1 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
4 (path_range_query::precompute_ranges_in_block): Rename to...
5 (path_range_query::compute_ranges_in_block): ...this.
6 (path_range_query::precompute_ranges): Rename to...
7 (path_range_query::compute_ranges): ...this.
8 (path_range_query::precompute_relations): Rename to...
9 (path_range_query::compute_relations): ...this.
10 (path_range_query::precompute_phi_relations): Rename to...
11 (path_range_query::compute_phi_relations): ...this.
12 * gimple-range-path.h: Rename precompute* to compute*.
13 * tree-ssa-threadbackward.c
14 (back_threader::find_taken_edge_switch): Same.
15 (back_threader::find_taken_edge_cond): Same.
16 * tree-ssa-threadedge.c
17 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
18 (hybrid_jt_state::register_equivs_stmt): Inline...
19 * tree-ssa-threadedge.h: ...here.
21 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
23 * tree-vrp.c (lhs_of_dominating_assert): Remove.
24 (class vrp_jt_state): Remove.
25 (class vrp_jt_simplifier): Remove.
26 (vrp_jt_simplifier::simplify): Remove.
27 (class vrp_jump_threader): Remove.
28 (vrp_jump_threader::vrp_jump_threader): Remove.
29 (vrp_jump_threader::~vrp_jump_threader): Remove.
30 (vrp_jump_threader::before_dom_children): Remove.
31 (vrp_jump_threader::after_dom_children): Remove.
33 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
35 * passes.def (pass_vrp_threader): New.
36 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
37 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
38 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
39 (hybrid_jt_simplifier::simplify): New.
40 (hybrid_jt_simplifier::compute_ranges_from_state): New.
41 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
42 (class hybrid_jt_simplifier): New.
43 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
45 (class hybrid_threader): New.
46 (hybrid_threader::hybrid_threader): New.
47 (hybrid_threader::~hybrid_threader): New.
48 (hybrid_threader::before_dom_children): New.
49 (hybrid_threader::after_dom_children): New.
50 (execute_vrp_threader): New.
51 (class pass_vrp_threader): New.
52 (make_pass_vrp_threader): New.
54 2021-09-27 Martin Liska <mliska@suse.cz>
56 * output.h (enum section_flag): New.
57 (SECTION_FORGET): Remove.
58 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
59 (SECTION_STYLE_MASK): Define it based on other enum
61 * varasm.c (switch_to_section): Remove unused handling of
64 2021-09-27 Martin Liska <mliska@suse.cz>
66 * common.opt: Add new variable flag_default_complex_method.
67 * opts.c (finish_options): Handle flags related to
68 x_flag_complex_method.
69 * toplev.c (process_options): Remove option handling related
70 to flag_complex_method.
72 2021-09-27 Richard Biener <rguenther@suse.de>
75 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
76 type_for_size, instead use int_mode_for_size.
78 2021-09-27 Andrew Pinski <apinski@marvell.com>
81 * gimplify.c (gimplify_save_expr): Return early
82 if the type of val is error_mark_node.
84 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
86 * tree-ssanames.c (ssa_name_has_boolean_range): Use
89 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
91 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
93 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
94 (cprop_into_stmt): Same.
95 (dom_opt_dom_walker::optimize_stmt): Same.
97 2021-09-27 Richard Biener <rguenther@suse.de>
99 PR tree-optimization/97351
100 PR tree-optimization/97352
101 PR tree-optimization/82426
102 * tree-vectorizer.h (dr_misalignment): Add vector type
104 (aligned_access_p): Likewise.
105 (known_alignment_for_access_p): Likewise.
106 (vect_supportable_dr_alignment): Likewise.
107 (vect_known_alignment_in_bytes): Likewise. Refactor.
108 (DR_MISALIGNMENT): Remove.
109 (vect_update_shared_vectype): Likewise.
110 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
111 a vector type with larger alignment requirement and apply
112 the negative step adjustment here.
113 (vect_calculate_target_alignment): Remove.
114 (vect_compute_data_ref_alignment): Get explicit vector type
115 argument, do not apply a negative step alignment adjustment
117 (vect_slp_analyze_node_alignment): Re-analyze alignment
118 when we re-visit the DR with a bigger desired alignment but
119 keep more precise results from smaller alignments.
120 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
121 (vect_slp_analyze_node_operations_1): Do not update the
122 shared vector type on stmts.
123 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
124 vector type of an SLP node to the representative stmt-info.
125 (vect_transform_stmt): Likewise.
127 2021-09-27 liuhongt <hongtao.liu@intel.com>
130 2021-09-09 liuhongt <hongtao.liu@intel.com>
133 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
134 (reduc_plus_scal_v4sf): .. this, New define_expand.
135 (reduc_plus_scal_v2df): .. and this, New define_expand.
137 2021-09-26 liuhongt <hongtao.liu@intel.com>
139 * doc/extend.texi (Half-Precision): Remove storage only
140 description for _Float16 w/o avx512fp16.
142 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
144 * config/pru/constraints.md (Rrio): New constraint.
145 * config/pru/predicates.md (regio_operand): New predicate.
146 * config/pru/pru-pragma.c (pru_register_pragmas): Register
147 the __regio_symbol address space.
148 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
149 * config/pru/pru.c (pru_symref2ioregno): New helper function.
150 (pru_legitimate_address_p): Remove.
151 (pru_addr_space_legitimate_address_p): Use the address space
153 (pru_nongeneric_pointer_addrspace): New helper function.
154 (pru_insert_attributes): New function to validate __regio_symbol
156 (TARGET_INSERT_ATTRIBUTES): New macro.
157 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
158 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
159 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
160 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
162 (*regio_nozext_writesi): New pattern to write to I/O registers.
163 (*regio_zext_write_r30<EQS0:mode>): Ditto.
164 * doc/extend.texi: Document the new PRU Named Address Space.
166 2021-09-24 Patrick Palka <ppalka@redhat.com>
170 * real.c (encode_ieee_double): Avoid unwanted sign extension.
171 (encode_ieee_quad): Likewise.
173 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
175 PR rtl-optimization/102147
176 * ira-build.c (ira_conflict_vector_profitable_p): Make
177 profitability calculation independent of host compiler pointer and
180 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
182 * gimple-range-path.cc (path_range_query::path_range_query):
183 Move debugging header...
184 (path_range_query::precompute_ranges): ...here.
185 (path_range_query::internal_range_of_expr): Do not call
186 range_on_path_entry if NAME is defined in the current block.
188 2021-09-24 Richard Biener <rguenther@suse.de>
190 * cfghooks.c (verify_flow_info): Verify unallocated BB and
191 edge flags are not set.
193 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
195 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
197 (jt_path_registry::register_jump_thread): Call
198 cancel_invalid_paths.
199 * tree-ssa-threadupdate.h (class jt_path_registry): Add
200 cancel_invalid_paths.
202 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
204 PR tree-optimization/102400
205 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
206 result_vdef to zero value.
208 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
210 PR tree-optimization/102451
211 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
214 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
216 * config/i386/sse.md (cond_<insn><mode>): Extend to support
218 (cond_mul<mode>): Likewise.
219 (cond_div<mode>): Likewise.
220 (cond_<code><mode>): Likewise.
221 (cond_fma<mode>): Likewise.
222 (cond_fms<mode>): Likewise.
223 (cond_fnma<mode>): Likewise.
224 (cond_fnms<mode>): Likewise.
226 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
228 PR tree-optimization/102463
229 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
230 there is no range-ops handler, don't look for a relation.
232 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
234 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
235 non-executable_edge flag as parameter.
236 * gimple-range-cache.h (ranger_cache): Adjust prototype.
237 * gimple-range-gori.cc (gori_compute::gori_compute): Take
238 non-executable_edge flag as parameter.
239 (gori_compute::outgoing_edge_range_p): Check new flag.
240 * gimple-range-gori.h (gori_compute): Adjust prototype.
241 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
242 (gimple_ranger::range_on_edge): Check new flag.
243 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
244 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
245 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
246 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
247 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
248 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
249 (simplify_using_ranges::simplify_switch_using_ranges): Clear
250 EDGE_EXECUTABLE like it originally did.
251 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
252 NON_EXECUTABLE flags.
253 (simplify_using_ranges::simplify_using_ranges): Adjust.
254 * vr-values.h (class simplify_using_ranges): Adjust.
255 (simplify_using_ranges::set_range_query): Add non-executable flag param.
257 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
260 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
261 zero-width bit fields and return indicator.
262 (rs6000_discover_homogeneous_aggregate): Diagnose when the
263 presence of a zero-width bit field changes parameter passing in
266 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
268 * gimple-range-fold.cc (fold_using_range::range_of_phi):
269 Remove dominator check.
271 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
273 * gimple-range-path.cc (path_range_query::precompute_relations):
274 Hoist edge calculations before using EDGE_SUCC.
276 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
278 * configure.ac: Fix --with-multilib-list description.
279 * configure: Regenerate.
281 2021-09-23 Richard Biener <rguenther@suse.de>
283 PR tree-optimization/102448
284 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
285 Clear alignment info copied from DR_PTR_INFO.
287 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
289 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
291 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
292 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
293 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
294 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
295 (vcond<mode><mode>): New expander.
296 (vcond<mode><sseintvecmodelower>): Likewise.
297 (vcond<sseintvecmodelower><mode>): Likewise.
298 (vcondu<mode><sseintvecmodelower>): Likewise.
300 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
302 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
304 (extendv4hf<mode>2): Likewise.
305 (extendv2hfv2df2): Likewise.
306 (trunc<mode><ssePHmodelower>2): Likewise.
307 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
308 (trunc<mode>v4hf2): ... this, and drop constraints.
309 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
310 (truncv2dfv2hf2): ... this, and likewise.
312 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
314 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
316 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
318 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
319 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
320 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
322 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
324 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
325 (fixuns_trunchfhi2): Likewise.
326 (*fixuns_trunchfsi2zext): New define_insn.
327 * config/i386/sse.md (ssePHmodelower): New mode_attr.
328 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
329 New expander for same element vector fix_truncate.
330 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
331 Likewise for V4HF to V4SI/V4DI fix_truncate.
332 (fix<fixunssuffix>_truncv2hfv2di2):
333 Likeise for V2HF to V2DI fix_truncate.
335 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
337 * config/i386/i386.md (<code>hf3): New expander.
339 2021-09-23 liuhongt <hongtao.liu@intel.com>
341 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
342 (VFH_SF_AVX512VL): Extend to handle HFmode.
343 (VF_SF_AVX512VL): Deleted.
345 2021-09-23 liuhongt <hongtao.liu@intel.com>
347 * config/i386/i386.md (rinthf2): New expander.
348 (nearbyinthf2): New expander.
350 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
352 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
353 (class dom_jt_state): ...this and provide virtual overrides.
354 (dom_jt_state::register_equiv): New.
355 (class dom_jt_simplifier): Rename from
356 dom_jump_threader_simplifier.
357 (dom_jump_threader_simplifier::simplify): Rename...
358 (dom_jt_simplifier::simplify): ...to this.
359 (pass_dominator::execute): Use dom_jt_simplifier and
361 * tree-ssa-threadedge.c (jump_threader::jump_threader):
363 (jt_state::register_equivs_stmt): Abstract out...
364 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
366 (jump_threader::thread_around_empty_blocks): Update state.
367 (jump_threader::thread_through_normal_block): Same.
368 (jt_state::jt_state): Remove.
369 (jt_state::push): Remove pass specific bits. Keep block vector
371 (jt_state::append_path): New.
372 (jt_state::pop): Remove pass specific bits.
373 (jt_state::register_equiv): Same.
374 (jt_state::record_ranges_from_stmt): Same.
375 (jt_state::register_equivs_on_edge): Same. Rename...
376 (jt_state::register_equivs_edge): ...to this.
377 (jt_state::dump): New.
378 (jt_state::debug): New.
379 (jump_threader_simplifier::simplify): Remove.
380 (jt_state::get_path): New.
381 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
382 class. Expose common functionality as virtual methods.
383 (class jump_threader_simplifier): Same. Rename...
384 (class jt_simplifier): ...to this.
385 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
386 (class vrp_jt_simplifier): ...to this. Provide pass specific
388 (class vrp_jt_state): New.
389 (vrp_jump_threader_simplifier::simplify): Rename...
390 (vrp_jt_simplifier::simplify): ...to this. Inline code from
391 what used to be the base class.
392 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
395 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
398 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
401 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
402 Richard Biener <rguenther@suse.de>
404 * match.pd (negation simplifications): Implement some negation
405 folding transformations from fold-const.c's fold_negate_expr.
406 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
407 argument, to control whether the op should be simplified prior
408 to looking up/assigning a value number.
409 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
410 (vn_nary_simplify): Likewise.
411 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
412 now call vn_nary_build_or_lookup_1 disabling simplification.
414 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
416 PR tree-optimization/102087
417 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
418 Update bound/cmp/control for niter.
420 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
422 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
423 Move check for non-empty BB here.
424 (fur_source::register_outgoing_edges): ...from here.
426 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
428 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
429 Remove call to improve_range_with_equivs.
430 (path_range_query::improve_range_with_equivs): Remove
431 * gimple-range-path.h: Remove improve_range_with_equivs.
433 2021-09-22 dianhong xu <dianhong.xu@intel.com>
435 * config/i386/avx512fp16intrin.h:
436 (_mm512_mask_blend_ph): New intrinsic.
437 (_mm512_permutex2var_ph): Ditto.
438 (_mm512_permutexvar_ph): Ditto.
439 * config/i386/avx512fp16vlintrin.h:
440 (_mm256_mask_blend_ph): New intrinsic.
441 (_mm256_permutex2var_ph): Ditto.
442 (_mm256_permutexvar_ph): Ditto.
443 (_mm_mask_blend_ph): Ditto.
444 (_mm_permutex2var_ph): Ditto.
445 (_mm_permutexvar_ph): Ditto.
447 2021-09-22 dianhong xu <dianhong.xu@intel.com>
449 * config/i386/avx512fp16intrin.h: Add new intrinsics.
450 (_mm512_conj_pch): New intrinsic.
451 (_mm512_mask_conj_pch): Ditto.
