306 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11309 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11310 is BLKmode and bitpos not zero or mode change is needed.
11312 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11315 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11318 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11321 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11322 instead of MULT rtx. Update all corresponding splitters.
11323 (*saddl_se): Ditto.
11324 (*ssub<modesuffix>): Ditto.
11325 (*ssubl_se): Ditto.
11326 (*cmp_sadd_di): Update split patterns.
11327 (*cmp_sadd_si): Ditto.
11328 (*cmp_sadd_sidi): Ditto.
11329 (*cmp_ssub_di): Ditto.
11330 (*cmp_ssub_si): Ditto.
11331 (*cmp_ssub_sidi): Ditto.
11332 * config/alpha/predicates.md (const23_operand): New predicate.
11333 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11334 Look for ASHIFT, not MULT inner operand.
11335 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11337 2018-01-04 Martin Liska <mliska@suse.cz>
11339 PR gcov-profile/83669
11340 * gcov.c (output_intermediate_file): Add version to intermediate
11342 * doc/gcov.texi: Document new field 'version' in intermediate
11343 file format. Fix location of '-k' option of gcov command.
11345 2018-01-04 Martin Liska <mliska@suse.cz>
11348 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11350 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11352 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11354 2018-01-03 Martin Sebor <msebor@redhat.com>
11356 PR tree-optimization/83655
11357 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11358 checking calls with invalid arguments.
11360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11362 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11363 (vectorizable_mask_load_store): Delete.
11364 (vectorizable_call): Return false for masked loads and stores.
11365 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11366 instead of gimple_assign_rhs1.
11367 (vectorizable_load): Handle IFN_MASK_LOAD.
11368 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11372 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11374 (vectorizable_mask_load_store): ...here.
11375 (vectorizable_load): ...and here.
11377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11379 * tree-vect-stmts.c (vect_build_all_ones_mask)
11380 (vect_build_zero_merge_argument): New functions, split out from...
11381 (vectorizable_load): ...here.
11383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11385 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11387 (vectorizable_mask_load_store): ...here.
11388 (vectorizable_store): ...and here.
11390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11392 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11394 (vectorizable_mask_load_store): ...here.
11396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11398 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11399 (vect_model_store_cost): Take a vec_load_store_type instead of a
11401 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11402 (vect_model_store_cost): Take a vec_load_store_type instead of a
11404 (vectorizable_mask_load_store): Update accordingly.
11405 (vectorizable_store): Likewise.
11406 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11410 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11411 IFN_MASK_LOAD calls here rather than...
11412 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11415 Alan Hayward <alan.hayward@arm.com>
11416 David Sherwood <david.sherwood@arm.com>
11418 * expmed.c (extract_bit_field_1): For vector extracts,
11419 fall back to extract_bit_field_as_subreg if vec_extract
11422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11423 Alan Hayward <alan.hayward@arm.com>
11424 David Sherwood <david.sherwood@arm.com>
11426 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11427 they are variable or constant sized.
11428 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11429 slots for constant-sized data.
11431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11432 Alan Hayward <alan.hayward@arm.com>
11433 David Sherwood <david.sherwood@arm.com>
11435 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11436 handling COND_EXPRs with boolean comparisons, try to find a better
11437 basis for the mask type than the boolean itself.
11439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11441 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11442 is calculated and how it can be overridden.
11443 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11444 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11446 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11450 Alan Hayward <alan.hayward@arm.com>
11451 David Sherwood <david.sherwood@arm.com>
11453 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11454 Remove the mode argument.
11455 (aarch64_simd_valid_immediate): Remove the mode and inverse
11457 * config/aarch64/iterators.md (bitsize): New iterator.
11458 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11459 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11460 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11461 aarch64_simd_valid_immediate.
11462 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11463 (aarch64_reg_or_bic_imm): Likewise.
11464 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11465 with an insn_type enum and msl with a modifier_type enum.
11466 Replace element_width with a scalar_mode. Change the shift
11467 to unsigned int. Add constructors for scalar_float_mode and
11468 scalar_int_mode elements.
11469 (aarch64_vect_float_const_representable_p): Delete.
11470 (aarch64_can_const_movi_rtx_p)
11471 (aarch64_simd_scalar_immediate_valid_for_move)
11472 (aarch64_simd_make_constant): Update call to
11473 aarch64_simd_valid_immediate.
11474 (aarch64_advsimd_valid_immediate_hs): New function.
11475 (aarch64_advsimd_valid_immediate): Likewise.
11476 (aarch64_simd_valid_immediate): Remove mode and inverse
11477 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11478 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11479 and aarch64_float_const_representable_p on the result.
11480 (aarch64_output_simd_mov_immediate): Remove mode argument.
11481 Update call to aarch64_simd_valid_immediate and use of
11482 simd_immediate_info.
11483 (aarch64_output_scalar_simd_mov_immediate): Update call
11486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11487 Alan Hayward <alan.hayward@arm.com>
11488 David Sherwood <david.sherwood@arm.com>
11490 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11491 (mode_nunits): Likewise CONST_MODE_NUNITS.
11492 * machmode.def (ADJUST_NUNITS): Document.
11493 * genmodes.c (mode_data::need_nunits_adj): New field.
11494 (blank_mode): Update accordingly.
11495 (adj_nunits): New variable.
11496 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11498 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11499 listed in adj_nunits.
11500 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11501 listed in adj_nunits. Don't emit case statements for such modes.
11502 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11503 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11504 nothing if adj_nunits is nonnull.
11505 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11506 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11507 (emit_mode_fbit): Update use of print_maybe_const_decl.
11508 (emit_move_size): Likewise. Treat the array as non-const
11510 (emit_mode_adjustments): Handle adj_nunits.
11512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11514 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11515 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11516 (VECTOR_MODES): Use it.
11517 (make_vector_modes): Take the prefix as an argument.
11519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11520 Alan Hayward <alan.hayward@arm.com>
11521 David Sherwood <david.sherwood@arm.com>
11523 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11524 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11525 for MODE_VECTOR_BOOL.
11526 * machmode.def (VECTOR_BOOL_MODE): Document.
11527 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11528 (make_vector_bool_mode): New function.
11529 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11531 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11532 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11534 * stor-layout.c (int_mode_for_mode): Likewise.
11535 * tree.c (build_vector_type_for_mode): Likewise.
11536 * varasm.c (output_constant_pool_2): Likewise.
11537 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11538 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11539 for MODE_VECTOR_BOOL.
11540 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11541 of mode class checks.
11542 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11543 instead of a list of mode class checks.
11544 (expand_vector_scalar_condition): Likewise.