452 (_mm512_maskz_conj_pch): Ditto.
453 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
454 (_mm256_conj_pch): New intrinsic.
455 (_mm256_mask_conj_pch): Ditto.
456 (_mm256_maskz_conj_pch): Ditto.
457 (_mm_conj_pch): Ditto.
458 (_mm_mask_conj_pch): Ditto.
459 (_mm_maskz_conj_pch): Ditto.
461 2021-09-22 dianhong xu <dianhong.xu@intel.com>
463 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
464 (_mm512_reduce_add_ph): New intrinsic.
465 (_mm512_reduce_mul_ph): Ditto.
466 (_mm512_reduce_min_ph): Ditto.
467 (_mm512_reduce_max_ph): Ditto.
468 * config/i386/avx512fp16vlintrin.h
469 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
470 (_mm256_reduce_add_ph): New intrinsic.
471 (_mm256_reduce_mul_ph): Ditto.
472 (_mm256_reduce_min_ph): Ditto.
473 (_mm256_reduce_max_ph): Ditto.
474 (_mm_reduce_add_ph): Ditto.
475 (_mm_reduce_mul_ph): Ditto.
476 (_mm_reduce_min_ph): Ditto.
477 (_mm_reduce_max_ph): Ditto.
479 2021-09-22 dianhong xu <dianhong.xu@intel.com>
481 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
482 __m128h_u): New typedef.
483 (_mm512_load_ph): New intrinsic.
484 (_mm256_load_ph): Ditto.
485 (_mm_load_ph): Ditto.
486 (_mm512_loadu_ph): Ditto.
487 (_mm256_loadu_ph): Ditto.
488 (_mm_loadu_ph): Ditto.
489 (_mm512_store_ph): Ditto.
490 (_mm256_store_ph): Ditto.
491 (_mm_store_ph): Ditto.
492 (_mm512_storeu_ph): Ditto.
493 (_mm256_storeu_ph): Ditto.
494 (_mm_storeu_ph): Ditto.
495 (_mm512_abs_ph): Ditto.
496 * config/i386/avx512fp16vlintrin.h
498 (_mm256_abs_ph): Ditto.
500 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
502 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
504 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
507 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
508 is actually a full copy of the source operand into the target.
509 Don't emit a strict low part move if source and target mode match.
511 2021-09-22 Jakub Jelinek <jakub@redhat.com>
514 * omp-expand.c (expand_omp_single): If region->exit is NULL,
515 assert region->entry is GIMPLE_OMP_SCOPE region and return.
517 2021-09-22 Jakub Jelinek <jakub@redhat.com>
519 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
520 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
521 arguments from 2 to 3.
522 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
523 allocate clause allocator and print align if present.
524 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
525 for omp_default_mem_alloc if align modifier is present. If align
526 modifier is present, use TREE_LIST to encode both allocator and
528 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
529 Handle align modifier on allocator clause if present.
531 2021-09-22 liuhongt <hongtao.liu@intel.com>
533 * config/i386/i386.md (define_attr "isa"): Add
535 (define_attr "enabled"): Correspond fma_or_avx512vl to
536 TARGET_FMA || TARGET_AVX512VL.
537 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
542 2021-09-22 liuhongt <hongtao.liu@intel.com>
544 * config/i386/i386.md (cstorehf3): New define_expand.
546 2021-09-22 liuhongt <hongtao.liu@intel.com>
548 * config/i386/i386.md (<rounding_insn>hf2): New expander.
549 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
550 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
551 Extend from VF_128 to VFH_128.
553 2021-09-22 liuhongt <hongtao.liu@intel.com>
555 * config/i386/i386-features.c (i386-features.c): Handle
557 * config/i386/i386.md (sqrthf2): New expander.
558 (*sqrthf2): New define_insn.
560 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
563 2021-09-22 liuhongt <hongtao.liu@intel.com>
565 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
567 (_mm_mask3_fcmadd_sch): Likewise.
568 (_mm_maskz_fcmadd_sch): Likewise.
569 (_mm_fcmadd_sch): Likewise.
570 (_mm_mask_fmadd_sch): Likewise.
571 (_mm_mask3_fmadd_sch): Likewise.
572 (_mm_maskz_fmadd_sch): Likewise.
573 (_mm_fmadd_sch): Likewise.
574 (_mm_mask_fcmadd_round_sch): Likewise.
575 (_mm_mask3_fcmadd_round_sch): Likewise.
576 (_mm_maskz_fcmadd_round_sch): Likewise.
577 (_mm_fcmadd_round_sch): Likewise.
578 (_mm_mask_fmadd_round_sch): Likewise.
579 (_mm_mask3_fmadd_round_sch): Likewise.
580 (_mm_maskz_fmadd_round_sch): Likewise.
581 (_mm_fmadd_round_sch): Likewise.
582 (_mm_fcmul_sch): Likewise.
583 (_mm_mask_fcmul_sch): Likewise.
584 (_mm_maskz_fcmul_sch): Likewise.
585 (_mm_fmul_sch): Likewise.
586 (_mm_mask_fmul_sch): Likewise.
587 (_mm_maskz_fmul_sch): Likewise.
588 (_mm_fcmul_round_sch): Likewise.
589 (_mm_mask_fcmul_round_sch): Likewise.
590 (_mm_maskz_fcmul_round_sch): Likewise.
591 (_mm_fmul_round_sch): Likewise.
592 (_mm_mask_fmul_round_sch): Likewise.
593 (_mm_maskz_fmul_round_sch): Likewise.
594 * config/i386/i386-builtin.def: Add corresponding new builtins.
596 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
597 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
598 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
600 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
601 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
603 * config/i386/subst.md (mask_scalarcz_name): New.
604 (mask_scalarc_name): Ditto.
605 (mask_scalarc_operand3): Ditto.
606 (mask_scalarcz_operand4): Ditto.
607 (round_scalarcz_name): Ditto.
608 (round_scalarc_mask_operand3): Ditto.
609 (round_scalarcz_mask_operand4): Ditto.
610 (round_scalarc_mask_op3): Ditto.
611 (round_scalarcz_mask_op4): Ditto.
612 (round_scalarcz_constraint): Ditto.
613 (round_scalarcz_nimm_predicate): Ditto.
614 (mask_scalarcz): Ditto.
615 (mask_scalarc): Ditto.
616 (round_scalarcz): Ditto.
618 2021-09-22 liuhongt <hongtao.liu@intel.com>
620 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
622 (_mm512_mask_fcmadd_pch): Likewise.
623 (_mm512_mask3_fcmadd_pch): Likewise.
624 (_mm512_maskz_fcmadd_pch): Likewise.
625 (_mm512_fmadd_pch): Likewise.
626 (_mm512_mask_fmadd_pch): Likewise.
627 (_mm512_mask3_fmadd_pch): Likewise.
628 (_mm512_maskz_fmadd_pch): Likewise.
629 (_mm512_fcmadd_round_pch): Likewise.
630 (_mm512_mask_fcmadd_round_pch): Likewise.
631 (_mm512_mask3_fcmadd_round_pch): Likewise.
632 (_mm512_maskz_fcmadd_round_pch): Likewise.
633 (_mm512_fmadd_round_pch): Likewise.
634 (_mm512_mask_fmadd_round_pch): Likewise.
635 (_mm512_mask3_fmadd_round_pch): Likewise.
636 (_mm512_maskz_fmadd_round_pch): Likewise.
637 (_mm512_fcmul_pch): Likewise.
638 (_mm512_mask_fcmul_pch): Likewise.
639 (_mm512_maskz_fcmul_pch): Likewise.
640 (_mm512_fmul_pch): Likewise.
641 (_mm512_mask_fmul_pch): Likewise.
642 (_mm512_maskz_fmul_pch): Likewise.
643 (_mm512_fcmul_round_pch): Likewise.
644 (_mm512_mask_fcmul_round_pch): Likewise.
645 (_mm512_maskz_fcmul_round_pch): Likewise.
646 (_mm512_fmul_round_pch): Likewise.
647 (_mm512_mask_fmul_round_pch): Likewise.
648 (_mm512_maskz_fmul_round_pch): Likewise.
649 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
651 (_mm_mask_fmadd_pch): Likewise.
652 (_mm_mask3_fmadd_pch): Likewise.
653 (_mm_maskz_fmadd_pch): Likewise.
654 (_mm256_fmadd_pch): Likewise.
655 (_mm256_mask_fmadd_pch): Likewise.
656 (_mm256_mask3_fmadd_pch): Likewise.
657 (_mm256_maskz_fmadd_pch): Likewise.
658 (_mm_fcmadd_pch): Likewise.
659 (_mm_mask_fcmadd_pch): Likewise.
660 (_mm_mask3_fcmadd_pch): Likewise.
661 (_mm_maskz_fcmadd_pch): Likewise.
662 (_mm256_fcmadd_pch): Likewise.
663 (_mm256_mask_fcmadd_pch): Likewise.
664 (_mm256_mask3_fcmadd_pch): Likewise.
665 (_mm256_maskz_fcmadd_pch): Likewise.
666 (_mm_fmul_pch): Likewise.
667 (_mm_mask_fmul_pch): Likewise.
668 (_mm_maskz_fmul_pch): Likewise.
669 (_mm256_fmul_pch): Likewise.
670 (_mm256_mask_fmul_pch): Likewise.
671 (_mm256_maskz_fmul_pch): Likewise.
672 (_mm_fcmul_pch): Likewise.
673 (_mm_mask_fcmul_pch): Likewise.
674 (_mm_maskz_fcmul_pch): Likewise.
675 (_mm256_fcmul_pch): Likewise.
676 (_mm256_mask_fcmul_pch): Likewise.
677 (_mm256_maskz_fcmul_pch): Likewise.
678 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
679 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
680 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
681 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
682 * config/i386/i386-builtin.def: Add new builtins.
683 * config/i386/i386-expand.c: Handle new builtin types.
684 * config/i386/subst.md (SUBST_CV): New.
686 (maskc_operand3): Ditto.
688 (sdc_maskz_name): Ditto.
689 (sdc_mask_op4): Ditto.
690 (sdc_mask_op5): Ditto.
691 (sdc_mask_mode512bit_condition): Ditto.
693 (round_maskc_operand3): Ditto.
694 (round_sdc_mask_operand4): Ditto.
695 (round_maskc_op3): Ditto.
696 (round_sdc_mask_op4): Ditto.
697 (round_saeonly_sdc_mask_operand5): Ditto.
698 * config/i386/sse.md (unspec): Add complex fma unspecs.
699 (avx512fmaskcmode): New.
700 (UNSPEC_COMPLEX_F_C_MA): Ditto.
701 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
702 (complexopname): Ditto.
703 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
704 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
705 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
707 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
708 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
710 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
712 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
713 rs6000-density-size-threshold, rs6000-density-penalty,
714 rs6000-density-load-pct-threshold,
715 rs6000-density-load-num-threshold): New parameter.
716 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
717 corresponding parameters.
719 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
721 * gimple-range-path.cc (path_range_query::defined_outside_path):
723 (path_range_query::range_on_path_entry): New.
724 (path_range_query::internal_range_of_expr): Resolve unknowns
726 (path_range_query::improve_range_with_equivs): New.
727 (path_range_query::ssa_range_in_phi): Resolve unknowns with
729 * gimple-range-path.h (class path_range_query): Add
730 defined_outside_path, range_on_path_entry, and
731 improve_range_with_equivs.
733 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
735 * gimple-range-path.cc (path_range_query::add_to_imports): New.
736 (path_range_query::add_copies_to_imports): New.
737 (path_range_query::precompute_ranges): Call
738 add_copies_to_imports.
739 * gimple-range-path.h (class path_range_query): Add prototypes
740 for add_copies_to_imports and add_to_imports.
742 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
744 * gimple-range-path.cc (path_range_query::range_defined_in_block):
747 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
749 * gimple-range-fold.h (class fur_source): Make oracle protected.
750 * gimple-range-path.cc (path_range_query::path_range_query): Add
751 resolve argument. Initialize oracle.
752 (path_range_query::~path_range_query): Delete oracle.
753 (path_range_query::range_of_stmt): Adapt to use relations.
754 (path_range_query::precompute_ranges): Pre-compute relations.
755 (class jt_fur_source): New
756 (jt_fur_source::jt_fur_source): New.
757 (jt_fur_source::register_relation): New.
758 (jt_fur_source::query_relation): New.
759 (path_range_query::precompute_relations): New.
760 (path_range_query::precompute_phi_relations): New.
761 * gimple-range-path.h (path_range_query): Add resolve argument.
762 Add oracle, precompute_relations, precompute_phi_relations.
763 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
764 resolve argument to solver.
766 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
768 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
769 Rename postfold_gcond_edges to register_outgoing_edges and
771 (fold_using_range::postfold_gcond_edges): Rename...
772 (fur_source::register_outgoing_edges): ...to this.
773 * gimple-range-fold.h (postfold_gcond_edges): Rename to
774 register_outgoing_edges and move to fur_source.
776 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
778 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
779 dom_info_available_p.
781 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
783 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
784 and quick_grow_cleared instead of safe_grow_cleared.
786 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
789 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
792 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
794 * configure.ac: Detect when the assembler supports new-style
795 architecture extensions.
796 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
797 the full CPU string if the assembler can grok it.
798 (arm_rewrite_march): Likewise but for the architecture.
799 * config.in: Regenerate.
800 * configure: Regenerate.
802 2021-09-21 Richard Biener <rguenther@suse.de>
804 PR tree-optimization/102421
805 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
806 adjust alignment info.
808 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
810 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
813 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
815 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
816 undefined edges, apply an equivalence if appropriate.
817 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
818 UNDEFINED if EDGE_EXECUTABLE is not set.
819 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
820 as EXECUTABLE upon startup.
821 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
823 * vr-values.c (set_and_propagate_unexecutable): New.
824 (simplify_using_ranges::fold_cond): Call set_and_propagate.
825 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
826 * vr-values.h: Add prototype.
828 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
830 * value-relation.cc (equiv_oracle::register_initial_def): New.
831 (equiv_oracle::register_relation): Call register_initial_def.
832 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
833 (relation_oracle::register_stmt): Check def block of PHI arguments.