11545 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11548 Alan Hayward <alan.hayward@arm.com>
11549 David Sherwood <david.sherwood@arm.com>
11551 * machmode.h (mode_size): Change from unsigned short to
11553 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11554 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11555 or if measurement_type is not polynomial.
11556 (fixed_size_mode::includes_p): Check for constant-sized modes.
11557 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11558 return a poly_uint16 rather than an unsigned short.
11559 (emit_mode_size): Change the type of mode_size from unsigned short
11560 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11561 (emit_mode_adjustments): Cope with polynomial vector sizes.
11562 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11564 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11566 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11567 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11568 * caller-save.c (setup_save_areas): Likewise.
11569 (replace_reg_with_saved_mem): Likewise.
11570 * calls.c (emit_library_call_value_1): Likewise.
11571 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11572 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11573 (gen_lowpart_for_combine): Likewise.
11574 * convert.c (convert_to_integer_1): Likewise.
11575 * cse.c (equiv_constant, cse_insn): Likewise.
11576 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11577 (cselib_subst_to_values): Likewise.
11578 * dce.c (word_dce_process_block): Likewise.
11579 * df-problems.c (df_word_lr_mark_ref): Likewise.
11580 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11581 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11582 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11583 (rtl_for_decl_location): Likewise.
11584 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11585 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11586 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11587 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11588 (expand_expr_real_1): Likewise.
11589 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11590 (pad_below): Likewise.
11591 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11592 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11593 * ira.c (get_subreg_tracking_sizes): Likewise.
11594 * ira-build.c (ira_create_allocno_objects): Likewise.
11595 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11596 (ira_sort_regnos_for_alter_reg): Likewise.
11597 * ira-costs.c (record_operand_costs): Likewise.
11598 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11599 (resolve_simple_move): Likewise.
11600 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11601 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11602 (lra_constraints): Likewise.
11603 (CONST_POOL_OK_P): Reject variable-sized modes.
11604 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11605 (add_pseudo_to_slot, lra_spill): Likewise.
11606 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11607 * optabs-query.c (get_best_extraction_insn): Likewise.
11608 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11609 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11610 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11611 * recog.c (offsettable_address_addr_space_p): Likewise.
11612 * regcprop.c (maybe_mode_change): Likewise.
11613 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11614 * regrename.c (build_def_use): Likewise.
11615 * regstat.c (dump_reg_info): Likewise.
11616 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11617 (find_reloads, find_reloads_subreg_address): Likewise.
11618 * reload1.c (eliminate_regs_1): Likewise.
11619 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11620 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11621 (simplify_binary_operation_1, simplify_subreg): Likewise.
11622 * targhooks.c (default_function_arg_padding): Likewise.
11623 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11624 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11625 (verify_gimple_assign_ternary): Likewise.
11626 * tree-inline.c (estimate_move_cost): Likewise.
11627 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11628 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11629 (get_address_cost_ainc): Likewise.
11630 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11631 (vect_supportable_dr_alignment): Likewise.
11632 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11633 (vectorizable_reduction): Likewise.
11634 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11635 (vectorizable_operation, vectorizable_load): Likewise.
11636 * tree.c (build_same_sized_truth_vector_type): Likewise.
11637 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11638 * var-tracking.c (emit_note_insn_var_location): Likewise.
11639 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11640 (ADDR_VEC_ALIGN): Likewise.
11642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11643 Alan Hayward <alan.hayward@arm.com>
11644 David Sherwood <david.sherwood@arm.com>
11646 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11648 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11649 or if measurement_type is polynomial.
11650 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11651 * combine.c (make_extraction): Likewise.
11652 * dse.c (find_shift_sequence): Likewise.
11653 * dwarf2out.c (mem_loc_descriptor): Likewise.
11654 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11655 (extract_bit_field, extract_low_bits): Likewise.
11656 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11657 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11658 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11659 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11660 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11661 * reload.c (find_reloads): Likewise.
11662 * reload1.c (alter_reg): Likewise.
11663 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11664 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11665 * tree-if-conv.c (predicate_mem_writes): Likewise.
11666 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11667 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11668 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11669 * valtrack.c (dead_debug_insert_temp): Likewise.
11670 * varasm.c (mergeable_constant_section): Likewise.
11671 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11674 Alan Hayward <alan.hayward@arm.com>
11675 David Sherwood <david.sherwood@arm.com>
11677 * expr.c (expand_assignment): Cope with polynomial mode sizes
11678 when assigning to a CONCAT.
11680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11681 Alan Hayward <alan.hayward@arm.com>
11682 David Sherwood <david.sherwood@arm.com>
11684 * machmode.h (mode_precision): Change from unsigned short to
11686 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11688 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11689 or if measurement_type is not polynomial.
11690 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11691 in which the mode is already known to be a scalar_int_mode.
11692 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11693 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11695 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11696 for GET_MODE_PRECISION.
11697 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11698 for GET_MODE_PRECISION.
11699 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11701 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11702 (expand_field_assignment, make_extraction): Likewise.
11703 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11704 (get_last_value): Likewise.
11705 * convert.c (convert_to_integer_1): Likewise.
11706 * cse.c (cse_insn): Likewise.
11707 * expr.c (expand_expr_real_1): Likewise.
11708 * lra-constraints.c (simplify_operand_subreg): Likewise.
11709 * optabs-query.c (can_atomic_load_p): Likewise.
11710 * optabs.c (expand_atomic_load): Likewise.
11711 (expand_atomic_store): Likewise.
11712 * ree.c (combine_reaching_defs): Likewise.
11713 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11714 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11715 * tree.h (type_has_mode_precision_p): Likewise.
11716 * ubsan.c (instrument_si_overflow): Likewise.
11718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11719 Alan Hayward <alan.hayward@arm.com>
11720 David Sherwood <david.sherwood@arm.com>
11722 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11723 polynomial numbers of units.
11724 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11725 (valid_vector_subparts_p): New function.
11726 (build_vector_type): Remove temporary shim and take the number
11727 of units as a poly_uint64 rather than an int.
11728 (build_opaque_vector_type): Take the number of units as a
11729 poly_uint64 rather than an int.
11730 * tree.c (build_vector_from_ctor): Handle polynomial
11731 TYPE_VECTOR_SUBPARTS.
11732 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11733 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11734 (build_vector_from_val): If the number of units is variable,
11735 use build_vec_duplicate_cst for constant operands and
11736 VEC_DUPLICATE_EXPR otherwise.
11737 (make_vector_type): Remove temporary is_constant ().
11738 (build_vector_type, build_opaque_vector_type): Take the number of
11739 units as a poly_uint64 rather than an int.