834 * value-relation.h (equiv_oracle): Add new prototypes.
836 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
838 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
839 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
840 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
841 __ROUNDING_MATH__ according to their corresponding flags.
842 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
843 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
844 and __ROUNDING_MATH__.
846 2021-09-20 Richard Biener <rguenther@suse.de>
848 * tree-vect-stmts.c (vectorizable_load): Use the vectype
851 2021-09-20 Richard Biener <rguenther@suse.de>
853 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
854 Do not compute alignment of the vectorized access here.
856 2021-09-20 Richard Biener <rguenther@suse.de>
858 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
859 Store -1 for runtime alias peeling iterations.
861 2021-09-20 Richard Biener <rguenther@suse.de>
863 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
865 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
867 * input.c (string_concat_db::record_string_concatenation)
868 (string_concat_db::get_string_concatenation): Skip for
869 'RESERVED_LOCATION_P'.
871 2021-09-20 Richard Biener <rguenther@suse.de>
873 PR tree-optimization/65206
874 * tree-data-ref.h (struct data_reference): Add alt_indices,
876 * tree-data-ref.c (free_data_ref): Release alt_indices.
877 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
878 (create_data_ref): Adjust.
879 (initialize_data_dependence_relation): Split into head
880 and tail. When the base objects fail to match up try
881 again with pointer-based analysis of indices.
882 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
883 not compare the lazily computed alternate set of indices.
885 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
887 * gcc.c: Test for execute OK when we find the
888 programs for assembler linker and dsymutil and those
889 were specified at configure-time.
891 2021-09-19 Martin Sebor <msebor@redhat.com>
894 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
895 Correct a function pre/postcondition.
897 2021-09-19 Martin Sebor <msebor@redhat.com>
900 * tree-ssa-strlen.c (get_range): Handle null cfun.
902 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
904 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
905 unwinder shim as a convenience library.
907 2021-09-19 Andrew Pinski <apinski@marvell.com>
909 * doc/install.texi: Add note about
910 binutils 2.35 is required for LTO usage.
912 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
914 * tree-ssa-threadbackward.c
915 (back_threader_registry::register_path): Use push_edge.
916 * tree-ssa-threadedge.c
917 (jump_threader::thread_around_empty_blocks): Same.
918 (jump_threader::thread_through_normal_block): Same.
919 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
921 * tree-ssa-threadupdate.c
922 (jt_path_registry::allocate_thread_edge): Remove.
923 (jt_path_registry::push_edge): New.
924 (dump_jump_thread_path): Make static.
925 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
928 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
930 * gimple-range-path.cc (path_range_query::path_range_query): Add
932 (path_range_query::dump): Remove extern declaration of dump_ranger.
933 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
934 * gimple-range-trace.h (dump_ranger): Add prototype.
936 2021-09-19 John Ericson <git@JohnEricson.me>
938 * gcc.c (find_a_program): New function, factored out of...
940 (execute): Use find_a_program when looking for programs rather
943 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
945 * config/avr/avr-mcus.def: Add atmega324pb.
946 * doc/avr-mmcu.texi: Corresponding changes.
948 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
951 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
952 the same transformations as to @1. For comparisons against NaN,
953 don't check HONOR_SNANS but confirm that neither operand is a
956 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
958 * attribs.c (make_unique_name): Delete.
959 * attribs.h (make_unique_name): Delete.
961 2021-09-19 Andrew Pinski <apinski@marvell.com>
963 * lra-constraints.c (check_and_process_move): Assert
964 that dclass and sclass are greater than or equal to NO_REGS.
966 2021-09-18 Jakub Jelinek <jakub@redhat.com>
968 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
969 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
970 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
972 2021-09-18 liuhongt <hongtao.liu@intel.com>
974 * config/i386/i386-features.c (remove_partial_avx_dependency):
975 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
976 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
978 2021-09-18 Jakub Jelinek <jakub@redhat.com>
980 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
981 if file/namespace scope variable doesn't have predetermined sharing,
982 treat it as if there was default(none).
984 2021-09-18 liuhongt <hongtao.liu@intel.com>
986 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
988 (_mm_mask_fmadd_sh): Likewise.
989 (_mm_mask3_fmadd_sh): Likewise.
990 (_mm_maskz_fmadd_sh): Likewise.
991 (_mm_fmadd_round_sh): Likewise.
992 (_mm_mask_fmadd_round_sh): Likewise.
993 (_mm_mask3_fmadd_round_sh): Likewise.
994 (_mm_maskz_fmadd_round_sh): Likewise.
995 (_mm_fnmadd_sh): Likewise.
996 (_mm_mask_fnmadd_sh): Likewise.
997 (_mm_mask3_fnmadd_sh): Likewise.
998 (_mm_maskz_fnmadd_sh): Likewise.
999 (_mm_fnmadd_round_sh): Likewise.
1000 (_mm_mask_fnmadd_round_sh): Likewise.
1001 (_mm_mask3_fnmadd_round_sh): Likewise.
1002 (_mm_maskz_fnmadd_round_sh): Likewise.
1003 (_mm_fmsub_sh): Likewise.
1004 (_mm_mask_fmsub_sh): Likewise.
1005 (_mm_mask3_fmsub_sh): Likewise.
1006 (_mm_maskz_fmsub_sh): Likewise.
1007 (_mm_fmsub_round_sh): Likewise.
1008 (_mm_mask_fmsub_round_sh): Likewise.
1009 (_mm_mask3_fmsub_round_sh): Likewise.
1010 (_mm_maskz_fmsub_round_sh): Likewise.
1011 (_mm_fnmsub_sh): Likewise.
1012 (_mm_mask_fnmsub_sh): Likewise.
1013 (_mm_mask3_fnmsub_sh): Likewise.
1014 (_mm_maskz_fnmsub_sh): Likewise.
1015 (_mm_fnmsub_round_sh): Likewise.
1016 (_mm_mask_fnmsub_round_sh): Likewise.
1017 (_mm_mask3_fnmsub_round_sh): Likewise.
1018 (_mm_maskz_fnmsub_round_sh): Likewise.
1019 * config/i386/i386-builtin-types.def
1020 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1021 * config/i386/i386-builtin.def: Add new builtins.
1022 * config/i386/i386-expand.c: Handle new builtin type.
1023 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1024 Ajdust to support FP16.
1025 (fmai_vmfmsub_<mode><round_name>): Ditto.
1026 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1027 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1028 (*fmai_fmadd_<mode>): Ditto.
1029 (*fmai_fmsub_<mode>): Ditto.
1030 (*fmai_fnmadd_<mode><round_name>): Ditto.
1031 (*fmai_fnmsub_<mode><round_name>): Ditto.
1032 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1033 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1034 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1035 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1036 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1037 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1038 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1039 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1040 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1041 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1042 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1043 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1044 adjust to support FP16.
1045 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1046 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1047 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1050 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1052 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1054 (maskload<mode><avx512fmaskmodelower>): Ditto.
1055 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1057 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1059 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1061 (ix86_expand_copysign): Ditto.
1062 (ix86_expand_xorsign): Ditto.
1063 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1065 (ix86_build_signbit_mask): Ditto.
1066 (ix86_can_change_mode_class): Ditto.
1067 * config/i386/i386.md
1068 (SSEMODEF): Add HFmode.
1069 (ssevecmodef): Ditto.
1070 (<code>hf2): New define_expand.
1071 (*<code>hf2_1): New define_insn_and_split.
1072 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1073 (xorsign<mode>): Ditto.
1074 * config/i386/sse.md (VFB): New mode iterator.
1075 (VFB_128_256): Ditto.
1077 (sseintvecmode2): Support HF vector mode.
1078 (<code><mode>2): Use new mode iterator.
1079 (*<code><mode>2): Ditto.
1080 (copysign<mode>3): Ditto.
1081 (xorsign<mode>3): Ditto.
1082 (<code><mode>3<mask_name>): Ditto.
1083 (<code><mode>3<mask_name>): Ditto.
1084 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1085 (<sse>_andnot<mode>3<mask_name>): Ditto.
1086 (*<code><mode>3<mask_name>): Ditto.
1087 (*<code><mode>3<mask_name>): Ditto.
1089 2021-09-18 liuhongt <hongtao.liu@intel.com>
1091 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1093 (_mm512_mask3_fmadd_ph): Likewise.
1094 (_mm512_maskz_fmadd_ph): Likewise.
1095 (_mm512_fmadd_round_ph): Likewise.
1096 (_mm512_mask_fmadd_round_ph): Likewise.
1097 (_mm512_mask3_fmadd_round_ph): Likewise.
1098 (_mm512_maskz_fmadd_round_ph): Likewise.
1099 (_mm512_fnmadd_ph): Likewise.
1100 (_mm512_mask_fnmadd_ph): Likewise.
1101 (_mm512_mask3_fnmadd_ph): Likewise.
1102 (_mm512_maskz_fnmadd_ph): Likewise.
1103 (_mm512_fnmadd_round_ph): Likewise.
1104 (_mm512_mask_fnmadd_round_ph): Likewise.
1105 (_mm512_mask3_fnmadd_round_ph): Likewise.
1106 (_mm512_maskz_fnmadd_round_ph): Likewise.
1107 (_mm512_fmsub_ph): Likewise.
1108 (_mm512_mask_fmsub_ph): Likewise.
1109 (_mm512_mask3_fmsub_ph): Likewise.
1110 (_mm512_maskz_fmsub_ph): Likewise.
1111 (_mm512_fmsub_round_ph): Likewise.
1112 (_mm512_mask_fmsub_round_ph): Likewise.
1113 (_mm512_mask3_fmsub_round_ph): Likewise.
1114 (_mm512_maskz_fmsub_round_ph): Likewise.
1115 (_mm512_fnmsub_ph): Likewise.
1116 (_mm512_mask_fnmsub_ph): Likewise.
1117 (_mm512_mask3_fnmsub_ph): Likewise.
1118 (_mm512_maskz_fnmsub_ph): Likewise.
1119 (_mm512_fnmsub_round_ph): Likewise.
1120 (_mm512_mask_fnmsub_round_ph): Likewise.
1121 (_mm512_mask3_fnmsub_round_ph): Likewise.
1122 (_mm512_maskz_fnmsub_round_ph): Likewise.
1123 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1125 (_mm256_mask_fmadd_ph): Likewise.
1126 (_mm256_mask3_fmadd_ph): Likewise.
1127 (_mm256_maskz_fmadd_ph): Likewise.
1128 (_mm_fmadd_ph): Likewise.
1129 (_mm_mask_fmadd_ph): Likewise.
1130 (_mm_mask3_fmadd_ph): Likewise.
1131 (_mm_maskz_fmadd_ph): Likewise.
1132 (_mm256_fnmadd_ph): Likewise.
1133 (_mm256_mask_fnmadd_ph): Likewise.
1134 (_mm256_mask3_fnmadd_ph): Likewise.
1135 (_mm256_maskz_fnmadd_ph): Likewise.
1136 (_mm_fnmadd_ph): Likewise.
1137 (_mm_mask_fnmadd_ph): Likewise.
1138 (_mm_mask3_fnmadd_ph): Likewise.
1139 (_mm_maskz_fnmadd_ph): Likewise.
1140 (_mm256_fmsub_ph): Likewise.
1141 (_mm256_mask_fmsub_ph): Likewise.
1142 (_mm256_mask3_fmsub_ph): Likewise.
1143 (_mm256_maskz_fmsub_ph): Likewise.
1144 (_mm_fmsub_ph): Likewise.
1145 (_mm_mask_fmsub_ph): Likewise.
1146 (_mm_mask3_fmsub_ph): Likewise.
1147 (_mm_maskz_fmsub_ph): Likewise.
1148 (_mm256_fnmsub_ph): Likewise.
1149 (_mm256_mask_fnmsub_ph): Likewise.
1150 (_mm256_mask3_fnmsub_ph): Likewise.
1151 (_mm256_maskz_fnmsub_ph): Likewise.
1152 (_mm_fnmsub_ph): Likewise.
1153 (_mm_mask_fnmsub_ph): Likewise.
1154 (_mm_mask3_fnmsub_ph): Likewise.
1155 (_mm_maskz_fnmsub_ph): Likewise.
1156 * config/i386/i386-builtin.def: Add corresponding new builtins.
1157 * config/i386/sse.md
1158 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1159 support HF vector modes.
1160 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1162 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1163 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1164 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1165 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1166 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1167 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1168 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1170 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1171 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1172 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1173 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1174 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1175 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1177 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1178 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1179 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1180 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1181 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1182 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1183 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1185 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1186 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1187 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1188 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1189 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1191 2021-09-18 liuhongt <hongtao.liu@intel.com>
1193 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1195 (_mm512_mask_fmaddsub_ph): Likewise.
1196 (_mm512_mask3_fmaddsub_ph): Likewise.
1197 (_mm512_maskz_fmaddsub_ph): Likewise.
1198 (_mm512_fmaddsub_round_ph): Likewise.
1199 (_mm512_mask_fmaddsub_round_ph): Likewise.
1200 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1201 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1202 (_mm512_mask_fmsubadd_ph): Likewise.
1203 (_mm512_mask3_fmsubadd_ph): Likewise.
1204 (_mm512_maskz_fmsubadd_ph): Likewise.
1205 (_mm512_fmsubadd_round_ph): Likewise.
1206 (_mm512_mask_fmsubadd_round_ph): Likewise.
1207 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1208 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1209 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1211 (_mm256_mask_fmaddsub_ph): Likewise.
1212 (_mm256_mask3_fmaddsub_ph): Likewise.
1213 (_mm256_maskz_fmaddsub_ph): Likewise.
1214 (_mm_fmaddsub_ph): Likewise.
1215 (_mm_mask_fmaddsub_ph): Likewise.
1216 (_mm_mask3_fmaddsub_ph): Likewise.
1217 (_mm_maskz_fmaddsub_ph): Likewise.
1218 (_mm256_fmsubadd_ph): Likewise.
1219 (_mm256_mask_fmsubadd_ph): Likewise.
1220 (_mm256_mask3_fmsubadd_ph): Likewise.
1221 (_mm256_maskz_fmsubadd_ph): Likewise.