11740 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11742 * cfgexpand.c (expand_debug_expr): Likewise.
11743 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11744 (store_constructor, expand_expr_real_1): Likewise.
11745 (const_scalar_mask_from_tree): Likewise.
11746 * fold-const-call.c (fold_const_reduction): Likewise.
11747 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11748 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11749 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11750 (fold_relational_const): Likewise.
11751 (native_interpret_vector): Likewise. Change the size from an
11752 int to an unsigned int.
11753 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11754 TYPE_VECTOR_SUBPARTS.
11755 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11756 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11757 duplicating a non-constant operand into a variable-length vector.
11758 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11759 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11760 * ipa-icf.c (sem_variable::equals): Likewise.
11761 * match.pd: Likewise.
11762 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11763 * print-tree.c (print_node): Likewise.
11764 * stor-layout.c (layout_type): Likewise.
11765 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11766 * tree-cfg.c (verify_gimple_comparison): Likewise.
11767 (verify_gimple_assign_binary): Likewise.
11768 (verify_gimple_assign_ternary): Likewise.
11769 (verify_gimple_assign_single): Likewise.
11770 * tree-pretty-print.c (dump_generic_node): Likewise.
11771 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11772 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11773 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11774 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11775 (vect_shift_permute_load_chain): Likewise.
11776 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11777 (expand_vector_condition, optimize_vector_constructor): Likewise.
11778 (lower_vec_perm, get_compute_type): Likewise.
11779 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11780 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11781 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11782 (vect_recog_mask_conversion_pattern): Likewise.
11783 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11784 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11785 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11786 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11787 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11788 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11789 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11790 (supportable_widening_operation): Likewise.
11791 (supportable_narrowing_operation): Likewise.
11792 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11794 * varasm.c (output_constant): Likewise.
11796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11797 Alan Hayward <alan.hayward@arm.com>
11798 David Sherwood <david.sherwood@arm.com>
11800 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11801 so that both the length == 3 and length != 3 cases set up their
11802 own permute vectors. Add comments explaining why we know the
11803 number of elements is constant.
11804 (vect_permute_load_chain): Likewise.
11806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11807 Alan Hayward <alan.hayward@arm.com>
11808 David Sherwood <david.sherwood@arm.com>
11810 * machmode.h (mode_nunits): Change from unsigned char to
11812 (ONLY_FIXED_SIZE_MODES): New macro.
11813 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11814 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11815 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11817 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11818 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11819 or if measurement_type is not polynomial.
11820 * genmodes.c (ZERO_COEFFS): New macro.
11821 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11823 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11824 Use ZERO_COEFFS when emitting initializers.
11825 * data-streamer.h (bp_pack_poly_value): New function.
11826 (bp_unpack_poly_value): Likewise.
11827 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11828 for GET_MODE_NUNITS.
11829 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11830 for GET_MODE_NUNITS.
11831 * tree.c (make_vector_type): Remove temporary shim and make
11832 the real function take the number of units as a poly_uint64
11833 rather than an int.
11834 (build_vector_type_for_mode): Handle polynomial nunits.
11835 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11836 * emit-rtl.c (const_vec_series_p_1): Likewise.
11837 (gen_rtx_CONST_VECTOR): Likewise.
11838 * fold-const.c (test_vec_duplicate_folding): Likewise.
11839 * genrecog.c (validate_pattern): Likewise.
11840 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11841 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11842 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11843 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11844 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11845 * rtlanal.c (subreg_get_info): Likewise.
11846 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11847 (vect_grouped_load_supported): Likewise.
11848 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11849 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11850 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11851 (simplify_const_unary_operation, simplify_binary_operation_1)
11852 (simplify_const_binary_operation, simplify_ternary_operation)
11853 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11854 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11855 instead of CONST_VECTOR_NUNITS.
11856 * varasm.c (output_constant_pool_2): Likewise.
11857 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11858 explicit-encoded elements in the XVEC for variable-length vectors.
11860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11862 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11865 Alan Hayward <alan.hayward@arm.com>
11866 David Sherwood <david.sherwood@arm.com>
11868 * coretypes.h (fixed_size_mode): Declare.
11869 (fixed_size_mode_pod): New typedef.
11870 * builtins.h (target_builtins::x_apply_args_mode)
11871 (target_builtins::x_apply_result_mode): Change type to
11872 fixed_size_mode_pod.
11873 * builtins.c (apply_args_size, apply_result_size, result_vector)
11874 (expand_builtin_apply_args_1, expand_builtin_apply)
11875 (expand_builtin_return): Update accordingly.
11877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11879 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11880 * cselib.c (cselib_hash_rtx): Likewise.
11881 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11882 CONST_VECTOR encoding.
11884 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11885 Jeff Law <law@redhat.com>
11888 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11889 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11890 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11891 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11894 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11895 explicitly probe *sp in a noreturn function if there were any callee
11896 register saves or frame pointer is needed.
11898 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11901 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11902 BLKmode for ternary, binary or unary expressions.
11905 * var-tracking.c (delete_vta_debug_insn): New inline function.
11906 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11907 insns from get_insns () to NULL instead of each bb separately.
11908 Use delete_vta_debug_insn. No longer static.
11909 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11910 delete_vta_debug_insns callers.
11911 * rtl.h (delete_vta_debug_insns): Declare.
11912 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11913 instead of variable_tracking_main.
11915 2018-01-03 Martin Sebor <msebor@redhat.com>
11917 PR tree-optimization/83603
11918 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11919 arguments past the endof the argument list in functions declared
11920 without a prototype.
11921 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11922 Avoid checking when arguments are null.
11924 2018-01-03 Martin Sebor <msebor@redhat.com>
11927 * doc/extend.texi (attribute const): Fix a typo.
11928 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11929 issuing -Wsuggest-attribute for void functions.
11931 2018-01-03 Martin Sebor <msebor@redhat.com>
11933 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11934 offset_int::from instead of wide_int::to_shwi.
11935 (maybe_diag_overlap): Remove assertion.
11936 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11937 * gimple-ssa-sprintf.c (format_directive): Same.
11938 (parse_directive): Same.
11939 (sprintf_dom_walker::compute_format_length): Same.
11940 (try_substitute_return_value): Same.
11942 2018-01-03 Jeff Law <law@redhat.com>
11944 PR middle-end/83654
11945 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11946 non-constant residual for zero at runtime and avoid probing in
11947 that case. Reorganize code for trailing problem to mirror handling
11950 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11952 PR tree-optimization/83501
11953 * tree-ssa-strlen.c (get_string_cst): New.
11954 (handle_char_store): Call get_string_cst.