1222 (_mm_fmsubadd_ph): Likewise.
1223 (_mm_mask_fmsubadd_ph): Likewise.
1224 (_mm_mask3_fmsubadd_ph): Likewise.
1225 (_mm_maskz_fmsubadd_ph): Likewise.
1226 * config/i386/i386-builtin.def: Add corresponding new builtins.
1227 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1228 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1229 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1231 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1233 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1234 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1235 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1237 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1238 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1240 2021-09-18 liuhongt <hongtao.liu@intel.com>
1243 * config/i386/i386.c (ix86_print_operand): Handle
1244 V8HF/V16HF/V32HFmode.
1245 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1246 * config/i386/sse.md (avx512bcst): Remove.
1248 2021-09-17 Martin Sebor <msebor@redhat.com>
1250 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1251 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1252 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1253 (check_defs): Add comment.
1254 (can_skip_redundant_opnd): Update comment.
1255 (compute_uninit_opnds_pos): Adjust to namespace change.
1256 (find_pdom): Move to gimple-predicate-analysis.cc.
1258 (struct uninit_undef_val_t): New.
1259 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1260 (find_control_equiv_block): Same.
1261 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1262 (MAX_SWITCH_CASES): Same.
1263 (compute_control_dep_chain): Same.
1264 (find_uninit_use): Use predicate analyzer.
1265 (struct pred_info): Move to gimple-predicate-analysis.
1266 (convert_control_dep_chain_into_preds): Same.
1267 (find_predicates): Same.
1268 (collect_phi_def_edges): Same.
1269 (warn_uninitialized_phi): Use predicate analyzer.
1270 (find_def_preds): Move to gimple-predicate-analysis.
1271 (dump_pred_info): Same.
1272 (dump_pred_chain): Same.
1273 (dump_predicates): Same.
1274 (destroy_predicate_vecs): Remove.
1275 (execute_late_warn_uninitialized): New.
1276 (get_cmp_code): Move to gimple-predicate-analysis.
1277 (is_value_included_in): Same.
1278 (value_sat_pred_p): Same.
1279 (find_matching_predicate_in_rest_chains): Same.
1280 (is_use_properly_guarded): Same.
1281 (prune_uninit_phi_opnds): Same.
1282 (find_var_cmp_const): Same.
1283 (use_pred_not_overlap_with_undef_path_pred): Same.
1284 (pred_equal_p): Same.
1285 (is_neq_relop_p): Same.
1286 (is_neq_zero_form_p): Same.
1287 (pred_expr_equal_p): Same.
1288 (is_pred_expr_subset_of): Same.
1289 (is_pred_chain_subset_of): Same.
1290 (is_included_in): Same.
1291 (is_superset_of): Same.
1293 (simplify_pred): Same.
1294 (simplify_preds_2): Same.
1295 (simplify_preds_3): Same.
1296 (simplify_preds_4): Same.
1297 (simplify_preds): Same.
1299 (push_to_worklist): Same.
1300 (get_pred_info_from_cmp): Same.
1301 (is_degenerated_phi): Same.
1302 (normalize_one_pred_1): Same.
1303 (normalize_one_pred): Same.
1304 (normalize_one_pred_chain): Same.
1305 (normalize_preds): Same.
1306 (can_one_predicate_be_invalidated_p): Same.
1307 (can_chain_union_be_invalidated_p): Same.
1308 (uninit_uses_cannot_happen): Same.
1309 (pass_late_warn_uninitialized::execute): Define.
1310 * gimple-predicate-analysis.cc: New file.
1311 * gimple-predicate-analysis.h: New file.
1313 2021-09-17 Julian Brown <julian@codesourcery.com>
1315 * config/gcn/gcn.c (gimple.h): Include.
1316 (gcn_fork_join): Emit barrier for worker-level joins.
1317 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1318 writes_gang_private bitmap parameter. Set bit for blocks
1319 containing gang-private variable writes.
1320 (worker_single_simple): Don't emit barrier after predicated block.
1321 (worker_single_copy): Don't emit barrier if we're not broadcasting
1322 anything and the block contains no gang-private writes.
1323 (neuter_worker_single): Don't predicate blocks that only contain
1324 NOPs or internal marker functions. Pass has_gang_private_write
1325 argument to worker_single_copy.
1326 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1328 2021-09-17 Julian Brown <julian@codesourcery.com>
1330 * config/gcn/gcn-protos.h
1331 (gcn_goacc_create_worker_broadcast_record): Update prototype.
1332 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
1333 preallocated block of LDS memory. Do not cache/share decls for
1334 reduction temporaries between invocations.
1335 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
1336 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
1337 and return temporary LDS space at that offset. Return pointer in
1339 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
1341 (ACC_LDS_SIZE): Define as acc_lds_size.
1342 (gcn_init_machine_status): Don't initialise lds_allocated,
1343 lds_allocs, reduc_decls fields of machine function struct.
1344 (gcn_option_override): Handle default size for gang-private
1345 variables and -mgang-private-size option.
1346 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
1347 initialising M0_REG.
1348 (gcn_shared_mem_layout): New function.
1349 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
1350 gang_private_hwm variable.
1351 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
1352 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
1353 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
1354 * config/gcn/gcn.opt (gang_private_size_opt): New global.
1355 (mgang-private-size=): New option.
1356 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
1358 * doc/tm.texi: Regenerate.
1359 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
1361 (build_sender_ref): Handle sender_decl being pointer.
1362 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
1363 parameters. Pass placement argument to
1364 create_worker_broadcast_record hook invocations. Handle
1365 sender_decl being pointer and isolate_broadcasts inserting extra
1367 (blk_offset_map_t): Add typedef.
1368 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
1369 preallocated range to worker_single_copy call.
1370 (dfs_broadcast_reachable_1): New function.
1371 (idx_decl_pair_t, used_range_vec_t): New typedefs.
1372 (sort_size_descending): New function.
1373 (addr_range): New class.
1374 (splay_tree_compare_addr_range, splay_tree_free_key)
1375 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
1376 (execute_omp_oacc_neuter_broadcast): Rename to...
1377 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
1378 parameters. Arrange layout of shared memory for broadcast
1380 (execute_omp_oacc_neuter_broadcast): New function.
1381 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
1382 handling from here. Enable pass for all OpenACC routines in order
1383 to call shared memory-layout hook.
1384 * target.def (create_worker_broadcast_record): Add OFFSET
1386 (shared_mem_layout): New hook.
1388 2021-09-17 Julian Brown <julian@codesourcery.com>
1389 Thomas Schwinge <thomas@codesourcery.com>
1391 * omp-oacc-neuter-broadcast.cc
1392 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
1394 (execute_omp_oacc_neuter_broadcast): Adjust.
1396 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1398 * value-relation.cc (class equiv_chain): Move to header file.
1399 (path_oracle::path_oracle): New.
1400 (path_oracle::~path_oracle): New.
1401 (path_oracle::register_relation): New.
1402 (path_oracle::query_relation): New.
1403 (path_oracle::reset_path): New.
1404 (path_oracle::dump): New.
1405 * value-relation.h (class equiv_chain): Move to here.
1406 (class path_oracle): New.
1408 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1410 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
1412 * gimple-range-fold.cc (fur_depend::register_relation): Use
1413 register_stmt/edge routines.
1414 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
1415 (equiv_oracle::equiv_oracle): Create self equivalence cache.
1416 (equiv_oracle::~equiv_oracle): Release same.
1417 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
1418 are no equivalences.
1419 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
1420 (equiv_oracle::register_relation): Rename from register_equiv.
1421 (relation_chain_head::find_relation): Relocate from dom_oracle.
1422 (relation_oracle::register_stmt): New.
1423 (relation_oracle::register_edge): New.
1424 (dom_oracle::*): Rename from relation_oracle.
1425 (dom_oracle::register_relation): Adjust to call equiv_oracle.
1426 (dom_oracle::set_one_relation): Split from register_relation.
1427 (dom_oracle::register_transitives): Consolidate 2 methods.
1428 (dom_oracle::find_relation_block): Move core to relation_chain.
1429 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
1430 * value-relation.h (class relation_oracle): New pure virtual base.
1431 (class equiv_oracle): Inherit from relation_oracle and adjust.
1432 (class dom_oracle): Rename from old relation_oracle and adjust.
1434 2021-09-17 Martin Sebor <msebor@redhat.com>
1436 PR middle-end/102200
1437 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
1438 (handle_min_max_size): Change argument. Store original SSA_NAME for
1439 operands to potentially distinct (sub)objects.
1440 (compute_objsize_r): Adjust call to the above.
1442 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1444 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
1445 (rs6000_new_builtin_vectorized_function): New function.
1446 (rs6000_new_builtin_md_vectorized_function): Likewise.
1447 (rs6000_builtin_vectorized_function): Call
1448 rs6000_new_builtin_vectorized_function.
1449 (rs6000_builtin_md_vectorized_function): Call
1450 rs6000_new_builtin_md_vectorized_function.
1452 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1454 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
1455 (ASSEMBLE_PAIR): Likewise.
1456 (BUILD_ACC): Likewise.
1457 (DISASSEMBLE_ACC): Likewise.
1458 (DISASSEMBLE_PAIR): Likewise.
1459 (PMXVBF16GER2): Likewise.
1460 (PMXVBF16GER2NN): Likewise.
1461 (PMXVBF16GER2NP): Likewise.
1462 (PMXVBF16GER2PN): Likewise.
1463 (PMXVBF16GER2PP): Likewise.
1464 (PMXVF16GER2): Likewise.
1465 (PMXVF16GER2NN): Likewise.
1466 (PMXVF16GER2NP): Likewise.
1467 (PMXVF16GER2PN): Likewise.
1468 (PMXVF16GER2PP): Likewise.
1469 (PMXVF32GER): Likewise.
1470 (PMXVF32GERNN): Likewise.
1471 (PMXVF32GERNP): Likewise.
1472 (PMXVF32GERPN): Likewise.
1473 (PMXVF32GERPP): Likewise.
1474 (PMXVF64GER): Likewise.
1475 (PMXVF64GERNN): Likewise.
1476 (PMXVF64GERNP): Likewise.
1477 (PMXVF64GERPN): Likewise.
1478 (PMXVF64GERPP): Likewise.
1479 (PMXVI16GER2): Likewise.
1480 (PMXVI16GER2PP): Likewise.
1481 (PMXVI16GER2S): Likewise.
1482 (PMXVI16GER2SPP): Likewise.
1483 (PMXVI4GER8): Likewise.
1484 (PMXVI4GER8PP): Likewise.
1485 (PMXVI8GER4): Likewise.
1486 (PMXVI8GER4PP): Likewise.
1487 (PMXVI8GER4SPP): Likewise.
1488 (XVBF16GER2): Likewise.
1489 (XVBF16GER2NN): Likewise.
1490 (XVBF16GER2NP): Likewise.
1491 (XVBF16GER2PN): Likewise.
1492 (XVBF16GER2PP): Likewise.
1493 (XVF16GER2): Likewise.
1494 (XVF16GER2NN): Likewise.
1495 (XVF16GER2NP): Likewise.
1496 (XVF16GER2PN): Likewise.
1497 (XVF16GER2PP): Likewise.
1498 (XVF32GER): Likewise.
1499 (XVF32GERNN): Likewise.
1500 (XVF32GERNP): Likewise.
1501 (XVF32GERPN): Likewise.
1502 (XVF32GERPP): Likewise.
1503 (XVF64GER): Likewise.
1504 (XVF64GERNN): Likewise.
1505 (XVF64GERNP): Likewise.
1506 (XVF64GERPN): Likewise.
1507 (XVF64GERPP): Likewise.
1508 (XVI16GER2): Likewise.
1509 (XVI16GER2PP): Likewise.
1510 (XVI16GER2S): Likewise.
1511 (XVI16GER2SPP): Likewise.
1512 (XVI4GER8): Likewise.
1513 (XVI4GER8PP): Likewise.
1514 (XVI8GER4): Likewise.
1515 (XVI8GER4PP): Likewise.
1516 (XVI8GER4SPP): Likewise.
1517 (XXMFACC): Likewise.
1518 (XXMTACC): Likewise.
1519 (XXSETACCZ): Likewise.
1520 (ASSEMBLE_PAIR_V): Likewise.
1521 (BUILD_PAIR): Likewise.
1522 (DISASSEMBLE_PAIR_V): Likewise.
1525 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
1526 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
1527 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
1528 (parse_bif_attrs): Handle ismmaint.
1529 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
1530 (write_bif_static_init): Handle ismmaint.
1532 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1534 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
1536 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
1537 (rs6000_new_builtin_valid_without_lhs): New function.
1538 (rs6000_gimple_fold_new_mma_builtin): Likewise.
1539 (rs6000_gimple_fold_new_builtin): Likewise.
1541 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
1543 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
1544 Destruct stale Value objects.
1545 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
1548 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
1551 * match.pd (shift optimizations): Disable recent sign-changing
1552 optimization for shifts by zero, these will be folded later.
1554 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1556 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
1557 [power9] to [always].
1559 2021-09-17 Richard Biener <rguenther@suse.de>
1561 * tree-vect-stmts.c (vectorizable_load): Do not frob
1564 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1566 * config/i386/i386-features.c (remove_partial_avx_dependency):
1567 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
1568 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
1570 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
1572 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1573 * config/i386/i386.md (SSE FP to FP splitters): Replace
1574 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
1575 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
1576 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
1577 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
1578 * config/i386/x86-tune.def
1579 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
1580 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1582 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1585 * config/i386/i386-features.c (remove_partial_avx_dependency):
1586 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
1587 before generating vxorps.
1589 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1591 * config/i386/i386-options.c (processor_cost_table): Use
1592 tremont_cost for Tremont.
1593 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
1594 (tremont_memset): Likewise.
1595 (tremont_cost): Likewise.
1596 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
1599 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1601 * common/config/i386/i386-common.c: Use Haswell scheduling model
1603 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
1605 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
1607 (ix86_adjust_cost): Handle Tremont.
1608 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
1610 (X86_TUNE_USE_LEAVE): Likewise.
1611 (X86_TUNE_PUSH_MEMORY): Likewise.