11956 2018-01-03 Martin Liska <mliska@suse.cz>
11958 PR tree-optimization/83593
11959 * tree-ssa-strlen.c: Include tree-cfg.h.
11960 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11961 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11962 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11964 (strlen_dom_walker::before_dom_children): Call
11965 gimple_purge_dead_eh_edges. Dump tranformation with details
11967 (strlen_dom_walker::before_dom_children): Update call by adding
11968 new argument cleanup_eh.
11969 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11971 2018-01-03 Martin Liska <mliska@suse.cz>
11974 * cif-code.def (VARIADIC_THUNK): New enum value.
11975 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11978 2018-01-03 Jan Beulich <jbeulich@suse.com>
11980 * sse.md (mov<mode>_internal): Tighten condition for when to use
11981 vmovdqu<ssescalarsize> for TI and OI modes.
11983 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11985 Update copyright years.
11987 2018-01-03 Martin Liska <mliska@suse.cz>
11990 * ipa-visibility.c (function_and_variable_visibility): Skip
11991 functions with noipa attribure.
11993 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11995 * gcc.c (process_command): Update copyright notice dates.
11996 * gcov-dump.c (print_version): Ditto.
11997 * gcov.c (print_version): Ditto.
11998 * gcov-tool.c (print_version): Ditto.
11999 * gengtype.c (create_file): Ditto.
12000 * doc/cpp.texi: Bump @copying's copyright year.
12001 * doc/cppinternals.texi: Ditto.
12002 * doc/gcc.texi: Ditto.
12003 * doc/gccint.texi: Ditto.
12004 * doc/gcov.texi: Ditto.
12005 * doc/install.texi: Ditto.
12006 * doc/invoke.texi: Ditto.
12008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12010 * vector-builder.h (vector_builder::m_full_nelts): Change from
12011 unsigned int to poly_uint64.
12012 (vector_builder::full_nelts): Update prototype accordingly.
12013 (vector_builder::new_vector): Likewise.
12014 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12015 (vector_builder::operator ==): Likewise.
12016 (vector_builder::finalize): Likewise.
12017 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12018 Take the number of elements as a poly_uint64 rather than an
12020 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12021 from unsigned int to poly_uint64.
12022 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12023 (vec_perm_indices::new_vector): Likewise.
12024 (vec_perm_indices::length): Likewise.
12025 (vec_perm_indices::nelts_per_input): Likewise.
12026 (vec_perm_indices::input_nelts): Likewise.
12027 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12028 number of elements per input as a poly_uint64 rather than an
12029 unsigned int. Use the original encoding for variable-length
12030 vectors, rather than clamping each individual element.
12031 For the second and subsequent elements in each pattern,
12032 clamp the step and base before clamping their sum.
12033 (vec_perm_indices::series_p): Handle polynomial element counts.
12034 (vec_perm_indices::all_in_range_p): Likewise.
12035 (vec_perm_indices_to_tree): Likewise.
12036 (vec_perm_indices_to_rtx): Likewise.
12037 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12038 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12039 (tree_vector_builder::new_binary_operation): Handle polynomial
12040 element counts. Return false if we need to know the number
12041 of elements at compile time.
12042 * fold-const.c (fold_vec_perm): Punt if the number of elements
12043 isn't known at compile time.
12045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12047 * vec-perm-indices.h (vec_perm_builder): Change element type
12048 from HOST_WIDE_INT to poly_int64.
12049 (vec_perm_indices::element_type): Update accordingly.
12050 (vec_perm_indices::clamp): Handle polynomial element_types.
12051 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12052 (vec_perm_indices::all_in_range_p): Likewise.
12053 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12055 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12056 polynomial vec_perm_indices element types.
12057 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12058 * fold-const.c (fold_vec_perm): Likewise.
12059 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12060 * tree-vect-generic.c (lower_vec_perm): Likewise.
12061 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12062 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12063 element type to HOST_WIDE_INT.
12065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12066 Alan Hayward <alan.hayward@arm.com>
12067 David Sherwood <david.sherwood@arm.com>
12069 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12070 rather than an int. Use plus_constant.
12071 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12072 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12075 Alan Hayward <alan.hayward@arm.com>
12076 David Sherwood <david.sherwood@arm.com>
12078 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12079 a HOST_WIDE_INT to a poly_int64.
12081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12082 Alan Hayward <alan.hayward@arm.com>
12083 David Sherwood <david.sherwood@arm.com>
12085 * calls.c (load_register_parameters): Cope with polynomial
12086 mode sizes. Require a constant size for BLKmode parameters
12087 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12088 forces a parameter to be padded at the lsb end in order to
12089 fill a complete number of words, require the parameter size
12090 to be ordered wrt UNITS_PER_WORD.
12092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12093 Alan Hayward <alan.hayward@arm.com>
12094 David Sherwood <david.sherwood@arm.com>
12096 * reload1.c (spill_stack_slot_width): Change element type
12097 from unsigned int to poly_uint64_pod.
12098 (alter_reg): Treat mode sizes as polynomial.
12100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12101 Alan Hayward <alan.hayward@arm.com>
12102 David Sherwood <david.sherwood@arm.com>
12104 * reload.c (complex_word_subreg_p): New function.
12105 (reload_inner_reg_of_subreg, push_reload): Use it.
12107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12108 Alan Hayward <alan.hayward@arm.com>
12109 David Sherwood <david.sherwood@arm.com>
12111 * lra-constraints.c (process_alt_operands): Reject matched
12112 operands whose sizes aren't ordered.
12113 (match_reload): Refer to this check here.
12115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12116 Alan Hayward <alan.hayward@arm.com>
12117 David Sherwood <david.sherwood@arm.com>
12119 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12120 that the mode size is in the set {1, 2, 4, 8, 16}.
12122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12123 Alan Hayward <alan.hayward@arm.com>
12124 David Sherwood <david.sherwood@arm.com>
12126 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12127 Use plus_constant instead of gen_rtx_PLUS.
12129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12130 Alan Hayward <alan.hayward@arm.com>
12131 David Sherwood <david.sherwood@arm.com>
12133 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12134 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12135 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12136 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12137 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12138 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12139 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12140 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12141 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12142 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12144 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12145 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12146 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12147 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12148 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12149 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12150 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12151 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12152 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12153 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12155 * expr.c (emit_move_resolve_push): Treat the input and result
12156 of PUSH_ROUNDING as a poly_int64.
12157 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12158 (emit_push_insn): Likewise.
12159 * lra-eliminations.c (mark_not_eliminable): Likewise.
12160 * recog.c (push_operand): Likewise.
12161 * reload1.c (elimination_effects): Likewise.
12162 * rtlanal.c (nonzero_bits1): Likewise.
12163 * calls.c (store_one_arg): Likewise. Require the padding to be
12164 known at compile time.
12166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12167 Alan Hayward <alan.hayward@arm.com>
12168 David Sherwood <david.sherwood@arm.com>
12170 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12171 Use plus_constant instead of gen_rtx_PLUS.
12173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12174 Alan Hayward <alan.hayward@arm.com>
12175 David Sherwood <david.sherwood@arm.com>
12177 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12178 rather than an int.
12180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12181 Alan Hayward <alan.hayward@arm.com>
12182 David Sherwood <david.sherwood@arm.com>
12184 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12185 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12186 via stack temporaries. Treat the mode size as polynomial too.
12188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12189 Alan Hayward <alan.hayward@arm.com>
12190 David Sherwood <david.sherwood@arm.com>
12192 * expr.c (expand_expr_real_2): When handling conversions involving
12193 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12194 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12195 as a poly_uint64 too.
12197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12198 Alan Hayward <alan.hayward@arm.com>
12199 David Sherwood <david.sherwood@arm.com>
12201 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12204 Alan Hayward <alan.hayward@arm.com>
12205 David Sherwood <david.sherwood@arm.com>
12207 * combine.c (can_change_dest_mode): Handle polynomial
12208 REGMODE_NATURAL_SIZE.
12209 * expmed.c (store_bit_field_1): Likewise.
12210 * expr.c (store_constructor): Likewise.
12211 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12212 and polynomial REGMODE_NATURAL_SIZE.
12213 (gen_lowpart_common): Likewise.
12214 * reginfo.c (record_subregs_of_mode): Likewise.
12215 * rtlanal.c (read_modify_subreg_p): Likewise.
12217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12218 Alan Hayward <alan.hayward@arm.com>
12219 David Sherwood <david.sherwood@arm.com>
12221 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12222 numbers of elements.
12224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12225 Alan Hayward <alan.hayward@arm.com>
12226 David Sherwood <david.sherwood@arm.com>
12228 * match.pd: Cope with polynomial numbers of vector elements.
12230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12231 Alan Hayward <alan.hayward@arm.com>
12232 David Sherwood <david.sherwood@arm.com>
12234 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12235 in a POINTER_PLUS_EXPR.
12237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12238 Alan Hayward <alan.hayward@arm.com>
12239 David Sherwood <david.sherwood@arm.com>
12241 * omp-simd-clone.c (simd_clone_subparts): New function.
12242 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12243 (ipa_simd_modify_function_body): Likewise.
12245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12246 Alan Hayward <alan.hayward@arm.com>
12247 David Sherwood <david.sherwood@arm.com>
12249 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12250 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12251 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12252 (expand_vector_condition, vector_element): Likewise.
12253 (subparts_gt): New function.
12254 (get_compute_type): Use subparts_gt.
12255 (count_type_subparts): Delete.
12256 (expand_vector_operations_1): Use subparts_gt instead of
12257 count_type_subparts.
12259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12260 Alan Hayward <alan.hayward@arm.com>
12261 David Sherwood <david.sherwood@arm.com>
12263 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12264 (vect_compile_time_alias): ...this new function. Do the calculation
12265 on poly_ints rather than trees.
12266 (vect_prune_runtime_alias_test_list): Update call accordingly.
12268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12269 Alan Hayward <alan.hayward@arm.com>
12270 David Sherwood <david.sherwood@arm.com>
12272 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12274 (vect_schedule_slp_instance): Likewise.
12276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12277 Alan Hayward <alan.hayward@arm.com>
12278 David Sherwood <david.sherwood@arm.com>
12280 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12281 constant and extern definitions for variable-length vectors.
12282 (vect_get_constant_vectors): Note that the number of units
12283 is known to be constant.
12285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12286 Alan Hayward <alan.hayward@arm.com>
12287 David Sherwood <david.sherwood@arm.com>
12289 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12290 of units as polynomial. Choose between WIDE and NARROW based
12293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12294 Alan Hayward <alan.hayward@arm.com>
12295 David Sherwood <david.sherwood@arm.com>
12297 * tree-vect-stmts.c (simd_clone_subparts): New function.
12298 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12301 Alan Hayward <alan.hayward@arm.com>
12302 David Sherwood <david.sherwood@arm.com>
12304 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12305 vectors as polynomial. Use build_index_vector for
12306 IFN_GOMP_SIMD_LANE.
12308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12309 Alan Hayward <alan.hayward@arm.com>
12310 David Sherwood <david.sherwood@arm.com>
12312 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12313 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12314 for variable-length vectors.
12315 (vectorizable_mask_load_store): Treat the number of units as
12316 polynomial, asserting that it is constant if the condition has
12317 already been enforced.
12318 (vectorizable_store, vectorizable_load): Likewise.
12320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12321 Alan Hayward <alan.hayward@arm.com>
12322 David Sherwood <david.sherwood@arm.com>
12324 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12325 of units as polynomial. Punt if we can't tell at compile time
12326 which vector contains the final result.
12328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12329 Alan Hayward <alan.hayward@arm.com>
12330 David Sherwood <david.sherwood@arm.com>
12332 * tree-vect-loop.c (vectorizable_induction): Treat the number
12333 of units as polynomial. Punt on SLP inductions. Use an integer
12334 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12335 cast of such a series for variable-length floating-point
12338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12339 Alan Hayward <alan.hayward@arm.com>
12340 David Sherwood <david.sherwood@arm.com>
12342 * tree.h (build_index_vector): Declare.
12343 * tree.c (build_index_vector): New function.
12344 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12345 of units as polynomial, forcibly converting it to a constant if
12346 vectorizable_reduction has already enforced the condition.
12347 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12348 to create a {1,2,3,...} vector.
12349 (vectorizable_reduction): Treat the number of units as polynomial.
12350 Choose vectype_in based on the largest scalar element size rather
12351 than the smallest number of units. Enforce the restrictions
12354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12355 Alan Hayward <alan.hayward@arm.com>
12356 David Sherwood <david.sherwood@arm.com>
12358 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12359 number of units as polynomial.
12361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12362 Alan Hayward <alan.hayward@arm.com>
12363 David Sherwood <david.sherwood@arm.com>
12365 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12366 * target.def (autovectorize_vector_sizes): Return the vector sizes
12367 by pointer, using vector_sizes rather than a bitmask.
12368 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12369 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12370 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12372 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12373 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12374 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12375 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12376 * omp-general.c (omp_max_vf): Likewise.
12377 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12378 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12379 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12380 * tree-vect-slp.c (vect_slp_bb): Likewise.