1612 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
1613 (X86_TUNE_USE_CLTD): Likewise.
1614 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
1615 (X86_TUNE_AVOID_MFENCE): Likewise.
1616 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
1617 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
1618 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
1619 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
1620 (X86_TUNE_OPT_AGU): Likewise.
1621 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
1622 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
1623 (X86_TUNE_EXPAND_ABS): Likewise.
1624 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
1625 (X86_TUNE_SLOW_PSHUFB): Likewise.
1627 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
1629 PR rtl-optimization/102306
1630 * combine.c (try_combine): Abort the combination if we are about to
1631 duplicate volatile references.
1633 2021-09-17 liuhongt <hongtao.liu@intel.com>
1635 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
1637 (_mm256_undefined_ph): Likewise.
1638 (_mm512_undefined_ph): Likewise.
1639 (_mm_cvtsh_h): Likewise.
1640 (_mm256_cvtsh_h): Likewise.
1641 (_mm512_cvtsh_h): Likewise.
1642 (_mm512_castph_ps): Likewise.
1643 (_mm512_castph_pd): Likewise.
1644 (_mm512_castph_si512): Likewise.
1645 (_mm512_castph512_ph128): Likewise.
1646 (_mm512_castph512_ph256): Likewise.
1647 (_mm512_castph128_ph512): Likewise.
1648 (_mm512_castph256_ph512): Likewise.
1649 (_mm512_zextph128_ph512): Likewise.
1650 (_mm512_zextph256_ph512): Likewise.
1651 (_mm512_castps_ph): Likewise.
1652 (_mm512_castpd_ph): Likewise.
1653 (_mm512_castsi512_ph): Likewise.
1654 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
1656 (_mm256_castph_ps): Likewise.
1657 (_mm_castph_pd): Likewise.
1658 (_mm256_castph_pd): Likewise.
1659 (_mm_castph_si128): Likewise.
1660 (_mm256_castph_si256): Likewise.
1661 (_mm_castps_ph): Likewise.
1662 (_mm256_castps_ph): Likewise.
1663 (_mm_castpd_ph): Likewise.
1664 (_mm256_castpd_ph): Likewise.
1665 (_mm_castsi128_ph): Likewise.
1666 (_mm256_castsi256_ph): Likewise.
1667 (_mm256_castph256_ph128): Likewise.
1668 (_mm256_castph128_ph256): Likewise.
1669 (_mm256_zextph128_ph256): Likewise.
1671 2021-09-17 liuhongt <hongtao.liu@intel.com>
1673 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
1675 (_mm_mask_cvtsh_ss): Likewise.
1676 (_mm_maskz_cvtsh_ss): Likewise.
1677 (_mm_cvtsh_sd): Likewise.
1678 (_mm_mask_cvtsh_sd): Likewise.
1679 (_mm_maskz_cvtsh_sd): Likewise.
1680 (_mm_cvt_roundsh_ss): Likewise.
1681 (_mm_mask_cvt_roundsh_ss): Likewise.
1682 (_mm_maskz_cvt_roundsh_ss): Likewise.
1683 (_mm_cvt_roundsh_sd): Likewise.
1684 (_mm_mask_cvt_roundsh_sd): Likewise.
1685 (_mm_maskz_cvt_roundsh_sd): Likewise.
1686 (_mm_cvtss_sh): Likewise.
1687 (_mm_mask_cvtss_sh): Likewise.
1688 (_mm_maskz_cvtss_sh): Likewise.
1689 (_mm_cvtsd_sh): Likewise.
1690 (_mm_mask_cvtsd_sh): Likewise.
1691 (_mm_maskz_cvtsd_sh): Likewise.
1692 (_mm_cvt_roundss_sh): Likewise.
1693 (_mm_mask_cvt_roundss_sh): Likewise.
1694 (_mm_maskz_cvt_roundss_sh): Likewise.
1695 (_mm_cvt_roundsd_sh): Likewise.
1696 (_mm_mask_cvt_roundsd_sh): Likewise.
1697 (_mm_maskz_cvt_roundsd_sh): Likewise.
1698 * config/i386/i386-builtin-types.def
1699 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
1700 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
1701 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
1702 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
1703 * config/i386/i386-builtin.def: Add corrresponding new builtins.
1704 * config/i386/i386-expand.c: Handle new builtin types.
1705 * config/i386/sse.md (VF48_128): New mode iterator.
1706 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
1708 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
1711 2021-09-17 liuhongt <hongtao.liu@intel.com>
1713 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
1715 (_mm512_mask_cvtph_pd): Likewise.
1716 (_mm512_maskz_cvtph_pd): Likewise.
1717 (_mm512_cvt_roundph_pd): Likewise.
1718 (_mm512_mask_cvt_roundph_pd): Likewise.
1719 (_mm512_maskz_cvt_roundph_pd): Likewise.
1720 (_mm512_cvtxph_ps): Likewise.
1721 (_mm512_mask_cvtxph_ps): Likewise.
1722 (_mm512_maskz_cvtxph_ps): Likewise.
1723 (_mm512_cvtx_roundph_ps): Likewise.
1724 (_mm512_mask_cvtx_roundph_ps): Likewise.
1725 (_mm512_maskz_cvtx_roundph_ps): Likewise.
1726 (_mm512_cvtxps_ph): Likewise.
1727 (_mm512_mask_cvtxps_ph): Likewise.
1728 (_mm512_maskz_cvtxps_ph): Likewise.
1729 (_mm512_cvtx_roundps_ph): Likewise.
1730 (_mm512_mask_cvtx_roundps_ph): Likewise.
1731 (_mm512_maskz_cvtx_roundps_ph): Likewise.
1732 (_mm512_cvtpd_ph): Likewise.
1733 (_mm512_mask_cvtpd_ph): Likewise.
1734 (_mm512_maskz_cvtpd_ph): Likewise.
1735 (_mm512_cvt_roundpd_ph): Likewise.
1736 (_mm512_mask_cvt_roundpd_ph): Likewise.
1737 (_mm512_maskz_cvt_roundpd_ph): Likewise.
1738 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
1740 (_mm_mask_cvtph_pd): Likewise.
1741 (_mm_maskz_cvtph_pd): Likewise.
1742 (_mm256_cvtph_pd): Likewise.
1743 (_mm256_mask_cvtph_pd): Likewise.
1744 (_mm256_maskz_cvtph_pd): Likewise.
1745 (_mm_cvtxph_ps): Likewise.
1746 (_mm_mask_cvtxph_ps): Likewise.
1747 (_mm_maskz_cvtxph_ps): Likewise.
1748 (_mm256_cvtxph_ps): Likewise.
1749 (_mm256_mask_cvtxph_ps): Likewise.
1750 (_mm256_maskz_cvtxph_ps): Likewise.
1751 (_mm_cvtxps_ph): Likewise.
1752 (_mm_mask_cvtxps_ph): Likewise.
1753 (_mm_maskz_cvtxps_ph): Likewise.
1754 (_mm256_cvtxps_ph): Likewise.
1755 (_mm256_mask_cvtxps_ph): Likewise.
1756 (_mm256_maskz_cvtxps_ph): Likewise.
1757 (_mm_cvtpd_ph): Likewise.
1758 (_mm_mask_cvtpd_ph): Likewise.
1759 (_mm_maskz_cvtpd_ph): Likewise.
1760 (_mm256_cvtpd_ph): Likewise.
1761 (_mm256_mask_cvtpd_ph): Likewise.
1762 (_mm256_maskz_cvtpd_ph): Likewise.
1763 * config/i386/i386-builtin.def: Add corresponding new builtins.
1764 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
1765 * config/i386/i386-expand.c: Handle new builtin types.
1766 * config/i386/sse.md
1767 (VF4_128_8_256): New.
1768 (VF48H_AVX512VL): Ditto.
1769 (ssePHmode): Add HF vector modes.
1770 (castmode): Add new convertable modes.
1773 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
1774 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
1775 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
1776 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
1777 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
1778 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
1779 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
1781 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
1782 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
1783 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
1784 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
1786 2021-09-17 liuhongt <hongtao.liu@intel.com>
1788 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
1790 (_mm_cvttsh_u32): Likewise.
1791 (_mm_cvtt_roundsh_i32): Likewise.
1792 (_mm_cvtt_roundsh_u32): Likewise.
1793 (_mm_cvttsh_i64): Likewise.
1794 (_mm_cvttsh_u64): Likewise.
1795 (_mm_cvtt_roundsh_i64): Likewise.
1796 (_mm_cvtt_roundsh_u64): Likewise.
1797 * config/i386/i386-builtin.def: Add corresponding new builtins.
1798 * config/i386/sse.md
1799 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
1802 2021-09-17 liuhongt <hongtao.liu@intel.com>
1804 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
1806 (_mm512_mask_cvttph_epi32): Likewise.
1807 (_mm512_maskz_cvttph_epi32): Likewise.
1808 (_mm512_cvtt_roundph_epi32): Likewise.
1809 (_mm512_mask_cvtt_roundph_epi32): Likewise.
1810 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
1811 (_mm512_cvttph_epu32): Likewise.
1812 (_mm512_mask_cvttph_epu32): Likewise.
1813 (_mm512_maskz_cvttph_epu32): Likewise.
1814 (_mm512_cvtt_roundph_epu32): Likewise.
1815 (_mm512_mask_cvtt_roundph_epu32): Likewise.
1816 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
1817 (_mm512_cvttph_epi64): Likewise.
1818 (_mm512_mask_cvttph_epi64): Likewise.
1819 (_mm512_maskz_cvttph_epi64): Likewise.
1820 (_mm512_cvtt_roundph_epi64): Likewise.
1821 (_mm512_mask_cvtt_roundph_epi64): Likewise.
1822 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
1823 (_mm512_cvttph_epu64): Likewise.
1824 (_mm512_mask_cvttph_epu64): Likewise.
1825 (_mm512_maskz_cvttph_epu64): Likewise.
1826 (_mm512_cvtt_roundph_epu64): Likewise.
1827 (_mm512_mask_cvtt_roundph_epu64): Likewise.
1828 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
1829 (_mm512_cvttph_epi16): Likewise.
1830 (_mm512_mask_cvttph_epi16): Likewise.
1831 (_mm512_maskz_cvttph_epi16): Likewise.
1832 (_mm512_cvtt_roundph_epi16): Likewise.
1833 (_mm512_mask_cvtt_roundph_epi16): Likewise.
1834 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
1835 (_mm512_cvttph_epu16): Likewise.
1836 (_mm512_mask_cvttph_epu16): Likewise.
1837 (_mm512_maskz_cvttph_epu16): Likewise.
1838 (_mm512_cvtt_roundph_epu16): Likewise.
1839 (_mm512_mask_cvtt_roundph_epu16): Likewise.
1840 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
1841 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
1843 (_mm_mask_cvttph_epi32): Likewise.
1844 (_mm_maskz_cvttph_epi32): Likewise.
1845 (_mm256_cvttph_epi32): Likewise.
1846 (_mm256_mask_cvttph_epi32): Likewise.
1847 (_mm256_maskz_cvttph_epi32): Likewise.
1848 (_mm_cvttph_epu32): Likewise.
1849 (_mm_mask_cvttph_epu32): Likewise.
1850 (_mm_maskz_cvttph_epu32): Likewise.
1851 (_mm256_cvttph_epu32): Likewise.
1852 (_mm256_mask_cvttph_epu32): Likewise.
1853 (_mm256_maskz_cvttph_epu32): Likewise.
1854 (_mm_cvttph_epi64): Likewise.
1855 (_mm_mask_cvttph_epi64): Likewise.
1856 (_mm_maskz_cvttph_epi64): Likewise.
1857 (_mm256_cvttph_epi64): Likewise.
1858 (_mm256_mask_cvttph_epi64): Likewise.
1859 (_mm256_maskz_cvttph_epi64): Likewise.
1860 (_mm_cvttph_epu64): Likewise.
1861 (_mm_mask_cvttph_epu64): Likewise.
1862 (_mm_maskz_cvttph_epu64): Likewise.
1863 (_mm256_cvttph_epu64): Likewise.
1864 (_mm256_mask_cvttph_epu64): Likewise.
1865 (_mm256_maskz_cvttph_epu64): Likewise.
1866 (_mm_cvttph_epi16): Likewise.
1867 (_mm_mask_cvttph_epi16): Likewise.
1868 (_mm_maskz_cvttph_epi16): Likewise.
1869 (_mm256_cvttph_epi16): Likewise.
1870 (_mm256_mask_cvttph_epi16): Likewise.
1871 (_mm256_maskz_cvttph_epi16): Likewise.
1872 (_mm_cvttph_epu16): Likewise.
1873 (_mm_mask_cvttph_epu16): Likewise.
1874 (_mm_maskz_cvttph_epu16): Likewise.
1875 (_mm256_cvttph_epu16): Likewise.
1876 (_mm256_mask_cvttph_epu16): Likewise.
1877 (_mm256_maskz_cvttph_epu16): Likewise.
1878 * config/i386/i386-builtin.def: Add new builtins.
1879 * config/i386/sse.md
1880 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
1882 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
1883 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
1884 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
1885 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
1887 2021-09-17 liuhongt <hongtao.liu@intel.com>
1889 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
1890 (_mm_cvtsh_u32): Likewise.
1891 (_mm_cvt_roundsh_i32): Likewise.
1892 (_mm_cvt_roundsh_u32): Likewise.
1893 (_mm_cvtsh_i64): Likewise.
1894 (_mm_cvtsh_u64): Likewise.
1895 (_mm_cvt_roundsh_i64): Likewise.
1896 (_mm_cvt_roundsh_u64): Likewise.
1897 (_mm_cvti32_sh): Likewise.
1898 (_mm_cvtu32_sh): Likewise.
1899 (_mm_cvt_roundi32_sh): Likewise.
1900 (_mm_cvt_roundu32_sh): Likewise.
1901 (_mm_cvti64_sh): Likewise.
1902 (_mm_cvtu64_sh): Likewise.
1903 (_mm_cvt_roundi64_sh): Likewise.
1904 (_mm_cvt_roundu64_sh): Likewise.