12381 * doc/tm.texi: Regenerate.
12382 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12384 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12385 the vector size as a poly_uint64 rather than an unsigned int.
12386 (current_vector_size): Change from an unsigned int to a poly_uint64.
12387 (get_vectype_for_scalar_type): Update accordingly.
12388 * tree.h (build_truth_vector_type): Take the size and number of
12389 units as a poly_uint64 rather than an unsigned int.
12390 (build_vector_type): Add a temporary overload that takes
12391 the number of units as a poly_uint64 rather than an unsigned int.
12392 * tree.c (make_vector_type): Likewise.
12393 (build_truth_vector_type): Take the number of units as a poly_uint64
12394 rather than an unsigned int.
12396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12397 Alan Hayward <alan.hayward@arm.com>
12398 David Sherwood <david.sherwood@arm.com>
12400 * target.def (get_mask_mode): Take the number of units and length
12401 as poly_uint64s rather than unsigned ints.
12402 * targhooks.h (default_get_mask_mode): Update accordingly.
12403 * targhooks.c (default_get_mask_mode): Likewise.
12404 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12405 * doc/tm.texi: Regenerate.
12407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12408 Alan Hayward <alan.hayward@arm.com>
12409 David Sherwood <david.sherwood@arm.com>
12411 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12412 * omp-general.c (omp_max_vf): Likewise.
12413 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12414 (expand_omp_simd): Handle polynomial safelen.
12415 * omp-low.c (omplow_simd_context): Add a default constructor.
12416 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12417 (lower_rec_simd_input_clauses): Update accordingly.
12418 (lower_rec_input_clauses): Likewise.
12420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12421 Alan Hayward <alan.hayward@arm.com>
12422 David Sherwood <david.sherwood@arm.com>
12424 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12425 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12426 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12427 (vect_analyze_slp_cost): Likewise.
12428 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12429 (vect_model_load_cost): Likewise.
12431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12432 Alan Hayward <alan.hayward@arm.com>
12433 David Sherwood <david.sherwood@arm.com>
12435 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12436 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12437 from an unsigned int * to a poly_uint64_pod *.
12438 (calculate_unrolling_factor): New function.
12439 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12442 Alan Hayward <alan.hayward@arm.com>
12443 David Sherwood <david.sherwood@arm.com>
12445 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12446 from an unsigned int to a poly_uint64.
12447 (_loop_vec_info::slp_unrolling_factor): Likewise.
12448 (_loop_vec_info::vectorization_factor): Change from an int
12450 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12451 (vect_get_num_vectors): New function.
12452 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12453 (vect_get_num_copies): Use vect_get_num_vectors.
12454 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12455 to an unsigned int *.
12456 (vect_analyze_data_refs): Change min_vf from an int * to a
12458 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12459 than an unsigned HOST_WIDE_INT.
12460 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12461 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12462 to an unsigned int *.
12463 (vect_analyze_data_ref_dependences): Likewise.
12464 (vect_compute_data_ref_alignment): Handle polynomial vf.
12465 (vect_enhance_data_refs_alignment): Likewise.
12466 (vect_prune_runtime_alias_test_list): Likewise.
12467 (vect_shift_permute_load_chain): Likewise.
12468 (vect_supportable_dr_alignment): Likewise.
12469 (dependence_distance_ge_vf): Take the vectorization factor as a
12470 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12471 (vect_analyze_data_refs): Change min_vf from an int * to a
12473 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12474 vfm1 as a poly_uint64 rather than an int. Make the same change
12475 for the returned bound_scalar.
12476 (vect_gen_vector_loop_niters): Handle polynomial vf.
12477 (vect_do_peeling): Likewise. Update call to
12478 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12479 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12481 * tree-vect-loop.c (vect_determine_vectorization_factor)
12482 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12483 (vect_get_known_peeling_cost): Likewise.
12484 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12485 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12486 (vect_transform_loop): Likewise. Use the lowest possible VF when
12487 updating the upper bounds of the loop.
12488 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12489 rather than an int.
12490 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12491 polynomial unroll factors.
12492 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12493 (vect_make_slp_decision): Likewise.
12494 (vect_supported_load_permutation_p): Likewise, and polynomial
12496 (vect_analyze_slp_cost): Handle polynomial vf.
12497 (vect_slp_analyze_node_operations): Likewise.
12498 (vect_slp_analyze_bb_1): Likewise.
12499 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12500 than an unsigned HOST_WIDE_INT.
12501 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12502 (vectorizable_load): Handle polynomial vf.
12503 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12505 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12508 Alan Hayward <alan.hayward@arm.com>
12509 David Sherwood <david.sherwood@arm.com>
12511 * match.pd: Handle bit operations involving three constants
12512 and try to fold one pair.
12514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12516 * tree-vect-loop-manip.c: Include gimple-fold.h.
12517 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12518 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12519 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12520 Add a path that uses a step of VF instead of 1, but disable it
12522 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12523 and niters_no_overflow parameters. Update calls to
12524 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12525 Create a new SSA name if the latter choses to use a ste other
12526 than zero, and return it via niters_vector_mult_vf_var.
12527 * tree-vect-loop.c (vect_transform_loop): Update calls to
12528 vect_do_peeling, vect_gen_vector_loop_niters and
12529 slpeel_make_loop_iterate_ntimes.
12530 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12531 (vect_gen_vector_loop_niters): Update declarations after above changes.
12533 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12535 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12536 128-bit round to integer instructions.
12537 (ceil<mode>2): Likewise.
12538 (btrunc<mode>2): Likewise.
12539 (round<mode>2): Likewise.
12541 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12543 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12544 unaligned VSX load/store on P8/P9.
12545 (expand_block_clear): Allow the use of unaligned VSX
12546 load/store on P8/P9.
12548 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12550 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12552 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12553 swap associated with both a load and a store.
12555 2018-01-02 Andrew Waterman <andrew@sifive.com>
12557 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12558 * config/riscv/riscv.md (clear_cache): Use it.
12560 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12562 * web.c: Remove out-of-date comment.
12564 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12566 * expr.c (fixup_args_size_notes): Check that any existing
12567 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12568 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12569 (emit_single_push_insn): ...here.
12571 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12573 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12574 (const_vector_encoded_nelts): New function.
12575 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12576 (const_vector_int_elt, const_vector_elt): Declare.
12577 * emit-rtl.c (const_vector_int_elt_1): New function.
12578 (const_vector_elt): Likewise.
12579 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12580 of CONST_VECTOR_ELT.
12582 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12584 * expr.c: Include rtx-vector-builder.h.
12585 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12586 directly on the tree encoding.
12587 (const_vector_from_tree): Likewise.