1905 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
1906 * config/i386/i386-builtin.def: Add corresponding new builtins.
1907 * config/i386/i386-expand.c (ix86_expand_round_builtin):
1908 Handle new builtin types.
1909 * config/i386/sse.md
1910 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
1912 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
1913 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
1915 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
1917 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
1918 (altivec_resolve_new_overloaded_builtin): New forward decl.
1919 (rs6000_new_builtin_type_compatible): New function.
1920 (altivec_resolve_overloaded_builtin): Call
1921 altivec_resolve_new_overloaded_builtin.
1922 (altivec_build_new_resolved_builtin): New function.
1923 (altivec_resolve_new_overloaded_builtin): Likewise.
1924 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
1926 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
1927 name of rs6000_new_builtin_is_supported.
1929 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
1931 * config/i386/i386-protos.h (ix86_decompose_address):
1932 Change return type to bool.
1933 * config/i386/i386.c (ix86_decompose_address): Ditto.
1935 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
1938 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
1939 Added 'build/' to target, use build/%.o rule.
1940 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
1941 '$(build_exeext)' to target and 'build/' for the *.o files.
1942 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
1945 2021-09-16 Martin Jambor <mjambor@suse.cz>
1947 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
1948 the body has been removed. Remove trailing whitespace.
1950 2021-09-16 Richard Biener <rguenther@suse.de>
1952 PR middle-end/102360
1953 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
1954 of non-memory more robust.
1956 2021-09-16 Daniel Cederman <cederman@gaisler.com>
1958 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
1959 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
1960 (leon5_adjust_cost): Increase cost of store with data dependency
1961 on ALU instruction and FPU anti-dependencies.
1962 (sparc_option_override): Add LEON5 costs
1963 (sparc_adjust_cost): Add LEON5 cost adjustments
1964 * config/sparc/sparc.h: Add LEON5
1965 * config/sparc/sparc.md: Include LEON5 scheduling information
1966 * config/sparc/sparc.opt: Add LEON5
1967 * doc/invoke.texi: Add LEON5
1968 * config/sparc/leon5.md: New file.
1970 2021-09-16 Daniel Cederman <cederman@gaisler.com>
1972 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
1973 sensitive sequence for B2BST errata workaround.
1975 2021-09-16 Daniel Cederman <cederman@gaisler.com>
1977 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
1978 functions with atomic instruction in the UT700 errata workaround.
1980 2021-09-16 Daniel Cederman <cederman@gaisler.com>
1982 * config/sparc/sparc.c (next_active_non_empty_insn): New function
1983 that returns next active non empty assembly instruction.
1984 (sparc_do_work_around_errata): Use new function.
1986 2021-09-16 Daniel Cederman <cederman@gaisler.com>
1988 * config/sparc/sparc.c (store_insn_p): Add predicate for store
1990 (load_insn_p): Add predicate for load attributes.
1991 (sparc_do_work_around_errata): Use new predicates.
1993 2021-09-16 Andreas Larsson <andreas@gaisler.com>
1995 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
1998 2021-09-16 Martin Liska <mliska@suse.cz>
2000 * config/mips/netbsd.h: Fix typo in name of a macro.
2002 2021-09-16 liuhongt <hongtao.liu@intel.com>
2004 PR middle-end/102080
2005 * match.pd: Check mask type when doing cond_op related gimple
2007 * tree.c (is_truth_type_for): New function.
2008 * tree.h (is_truth_type_for): New declaration.
2010 2021-09-16 liuhongt <hongtao.liu@intel.com>
2012 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2014 (_mm512_mask_cvtepi32_ph): Likewise.
2015 (_mm512_maskz_cvtepi32_ph): Likewise.
2016 (_mm512_cvt_roundepi32_ph): Likewise.
2017 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2018 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2019 (_mm512_cvtepu32_ph): Likewise.
2020 (_mm512_mask_cvtepu32_ph): Likewise.
2021 (_mm512_maskz_cvtepu32_ph): Likewise.
2022 (_mm512_cvt_roundepu32_ph): Likewise.
2023 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2024 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2025 (_mm512_cvtepi64_ph): Likewise.
2026 (_mm512_mask_cvtepi64_ph): Likewise.
2027 (_mm512_maskz_cvtepi64_ph): Likewise.
2028 (_mm512_cvt_roundepi64_ph): Likewise.
2029 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2030 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2031 (_mm512_cvtepu64_ph): Likewise.
2032 (_mm512_mask_cvtepu64_ph): Likewise.
2033 (_mm512_maskz_cvtepu64_ph): Likewise.
2034 (_mm512_cvt_roundepu64_ph): Likewise.
2035 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2036 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2037 (_mm512_cvtepi16_ph): Likewise.
2038 (_mm512_mask_cvtepi16_ph): Likewise.
2039 (_mm512_maskz_cvtepi16_ph): Likewise.
2040 (_mm512_cvt_roundepi16_ph): Likewise.
2041 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2042 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2043 (_mm512_cvtepu16_ph): Likewise.
2044 (_mm512_mask_cvtepu16_ph): Likewise.
2045 (_mm512_maskz_cvtepu16_ph): Likewise.
2046 (_mm512_cvt_roundepu16_ph): Likewise.
2047 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2048 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2049 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2051 (_mm_mask_cvtepi32_ph): Likewise.
2052 (_mm_maskz_cvtepi32_ph): Likewise.
2053 (_mm256_cvtepi32_ph): Likewise.
2054 (_mm256_mask_cvtepi32_ph): Likewise.
2055 (_mm256_maskz_cvtepi32_ph): Likewise.
2056 (_mm_cvtepu32_ph): Likewise.
2057 (_mm_mask_cvtepu32_ph): Likewise.
2058 (_mm_maskz_cvtepu32_ph): Likewise.
2059 (_mm256_cvtepu32_ph): Likewise.
2060 (_mm256_mask_cvtepu32_ph): Likewise.
2061 (_mm256_maskz_cvtepu32_ph): Likewise.
2062 (_mm_cvtepi64_ph): Likewise.
2063 (_mm_mask_cvtepi64_ph): Likewise.
2064 (_mm_maskz_cvtepi64_ph): Likewise.
2065 (_mm256_cvtepi64_ph): Likewise.
2066 (_mm256_mask_cvtepi64_ph): Likewise.
2067 (_mm256_maskz_cvtepi64_ph): Likewise.
2068 (_mm_cvtepu64_ph): Likewise.
2069 (_mm_mask_cvtepu64_ph): Likewise.
2070 (_mm_maskz_cvtepu64_ph): Likewise.
2071 (_mm256_cvtepu64_ph): Likewise.
2072 (_mm256_mask_cvtepu64_ph): Likewise.
2073 (_mm256_maskz_cvtepu64_ph): Likewise.
2074 (_mm_cvtepi16_ph): Likewise.
2075 (_mm_mask_cvtepi16_ph): Likewise.
2076 (_mm_maskz_cvtepi16_ph): Likewise.
2077 (_mm256_cvtepi16_ph): Likewise.
2078 (_mm256_mask_cvtepi16_ph): Likewise.
2079 (_mm256_maskz_cvtepi16_ph): Likewise.
2080 (_mm_cvtepu16_ph): Likewise.
2081 (_mm_mask_cvtepu16_ph): Likewise.
2082 (_mm_maskz_cvtepu16_ph): Likewise.
2083 (_mm256_cvtepu16_ph): Likewise.
2084 (_mm256_mask_cvtepu16_ph): Likewise.
2085 (_mm256_maskz_cvtepu16_ph): Likewise.
2086 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2087 * config/i386/i386-builtin.def: Add corresponding new builtins.
2088 * config/i386/i386-expand.c
2089 (ix86_expand_args_builtin): Handle new builtin types.
2090 (ix86_expand_round_builtin): Ditto.
2091 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2092 * config/i386/sse.md (VI2H_AVX512VL): New.
2094 (sseintvecmode): Add HF vector modes.
2095 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2097 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2098 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2099 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2100 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2101 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2102 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2103 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2104 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2105 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2106 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2107 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2109 2021-09-16 liuhongt <hongtao.liu@intel.com>
2111 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2113 (_mm512_mask_cvtph_epi32): Likewise.
2114 (_mm512_maskz_cvtph_epi32): Likewise.
2115 (_mm512_cvt_roundph_epi32): Likewise.
2116 (_mm512_mask_cvt_roundph_epi32): Likewise.
2117 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2118 (_mm512_cvtph_epu32): Likewise.
2119 (_mm512_mask_cvtph_epu32): Likewise.
2120 (_mm512_maskz_cvtph_epu32): Likewise.
2121 (_mm512_cvt_roundph_epu32): Likewise.
2122 (_mm512_mask_cvt_roundph_epu32): Likewise.
2123 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2124 (_mm512_cvtph_epi64): Likewise.
2125 (_mm512_mask_cvtph_epi64): Likewise.
2126 (_mm512_maskz_cvtph_epi64): Likewise.
2127 (_mm512_cvt_roundph_epi64): Likewise.
2128 (_mm512_mask_cvt_roundph_epi64): Likewise.
2129 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2130 (_mm512_cvtph_epu64): Likewise.
2131 (_mm512_mask_cvtph_epu64): Likewise.
2132 (_mm512_maskz_cvtph_epu64): Likewise.
2133 (_mm512_cvt_roundph_epu64): Likewise.
2134 (_mm512_mask_cvt_roundph_epu64): Likewise.
2135 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2136 (_mm512_cvtph_epi16): Likewise.
2137 (_mm512_mask_cvtph_epi16): Likewise.
2138 (_mm512_maskz_cvtph_epi16): Likewise.
2139 (_mm512_cvt_roundph_epi16): Likewise.
2140 (_mm512_mask_cvt_roundph_epi16): Likewise.
2141 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2142 (_mm512_cvtph_epu16): Likewise.
2143 (_mm512_mask_cvtph_epu16): Likewise.
2144 (_mm512_maskz_cvtph_epu16): Likewise.
2145 (_mm512_cvt_roundph_epu16): Likewise.
2146 (_mm512_mask_cvt_roundph_epu16): Likewise.
2147 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2148 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2150 (_mm_mask_cvtph_epi32): Likewise.
2151 (_mm_maskz_cvtph_epi32): Likewise.
2152 (_mm256_cvtph_epi32): Likewise.
2153 (_mm256_mask_cvtph_epi32): Likewise.
2154 (_mm256_maskz_cvtph_epi32): Likewise.
2155 (_mm_cvtph_epu32): Likewise.
2156 (_mm_mask_cvtph_epu32): Likewise.
2157 (_mm_maskz_cvtph_epu32): Likewise.
2158 (_mm256_cvtph_epu32): Likewise.
2159 (_mm256_mask_cvtph_epu32): Likewise.
2160 (_mm256_maskz_cvtph_epu32): Likewise.
2161 (_mm_cvtph_epi64): Likewise.
2162 (_mm_mask_cvtph_epi64): Likewise.
2163 (_mm_maskz_cvtph_epi64): Likewise.
2164 (_mm256_cvtph_epi64): Likewise.
2165 (_mm256_mask_cvtph_epi64): Likewise.
2166 (_mm256_maskz_cvtph_epi64): Likewise.
2167 (_mm_cvtph_epu64): Likewise.
2168 (_mm_mask_cvtph_epu64): Likewise.
2169 (_mm_maskz_cvtph_epu64): Likewise.
2170 (_mm256_cvtph_epu64): Likewise.
2171 (_mm256_mask_cvtph_epu64): Likewise.
2172 (_mm256_maskz_cvtph_epu64): Likewise.
2173 (_mm_cvtph_epi16): Likewise.
2174 (_mm_mask_cvtph_epi16): Likewise.
2175 (_mm_maskz_cvtph_epi16): Likewise.
2176 (_mm256_cvtph_epi16): Likewise.
2177 (_mm256_mask_cvtph_epi16): Likewise.
2178 (_mm256_maskz_cvtph_epi16): Likewise.
2179 (_mm_cvtph_epu16): Likewise.
2180 (_mm_mask_cvtph_epu16): Likewise.
2181 (_mm_maskz_cvtph_epu16): Likewise.
2182 (_mm256_cvtph_epu16): Likewise.
2183 (_mm256_mask_cvtph_epu16): Likewise.
2184 (_mm256_maskz_cvtph_epu16): Likewise.
2185 * config/i386/i386-builtin-types.def: Add new builtin types.
2186 * config/i386/i386-builtin.def: Add new builtins.
2187 * config/i386/i386-expand.c
2188 (ix86_expand_args_builtin): Handle new builtin types.
2189 (ix86_expand_round_builtin): Ditto.
2190 * config/i386/sse.md (sseintconvert): New.
2192 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2193 (sseintconvertsignprefix): Ditto.
2194 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2197 2021-09-16 liuhongt <hongtao.liu@intel.com>
2199 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2201 (_mm_cvtsi128_si16): Likewise.
2202 (_mm_mask_load_sh): Likewise.
2203 (_mm_maskz_load_sh): Likewise.
2204 (_mm_mask_store_sh): Likewise.
2205 (_mm_move_sh): Likewise.
2206 (_mm_mask_move_sh): Likewise.
2207 (_mm_maskz_move_sh): Likewise.
2208 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2209 * config/i386/i386-builtin.def: Add corresponding new builtins.
2210 * config/i386/i386-expand.c
2211 (ix86_expand_special_args_builtin): Handle new builtin types.
2212 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2213 * config/i386/sse.md (VI2F): New mode iterator.
2214 (vec_set<mode>_0): Use new mode iterator.
2215 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2216 (avx512f_store<mode>_mask): Ditto.
2218 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2220 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2222 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2224 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2225 Proceed if no symbol summary or the symbol alias flag is false.
2227 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2231 * varasm.c (output_constructor_regular_field): Instead of assertion
2232 that array_size_for_constructor result is equal to size of
2233 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2234 or equal to array_size_for_constructor result and use type size as
2237 2021-09-15 Martin Liska <mliska@suse.cz>
2240 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2242 2021-09-15 Martin Liska <mliska@suse.cz>
2245 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2246 Check that we have a symbol summary for a symbol.
2248 2021-09-15 Richard Biener <rguenther@suse.de>
2251 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2252 to inherit from elfos.h
2254 2021-09-15 liuhongt <hongtao.liu@intel.com>
2257 * config/i386/i386-expand.c
2258 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2260 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2261 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2262 AVX512FP16 which supports 16bit vector load.
2263 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2265 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2266 extend to V32HFmode.
2267 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2268 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2270 (vec_interleave_highv8hi<mask_name>): Rename to ..
2271 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2272 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2274 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2275 this, and extend to V32HFmode.
2276 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2277 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2278 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2279 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2280 (sse4_1_pblendw): Rename to ..
2281 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2282 (avx2_pblendph): New define_expand.
2283 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2284 sseintmodesuffix instead of ssemodesuffix.
2285 (blendsuf): New mode attr.
2287 2021-09-15 Richard Biener <rguenther@suse.de>
2289 * tree-vectorizer.h (dr_misalignment): Move out of line.
2290 (dr_target_alignment): New.
2291 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2292 (set_dr_target_alignment): New.
2293 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2294 * tree-vect-data-refs.c (dr_misalignment): Compute and
2295 return the group members misalignment.
2296 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2297 (vect_analyze_data_refs_alignment): Compute alignment only
2298 for the first element of a DR group.
2299 (vect_slp_analyze_node_alignment): Likewise.
2301 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2303 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2304 * config/i386/avx512fp16vlintrin.h: Likewise.
2305 * config/i386/i386-builtin.def: Adjust builtin name and
2306 enumeration to match AVX512F style.
2308 2021-09-15 Richard Biener <rguenther@suse.de>
2310 PR tree-optimization/102318
2311 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2312 previous change and do the mode conversion separately from
2313 the sign conversion.
2315 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2316 Peter Cordes <peter@cordes.ca>
2319 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2320 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2321 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2324 2021-09-15 Richard Biener <rguenther@suse.de>
2326 * config.gcc: Remove vax-*-openbsd* configuration.
2328 2021-09-15 Richard Biener <rguenther@suse.de>
2330 * config.gcc: Remove m68k-openbsd.
2332 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
2335 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
2337 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
2339 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
2340 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
2341 (*mma_xxsetaccz): Delete.
2342 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
2343 Use UNSPECV_MMA_XXSETACCZ. Update comment.
2344 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
2346 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
2348 * Makefile.in: Remove variables related to applying no-PIE
2349 to the exes on $build.
2350 * configure: Regenerate.
2351 * configure.ac: Remove configuration related to applying
2352 no-PIE to the exes on $build.
2354 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
2356 * config/arc/arc.md (doloop_end): Add missing mode.
2357 (loop_end): Likewise.
2359 2021-09-14 Jakub Jelinek <jakub@redhat.com>
2361 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
2362 it to recursive calls, for depth above 7 just gimplify or return.
2363 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
2364 __builtin_clear_padding and TARGET_EXPR.
2365 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
2367 2021-09-14 liuhongt <hongtao.liu@intel.com>
2369 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
2371 (_mm_mask_fpclass_sh_mask): Likewise.
2372 (_mm512_mask_fpclass_ph_mask): Likewise.
2373 (_mm512_fpclass_ph_mask): Likewise.
2374 (_mm_getexp_sh): Likewise.
2375 (_mm_mask_getexp_sh): Likewise.
2376 (_mm_maskz_getexp_sh): Likewise.
2377 (_mm512_getexp_ph): Likewise.
2378 (_mm512_mask_getexp_ph): Likewise.
2379 (_mm512_maskz_getexp_ph): Likewise.
2380 (_mm_getexp_round_sh): Likewise.
2381 (_mm_mask_getexp_round_sh): Likewise.
2382 (_mm_maskz_getexp_round_sh): Likewise.
2383 (_mm512_getexp_round_ph): Likewise.
2384 (_mm512_mask_getexp_round_ph): Likewise.
2385 (_mm512_maskz_getexp_round_ph): Likewise.
2386 (_mm_getmant_sh): Likewise.
2387 (_mm_mask_getmant_sh): Likewise.
2388 (_mm_maskz_getmant_sh): Likewise.
2389 (_mm512_getmant_ph): Likewise.
2390 (_mm512_mask_getmant_ph): Likewise.
2391 (_mm512_maskz_getmant_ph): Likewise.
2392 (_mm_getmant_round_sh): Likewise.
2393 (_mm_mask_getmant_round_sh): Likewise.
2394 (_mm_maskz_getmant_round_sh): Likewise.
2395 (_mm512_getmant_round_ph): Likewise.
2396 (_mm512_mask_getmant_round_ph): Likewise.
2397 (_mm512_maskz_getmant_round_ph): Likewise.
2398 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
2400 (_mm_fpclass_ph_mask): Likewise.
2401 (_mm256_mask_fpclass_ph_mask): Likewise.
2402 (_mm256_fpclass_ph_mask): Likewise.
2403 (_mm256_getexp_ph): Likewise.
2404 (_mm256_mask_getexp_ph): Likewise.
2405 (_mm256_maskz_getexp_ph): Likewise.
2406 (_mm_getexp_ph): Likewise.
2407 (_mm_mask_getexp_ph): Likewise.
2408 (_mm_maskz_getexp_ph): Likewise.
2409 (_mm256_getmant_ph): Likewise.
2410 (_mm256_mask_getmant_ph): Likewise.
2411 (_mm256_maskz_getmant_ph): Likewise.
2412 (_mm_getmant_ph): Likewise.
2413 (_mm_mask_getmant_ph): Likewise.
2414 (_mm_maskz_getmant_ph): Likewise.
2415 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2416 * config/i386/i386-builtin.def: Add corresponding new builtins.
2417 * config/i386/i386-expand.c
2418 (ix86_expand_args_builtin): Handle new builtin types.
2419 (ix86_expand_round_builtin): Ditto.
2420 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
2421 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
2422 to support HF vector modes.
2423 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
2425 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
2426 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
2427 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2428 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2431 2021-09-14 liuhongt <hongtao.liu@intel.com>
2433 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
2435 (_mm512_mask_reduce_ph): Likewise.
2436 (_mm512_maskz_reduce_ph): Likewise.
2437 (_mm512_reduce_round_ph): Likewise.
2438 (_mm512_mask_reduce_round_ph): Likewise.
2439 (_mm512_maskz_reduce_round_ph): Likewise.
2440 (_mm_reduce_sh): Likewise.
2441 (_mm_mask_reduce_sh): Likewise.
2442 (_mm_maskz_reduce_sh): Likewise.
2443 (_mm_reduce_round_sh): Likewise.
2444 (_mm_mask_reduce_round_sh): Likewise.
2445 (_mm_maskz_reduce_round_sh): Likewise.
2446 (_mm512_roundscale_ph): Likewise.
2447 (_mm512_mask_roundscale_ph): Likewise.
2448 (_mm512_maskz_roundscale_ph): Likewise.
2449 (_mm512_roundscale_round_ph): Likewise.
2450 (_mm512_mask_roundscale_round_ph): Likewise.
2451 (_mm512_maskz_roundscale_round_ph): Likewise.
2452 (_mm_roundscale_sh): Likewise.
2453 (_mm_mask_roundscale_sh): Likewise.
2454 (_mm_maskz_roundscale_sh): Likewise.
2455 (_mm_roundscale_round_sh): Likewise.
2456 (_mm_mask_roundscale_round_sh): Likewise.
2457 (_mm_maskz_roundscale_round_sh): Likewise.
2458 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
2460 (_mm_mask_reduce_ph): Likewise.
2461 (_mm_maskz_reduce_ph): Likewise.
2462 (_mm256_reduce_ph): Likewise.
2463 (_mm256_mask_reduce_ph): Likewise.
2464 (_mm256_maskz_reduce_ph): Likewise.
2465 (_mm_roundscale_ph): Likewise.
2466 (_mm_mask_roundscale_ph): Likewise.
2467 (_mm_maskz_roundscale_ph): Likewise.
2468 (_mm256_roundscale_ph): Likewise.
2469 (_mm256_mask_roundscale_ph): Likewise.
2470 (_mm256_maskz_roundscale_ph): Likewise.
2471 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2472 * config/i386/i386-builtin.def: Add corresponding new builtins.
2473 * config/i386/i386-expand.c
2474 (ix86_expand_args_builtin): Handle new builtin types.
2475 (ix86_expand_round_builtin): Ditto.
2476 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
2478 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
2479 ... this, and adjust for round operands.
2480 (reduces<mode><mask_scalar_name>): Likewise, with ...
2481 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
2483 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
2484 Adjust for HF vector modes.
2485 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
2487 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2489 2021-09-14 liuhongt <hongtao.liu@intel.com>
2491 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
2493 (_mm512_mask_rcp_ph): Likewise.
2494 (_mm512_maskz_rcp_ph): Likewise.
2495 (_mm_rcp_sh): Likewise.
2496 (_mm_mask_rcp_sh): Likewise.
2497 (_mm_maskz_rcp_sh): Likewise.
2498 (_mm512_scalef_ph): Likewise.
2499 (_mm512_mask_scalef_ph): Likewise.
2500 (_mm512_maskz_scalef_ph): Likewise.
2501 (_mm512_scalef_round_ph): Likewise.
2502 (_mm512_mask_scalef_round_ph): Likewise.
2503 (_mm512_maskz_scalef_round_ph): Likewise.
2504 (_mm_scalef_sh): Likewise.
2505 (_mm_mask_scalef_sh): Likewise.
2506 (_mm_maskz_scalef_sh): Likewise.
2507 (_mm_scalef_round_sh): Likewise.
2508 (_mm_mask_scalef_round_sh): Likewise.
2509 (_mm_maskz_scalef_round_sh): Likewise.
2510 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
2512 (_mm256_rcp_ph): Likewise.
2513 (_mm_mask_rcp_ph): Likewise.
2514 (_mm256_mask_rcp_ph): Likewise.
2515 (_mm_maskz_rcp_ph): Likewise.
2516 (_mm256_maskz_rcp_ph): Likewise.
2517 (_mm_scalef_ph): Likewise.
2518 (_mm256_scalef_ph): Likewise.
2519 (_mm_mask_scalef_ph): Likewise.
2520 (_mm256_mask_scalef_ph): Likewise.
2521 (_mm_maskz_scalef_ph): Likewise.
2522 (_mm256_maskz_scalef_ph): Likewise.
2523 * config/i386/i386-builtin.def: Add new builtins.
2524 * config/i386/sse.md (VFH_AVX512VL): New.
2525 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
2526 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
2527 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
2528 Adjust to support HF vector modes.
2529 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
2531 2021-09-14 liuhongt <hongtao.liu@intel.com>
2533 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
2535 (_mm512_mask_sqrt_ph): Likewise.
2536 (_mm512_maskz_sqrt_ph): Likewise.
2537 (_mm512_sqrt_round_ph): Likewise.
2538 (_mm512_mask_sqrt_round_ph): Likewise.
2539 (_mm512_maskz_sqrt_round_ph): Likewise.
2540 (_mm512_rsqrt_ph): Likewise.
2541 (_mm512_mask_rsqrt_ph): Likewise.
2542 (_mm512_maskz_rsqrt_ph): Likewise.
2543 (_mm_rsqrt_sh): Likewise.
2544 (_mm_mask_rsqrt_sh): Likewise.
2545 (_mm_maskz_rsqrt_sh): Likewise.
2546 (_mm_sqrt_sh): Likewise.
2547 (_mm_mask_sqrt_sh): Likewise.
2548 (_mm_maskz_sqrt_sh): Likewise.
2549 (_mm_sqrt_round_sh): Likewise.
2550 (_mm_mask_sqrt_round_sh): Likewise.
2551 (_mm_maskz_sqrt_round_sh): Likewise.
2552 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
2553 (_mm256_sqrt_ph): Likewise.
2554 (_mm_mask_sqrt_ph): Likewise.
2555 (_mm256_mask_sqrt_ph): Likewise.
2556 (_mm_maskz_sqrt_ph): Likewise.
2557 (_mm256_maskz_sqrt_ph): Likewise.
2558 (_mm_rsqrt_ph): Likewise.
2559 (_mm256_rsqrt_ph): Likewise.
2560 (_mm_mask_rsqrt_ph): Likewise.
2561 (_mm256_mask_rsqrt_ph): Likewise.
2562 (_mm_maskz_rsqrt_ph): Likewise.
2563 (_mm256_maskz_rsqrt_ph): Likewise.
2564 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2565 * config/i386/i386-builtin.def: Add corresponding new builtins.
2566 * config/i386/i386-expand.c
2567 (ix86_expand_args_builtin): Handle new builtins.
2568 (ix86_expand_round_builtin): Ditto.
2569 * config/i386/sse.md (VF_AVX512FP16VL): New.
2570 (sqrt<mode>2): Adjust for HF vector modes.
2571 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
2572 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2574 (<sse>_rsqrt<mode>2<mask_name>): New.
2575 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
2577 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2580 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
2581 'RESERVED_LOCATION_P' locations.
2582 * warning-control.cc (get_nowarn_spec, suppress_warning)
2583 (copy_warning): Likewise.
2585 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2587 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
2588 (typedef key_type_t): ... this. Remove.
2589 (nowarn_map): Document.
2590 * diagnostic-spec.c (nowarn_map): Likewise.
2591 * warning-control.cc (convert_to_key): Evolve functions into...
2592 (get_location): ... these. Adjust all users.
2594 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2596 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
2598 2021-09-13 Jason Merrill <jason@redhat.com>
2600 * params.opt: Add destructive-interference-size and
2601 constructive-interference-size.
2602 * doc/invoke.texi: Document them.
2603 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2605 * config/arm/arm.c (arm_option_override): Set them.
2606 * config/i386/i386-options.c (ix86_option_override_internal):
2609 2021-09-13 Martin Liska <mliska@suse.cz>
2610 H.J. Lu <hjl.tools@gmail.com>
2613 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
2614 for x86-64 micro levels for __builtin_cpu_supports.
2615 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
2616 Add priorities for the micro-arch levels.
2617 (enum processor_features): Add new features.
2618 * common/config/i386/i386-isas.h: Add micro-arch features.