12588 * optabs.c: Include rtx-vector-builder.h.
12589 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12590 sequence of "u" values.
12591 * vec-perm-indices.c: Include rtx-vector-builder.h.
12592 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12593 directly on the vec_perm_indices encoding.
12595 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12597 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12598 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12599 * rtx-vector-builder.h: New file.
12600 * rtx-vector-builder.c: Likewise.
12601 * rtl.h (rtx_def::u2): Add a const_vector field.
12602 (CONST_VECTOR_NPATTERNS): New macro.
12603 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12604 (CONST_VECTOR_DUPLICATE_P): Likewise.
12605 (CONST_VECTOR_STEPPED_P): Likewise.
12606 (CONST_VECTOR_ENCODED_ELT): Likewise.
12607 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12608 (unwrap_const_vec_duplicate): Likewise.
12609 (const_vec_series_p): Check for a non-duplicated vector encoding.
12610 Say that the function only returns true for integer vectors.
12611 * emit-rtl.c: Include rtx-vector-builder.h.
12612 (gen_const_vec_duplicate_1): Delete.
12613 (gen_const_vector): Call gen_const_vec_duplicate instead of
12614 gen_const_vec_duplicate_1.
12615 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12616 (gen_const_vec_duplicate): Use rtx_vector_builder.
12617 (gen_const_vec_series): Likewise.
12618 (gen_rtx_CONST_VECTOR): Likewise.
12619 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12620 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12621 Build a new vector rather than modifying a CONST_VECTOR in-place.
12622 (handle_special_swappables): Update call accordingly.
12623 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12624 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12625 Build a new vector rather than modifying a CONST_VECTOR in-place.
12626 (handle_special_swappables): Update call accordingly.
12628 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12630 * simplify-rtx.c (simplify_const_binary_operation): Use
12631 CONST_VECTOR_ELT instead of XVECEXP.
12633 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12635 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12636 the selector elements to be different from the data elements
12637 if the selector is a VECTOR_CST.
12638 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12639 ssizetype for the selector.
12641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12643 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12644 before testing each element individually.
12645 * tree-vect-generic.c (lower_vec_perm): Likewise.
12647 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12649 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12650 * selftest-run-tests.c (selftest::run_tests): Call it.
12651 * vector-builder.h (vector_builder::operator ==): New function.
12652 (vector_builder::operator !=): Likewise.
12653 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12654 (vec_perm_indices::all_from_input_p): New function.
12655 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12656 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12657 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12658 instead of reading the VECTOR_CST directly. Detect whether both
12659 vector inputs are the same before constructing the vec_perm_indices,
12660 and update the number of inputs argument accordingly. Use the
12661 utility functions added above. Only construct sel2 if we need to.
12663 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12665 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12666 the broadcast of the low byte.
12667 (expand_mult_highpart): Use an explicit encoding for the permutes.
12668 * optabs-query.c (can_mult_highpart_p): Likewise.
12669 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12670 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12671 (vectorizable_bswap): Likewise.
12672 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12673 explicit encoding for the power-of-2 permutes.
12674 (vect_permute_store_chain): Likewise.
12675 (vect_grouped_load_supported): Likewise.
12676 (vect_permute_load_chain): Likewise.
12678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12680 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12681 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12682 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12683 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12684 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12685 (vect_gen_perm_mask_any): Likewise.
12687 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12689 * int-vector-builder.h: New file.
12690 * vec-perm-indices.h: Include int-vector-builder.h.
12691 (vec_perm_indices): Redefine as an int_vector_builder.
12692 (auto_vec_perm_indices): Delete.
12693 (vec_perm_builder): Redefine as a stand-alone class.
12694 (vec_perm_indices::vec_perm_indices): New function.
12695 (vec_perm_indices::clamp): Likewise.
12696 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12697 (vec_perm_indices::new_vector): New function.
12698 (vec_perm_indices::new_expanded_vector): Update for new
12699 vec_perm_indices class.
12700 (vec_perm_indices::rotate_inputs): New function.
12701 (vec_perm_indices::all_in_range_p): Operate directly on the
12702 encoded form, without computing elided elements.
12703 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12704 encoding. Update for new vec_perm_indices class.
12705 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12706 the given vec_perm_builder.
12707 (expand_vec_perm_var): Update vec_perm_builder constructor.
12708 (expand_mult_highpart): Use vec_perm_builder instead of
12709 auto_vec_perm_indices.
12710 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12711 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12712 or double series encoding as appropriate.
12713 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12714 vec_perm_indices instead of auto_vec_perm_indices.
12715 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12716 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12717 (vect_permute_store_chain): Likewise.
12718 (vect_grouped_load_supported): Likewise.
12719 (vect_permute_load_chain): Likewise.
12720 (vect_shift_permute_load_chain): Likewise.
12721 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12722 (vect_transform_slp_perm_load): Likewise.
12723 (vect_schedule_slp_instance): Likewise.
12724 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12725 (vectorizable_mask_load_store): Likewise.
12726 (vectorizable_bswap): Likewise.
12727 (vectorizable_store): Likewise.
12728 (vectorizable_load): Likewise.
12729 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12730 vec_perm_indices instead of auto_vec_perm_indices. Use
12731 tree_to_vec_perm_builder to read the vector from a tree.
12732 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12733 vec_perm_builder instead of a vec_perm_indices.
12734 (have_whole_vector_shift): Use vec_perm_builder and
12735 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12736 truncation to calc_vec_perm_mask_for_shift.
12737 (vect_create_epilog_for_reduction): Likewise.
12738 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12739 from auto_vec_perm_indices to vec_perm_indices.
12740 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12741 instead of changing individual elements.
12742 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12743 the vector in d.perm.
12744 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12745 from auto_vec_perm_indices to vec_perm_indices.
12746 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12747 instead of changing individual elements.
12748 (arm_vectorize_vec_perm_const): Use new_vector to install
12749 the vector in d.perm.
12750 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12751 Update vec_perm_builder constructor.
12752 (rs6000_expand_interleave): Likewise.
12753 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12754 (rs6000_expand_interleave): Likewise.
12756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12758 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12759 to qimode could truncate the indices.
12760 * optabs.c (expand_vec_perm_var): Likewise.
12762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12764 * Makefile.in (OBJS): Add vec-perm-indices.o.
12765 * vec-perm-indices.h: New file.
12766 * vec-perm-indices.c: Likewise.
12767 * target.h (vec_perm_indices): Replace with a forward class
12769 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12770 * optabs.h: Include vec-perm-indices.h.
12771 (expand_vec_perm): Delete.
12772 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12773 (expand_vec_perm_const): Declare.
12774 * target.def (vec_perm_const_ok): Replace with...
12775 (vec_perm_const): ...this new hook.
12776 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12777 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12778 * doc/tm.texi: Regenerate.
12779 * optabs.def (vec_perm_const): Delete.
12780 * doc/md.texi (vec_perm_const): Likewise.
12781 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12782 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12783 expand_vec_perm for constant permutation vectors. Assert that
12784 the mode of variable permutation vectors is the integer equivalent
12785 of the mode that is being permuted.
12786 * optabs-query.h (selector_fits_mode_p): Declare.
12787 * optabs-query.c: Include vec-perm-indices.h.
12788 (selector_fits_mode_p): New function.
12789 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12790 is defined, instead of checking whether the vec_perm_const_optab
12791 exists. Use targetm.vectorize.vec_perm_const instead of
12792 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12793 fit in the vector mode before using a variable permute.
12794 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12795 vec_perm_indices instead of an rtx.
12796 (expand_vec_perm): Replace with...
12797 (expand_vec_perm_const): ...this new function. Take the selector
12798 as a vec_perm_indices rather than an rtx. Also take the mode of
12799 the selector. Update call to shift_amt_for_vec_perm_mask.
12800 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12801 Use vec_perm_indices::new_expanded_vector to expand the original
12802 selector into bytes. Check whether the indices fit in the vector
12803 mode before using a variable permute.
12804 (expand_vec_perm_var): Make global.
12805 (expand_mult_highpart): Use expand_vec_perm_const.
12806 * fold-const.c: Includes vec-perm-indices.h.
12807 * tree-ssa-forwprop.c: Likewise.
12808 * tree-vect-data-refs.c: Likewise.
12809 * tree-vect-generic.c: Likewise.
12810 * tree-vect-loop.c: Likewise.
12811 * tree-vect-slp.c: Likewise.
12812 * tree-vect-stmts.c: Likewise.
12813 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12815 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12816 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12817 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12818 (aarch64_vectorize_vec_perm_const): ...this new function.
12819 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12820 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12821 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12822 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12823 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12824 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12825 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12827 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12828 check for NEON modes.
12829 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12830 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12831 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12832 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12834 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12835 the old VEC_PERM_CONST conditions.
12836 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12837 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12838 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12839 (ia64_vectorize_vec_perm_const_ok): Merge into...
12840 (ia64_vectorize_vec_perm_const): ...this new function.
12841 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12842 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12843 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12844 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12845 * config/mips/mips.c (mips_expand_vec_perm_const)
12846 (mips_vectorize_vec_perm_const_ok): Merge into...
12847 (mips_vectorize_vec_perm_const): ...this new function.
12848 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12849 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12850 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12851 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12852 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12853 (rs6000_expand_vec_perm_const): Delete.
12854 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12856 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12857 (altivec_expand_vec_perm_const_le): Take each operand individually.
12858 Operate on constant selectors rather than rtxes.
12859 (altivec_expand_vec_perm_const): Likewise. Update call to
12860 altivec_expand_vec_perm_const_le.
12861 (rs6000_expand_vec_perm_const): Delete.
12862 (rs6000_vectorize_vec_perm_const_ok): Delete.
12863 (rs6000_vectorize_vec_perm_const): New function.
12864 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12865 an element count and rtx array.
12866 (rs6000_expand_extract_even): Update call accordingly.
12867 (rs6000_expand_interleave): Likewise.
12868 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12869 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12870 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12871 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12872 (rs6000_expand_vec_perm_const): Delete.
12873 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12874 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12875 (altivec_expand_vec_perm_const_le): Take each operand individually.
12876 Operate on constant selectors rather than rtxes.
12877 (altivec_expand_vec_perm_const): Likewise. Update call to
12878 altivec_expand_vec_perm_const_le.
12879 (rs6000_expand_vec_perm_const): Delete.
12880 (rs6000_vectorize_vec_perm_const_ok): Delete.
12881 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12882 reference to the SPE evmerge intructions.
12883 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12884 an element count and rtx array.
12885 (rs6000_expand_extract_even): Update call accordingly.
12886 (rs6000_expand_interleave): Likewise.
12887 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12888 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12890 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12892 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12894 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12895 vector mode and that that mode matches the mode of the data
12897 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12898 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12899 directly using expand_vec_perm_1 when forcing selectors into
12901 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12903 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12905 * optabs-query.h (can_vec_perm_p): Delete.
12906 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12907 * optabs-query.c (can_vec_perm_p): Split into...
12908 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12909 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12910 particular selector is valid.
12911 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12912 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12913 (vect_grouped_load_supported): Likewise.
12914 (vect_shift_permute_load_chain): Likewise.
12915 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12916 (vect_transform_slp_perm_load): Likewise.
12917 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12918 (vectorizable_bswap): Likewise.
12919 (vect_gen_perm_mask_checked): Likewise.
12920 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12921 implementations of variable permutation vectors into account
12922 when deciding which selector to use.
12923 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12924 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12925 with a false third argument.
12926 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12927 to test whether the constant selector is valid and can_vec_perm_var_p
12928 to test whether a variable selector is valid.
12930 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12932 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12933 * optabs-query.c (can_vec_perm_p): Likewise.
12934 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12935 instead of vec_perm_indices.
12936 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12937 (vect_gen_perm_mask_checked): Likewise,
12938 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12939 (vect_gen_perm_mask_checked): Likewise,
12941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12943 * optabs-query.h (qimode_for_vec_perm): Declare.
12944 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12945 (qimode_for_vec_perm): ...this new function.
12946 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12948 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12950 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12951 does not have a conditional at the top.
12953 2018-01-02 Richard Biener <rguenther@suse.de>
12955 * ipa-inline.c (big_speedup_p): Fix expression.
12957 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12960 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12963 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12967 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12968 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12969 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12970 cond_taken_branch_cost 3->4.
12972 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12974 PR tree-optimization/83581
12975 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12976 TODO_cleanup_cfg if any changes have been made.
12978 PR middle-end/83608
12979 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12980 convert_modes if target mode has the right side, but different mode
12983 PR middle-end/83609
12984 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12985 last argument when extracting from CONCAT. If either from_real or
12986 from_imag is NULL, use expansion through memory. If result is not
12987 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12988 the parts directly to inner mode, if even that fails, use expansion
12991 PR middle-end/83623
12992 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12993 check for bswap in mode rather than HImode and use that in expand_unop
12996 Copyright (C) 2018 Free Software Foundation, Inc.
12998 Copying and distribution of this file, with or without modification,
12999 are permitted in any medium without royalty provided the copyright
13000 notice and this notice are preserved.