2619 * config/i386/i386-builtins.c (get_builtin_code_for_version):
2620 Support the micro-arch levels by callsing
2621 __builtin_cpu_supports.
2622 * doc/extend.texi: Document that the levels are support by
2623 __builtin_cpu_supports.
2625 2021-09-13 Andrew Pinski <apinski@marvell.com>
2628 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
2630 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
2631 (aarch64_general_gimple_fold_builtin): Likewise.
2633 2021-09-13 Andrew Pinski <apinski@marvell.com>
2635 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
2636 to the Unsupported targets list.
2637 Remove support for m32r-*-linux* and m32rle-*-linux*.
2638 * config/m32r/linux.h: Removed.
2639 * config/m32r/t-linux: Removed.
2641 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2644 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
2645 register index for SVE predicate modes.
2647 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
2649 * tree-ssa-threadbackward.c
2650 (back_threader_profitability::profitable_path_p): Remove FSM
2652 (back_threader_registry::register_path): Same.
2653 * tree-ssa-threadedge.c
2654 (jump_threader::simplify_control_stmt_condition): Same.
2655 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
2656 Add backedge_threads argument.
2657 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
2658 backedge_threads argument.
2659 (back_jt_path_registry::back_jt_path_registry): Same.
2660 (dump_jump_thread_path): Adjust for FSM removal.
2661 (back_jt_path_registry::rewire_first_differing_edge): Same.
2662 (back_jt_path_registry::adjust_paths_after_duplication): Same.
2663 (back_jt_path_registry::update_cfg): Same.
2664 (jt_path_registry::register_jump_thread): Same.
2665 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
2667 (class back_jt_path_registry): Add backedge_threads to
2670 2021-09-13 Martin Liska <mliska@suse.cz>
2673 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
2675 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
2678 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
2679 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
2680 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
2681 TARGET_AVX256_STORE_BY_PIECES instead of
2682 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
2683 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
2684 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
2685 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
2686 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
2687 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
2689 2021-09-13 liuhongt <hongtao.liu@intel.com>
2692 * expmed.c (extract_bit_field_using_extv): Use
2693 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
2695 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
2697 * Makefile.in (OBJS): Add value-pointer-equiv.o.
2698 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
2699 value-pointer-equiv.*.
2700 (ssa_equiv_stack::ssa_equiv_stack): Same.
2701 (ssa_equiv_stack::enter): Same.
2702 (ssa_equiv_stack::leave): Same.
2703 (ssa_equiv_stack::push_replacement): Same.
2704 (ssa_equiv_stack::get_replacement): Same.
2705 (is_pointer_ssa): Same.
2706 (class pointer_equiv_analyzer): Same.
2707 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
2708 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
2709 (pointer_equiv_analyzer::set_global_equiv): Same.
2710 (pointer_equiv_analyzer::set_cond_equiv): Same.
2711 (pointer_equiv_analyzer::get_equiv): Same.
2712 (pointer_equiv_analyzer::enter): Same.
2713 (pointer_equiv_analyzer::leave): Same.
2714 (pointer_equiv_analyzer::get_equiv_expr): Same.
2715 (pta_valueize): Same.
2716 (pointer_equiv_analyzer::visit_stmt): Same.
2717 (pointer_equiv_analyzer::visit_edge): Same.
2718 (hybrid_folder::value_of_expr): Same.
2719 (hybrid_folder::value_on_edge): Same.
2720 * value-pointer-equiv.cc: New file.
2721 * value-pointer-equiv.h: New file.
2723 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2726 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
2727 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
2729 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2732 * config/arm/arm.md (movmisaligndi): New define_expand.
2733 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
2735 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2738 * emit-rtl.c (gen_highpart): Use adjust_address to handle
2739 MEM rather than calling simplify_gen_subreg.
2741 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
2743 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
2744 define into a block.
2746 2021-09-13 Richard Biener <rguenther@suse.de>
2748 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
2749 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
2750 change based on DARWIN_PREFER_DWARF not being defined.
2752 2021-09-13 Richard Biener <rguenther@suse.de>
2754 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2755 to inherit from elfos.h
2757 2021-09-13 Richard Biener <rguenther@suse.de>
2759 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
2761 2021-09-13 Richard Biener <rguenther@suse.de>
2763 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
2764 override, pick up DWARF2_DEBUG define from elfos.h
2766 2021-09-13 Richard Biener <rguenther@suse.de>
2768 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
2771 2021-09-13 Richard Biener <rguenther@suse.de>
2773 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
2776 2021-09-13 Richard Biener <rguenther@suse.de>
2778 * config/i386/cygming.h: Always default to DWARF2 debugging.
2779 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
2781 * doc/install.texi: Document binutils 2.16 as minimum
2782 requirement for mingw.
2784 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
2786 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
2787 nstmts, nloads and extra_ctor_cost.
2788 (rs6000_density_test): Add load density related heuristics. Do
2789 extra costing on vector construction statements if need.
2790 (rs6000_init_cost): Init new members.
2791 (rs6000_update_target_cost_per_stmt): New function.
2792 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
2793 rs6000_update_target_cost_per_stmt and call it.
2795 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
2797 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
2798 (rs6000_init_cost): Adjust.
2800 2021-09-13 liuhongt <hongtao.liu@intel.com>
2802 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
2803 (UNSPEC_XORSIGN): Ditto.
2805 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
2807 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
2808 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
2811 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
2813 * tree-ssa-threadbackward.c (class back_threader_registry): Use
2814 back_jt_path_registry.
2815 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
2816 fwd_jt_path_registry.
2817 * tree-ssa-threadedge.h (class jump_threader): Same..
2818 * tree-ssa-threadupdate.c
2819 (jump_thread_path_registry::jump_thread_path_registry): Rename...
2820 (jt_path_registry::jt_path_registry): ...to this.
2821 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
2822 (jt_path_registry::~jt_path_registry): ...this.
2823 (fwd_jt_path_registry::fwd_jt_path_registry): New.
2824 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
2825 (jump_thread_path_registry::allocate_thread_edge): Rename...
2826 (jt_path_registry::allocate_thread_edge): ...to this.
2827 (jump_thread_path_registry::allocate_thread_path): Rename...
2828 (jt_path_registry::allocate_thread_path): ...to this.
2829 (jump_thread_path_registry::lookup_redirection_data): Rename...
2830 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
2831 (jump_thread_path_registry::thread_block_1): Rename...
2832 (fwd_jt_path_registry::thread_block_1): ...to this.
2833 (jump_thread_path_registry::thread_block): Rename...
2834 (fwd_jt_path_registry::thread_block): ...to this.
2835 (jt_path_registry::thread_through_loop_header): Rename...
2836 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
2837 (jump_thread_path_registry::mark_threaded_blocks): Rename...
2838 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
2839 (jump_thread_path_registry::debug_path): Rename...
2840 (jt_path_registry::debug_path): ...to this.
2841 (jump_thread_path_registry::dump): Rename...
2842 (jt_path_registry::debug): ...to this.
2843 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
2844 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
2845 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
2846 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
2847 (jump_thread_path_registry::duplicate_thread_path): Rename...
2848 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
2849 drop ill-formed candidates.
2850 (jump_thread_path_registry::remove_jump_threads_including): Rename...
2851 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
2852 (jt_path_registry::thread_through_all_blocks): New.
2853 (back_jt_path_registry::update_cfg): New.
2854 (fwd_jt_path_registry::update_cfg): New.
2855 (jump_thread_path_registry::register_jump_thread): Rename...
2856 (jt_path_registry::register_jump_thread): ...to this.
2857 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
2859 (class jt_path_registry): ...here.
2860 (class fwd_jt_path_registry): New.
2861 (class back_jt_path_registry): New.
2863 2021-09-10 liuhongt <hongtao.liu@intel.com>
2866 2021-09-01 liuhongt <hongtao.liu@intel.com>
2868 * emit-rtl.c (validate_subreg): Get rid of all float-int
2871 2021-09-10 Jakub Jelinek <jakub@redhat.com>
2873 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
2874 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
2875 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
2876 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
2877 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
2878 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
2879 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
2880 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
2882 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
2883 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
2885 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
2887 * tree.h (OMP_ATOMIC_WEAK): Define.
2888 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
2889 fail memory order being encoded in the same enum and also print
2890 fail clause if present.
2891 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
2892 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
2893 handle pre_p == NULL case as a test mode that only returns value
2894 but doesn't change gimplify nor change anything otherwise, adjust
2895 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
2896 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
2897 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
2899 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
2900 COND_EXPR rhs. Set weak flag on gimple load/store for
2902 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
2903 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
2905 (expand_omp_atomic_cas): New function.
2906 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
2908 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
2909 using expand_omp_atomic_cas.
2911 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
2912 Michael Matz <matz@suse.de>
2914 * tree-pass.h (PROP_loop_opts_done): New.
2915 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
2916 Intersect with global range.
2917 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
2918 * tree-ssa-threadbackward.c
2919 (back_threader_profitability::profitable_path_p): Disable
2920 threading through latches until after loop optimizations have run.
2922 2021-09-10 David Faust <david.faust@oracle.com>
2924 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
2927 2021-09-10 David Faust <david.faust@oracle.com>
2929 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
2930 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
2931 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
2932 (bpf_expand_cbranch): New function.
2933 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
2935 (zero_extendsidi2): Only use mov32 instruction if it is available.
2936 (SIM mode iterator): Conditionalize support for SI mode.
2937 (JM mode iterator): New.
2938 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
2939 (*branch_on_di): Update name, use new JM iterator.
2940 * config/bpf/bpf.opt: (mjmpext): New option.
2944 (bpf_isa): New enum.
2946 2021-09-10 David Faust <david.faust@oracle.com>
2948 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
2949 for register-to-register extensions.
2950 (zero_extendqidi2): Likewise.
2952 2021-09-10 Richard Biener <rguenther@suse.de>
2954 PR middle-end/102273
2955 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
2957 2021-09-10 Richard Biener <rguenther@suse.de>
2959 PR middle-end/102269
2960 * gimplify.c (is_var_need_auto_init): Empty types do not need
2963 2021-09-10 Richard Biener <rguenther@suse.de>
2965 * configure.ac (--with-stabs): Remove.
2966 * configure: Regenerate.
2967 * doc/install.texi: Remove --with-stabs documentation.
2969 2021-09-10 liuhongt <hongtao.liu@intel.com>
2971 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
2973 (_mm512_mask_cmp_ph_mask): Likewise.
2974 (_mm512_cmp_round_ph_mask): Likewise.
2975 (_mm512_mask_cmp_round_ph_mask): Likewise.
2976 (_mm_cmp_sh_mask): Likewise.
2977 (_mm_mask_cmp_sh_mask): Likewise.
2978 (_mm_cmp_round_sh_mask): Likewise.
2979 (_mm_mask_cmp_round_sh_mask): Likewise.
2980 (_mm_comieq_sh): Likewise.
2981 (_mm_comilt_sh): Likewise.
2982 (_mm_comile_sh): Likewise.
2983 (_mm_comigt_sh): Likewise.
2984 (_mm_comige_sh): Likewise.
2985 (_mm_comineq_sh): Likewise.
2986 (_mm_ucomieq_sh): Likewise.
2987 (_mm_ucomilt_sh): Likewise.
2988 (_mm_ucomile_sh): Likewise.
2989 (_mm_ucomigt_sh): Likewise.
2990 (_mm_ucomige_sh): Likewise.
2991 (_mm_ucomineq_sh): Likewise.
2992 (_mm_comi_round_sh): Likewise.
2993 (_mm_comi_sh): Likewise.
2994 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
2995 (_mm_mask_cmp_ph_mask): Likewise.
2996 (_mm256_cmp_ph_mask): Likewise.
2997 (_mm256_mask_cmp_ph_mask): Likewise.
2998 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2999 * config/i386/i386-builtin.def: Add corresponding new builtins.
3000 * config/i386/i386-expand.c
3001 (ix86_expand_args_builtin): Handle new builtin types.
3002 (ix86_expand_round_builtin): Ditto.
3003 * config/i386/i386.md (ssevecmode): Add HF mode.
3004 (MODEFH): New mode iterator.
3005 * config/i386/sse.md
3006 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3007 Ajdust corresponding description.
3008 (ssecmpintprefix): New.
3009 (VI12_AVX512VL): Adjust to support HF vector modes.
3010 (cmp_imm_predicate): Likewise.
3011 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3013 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3014 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3015 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3017 2021-09-10 liuhongt <hongtao.liu@intel.com>
3019 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3020 (_mm512_mask_max_ph): Likewise.
3021 (_mm512_maskz_max_ph): Likewise.
3022 (_mm512_min_ph): Likewise.
3023 (_mm512_mask_min_ph): Likewise.
3024 (_mm512_maskz_min_ph): Likewise.
3025 (_mm512_max_round_ph): Likewise.
3026 (_mm512_mask_max_round_ph): Likewise.
3027 (_mm512_maskz_max_round_ph): Likewise.
3028 (_mm512_min_round_ph): Likewise.
3029 (_mm512_mask_min_round_ph): Likewise.
3030 (_mm512_maskz_min_round_ph): Likewise.
3031 (_mm_max_sh): Likewise.
3032 (_mm_mask_max_sh): Likewise.
3033 (_mm_maskz_max_sh): Likewise.
3034 (_mm_min_sh): Likewise.
3035 (_mm_mask_min_sh): Likewise.
3036 (_mm_maskz_min_sh): Likewise.
3037 (_mm_max_round_sh): Likewise.
3038 (_mm_mask_max_round_sh): Likewise.
3039 (_mm_maskz_max_round_sh): Likewise.
3040 (_mm_min_round_sh): Likewise.
3041 (_mm_mask_min_round_sh): Likewise.
3042 (_mm_maskz_min_round_sh): Likewise.
3043 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3044 (_mm256_max_ph): Likewise.
3045 (_mm_mask_max_ph): Likewise.
3046 (_mm256_mask_max_ph): Likewise.
3047 (_mm_maskz_max_ph): Likewise.
3048 (_mm256_maskz_max_ph): Likewise.
3049 (_mm_min_ph): Likewise.
3050 (_mm256_min_ph): Likewise.
3051 (_mm_mask_min_ph): Likewise.
3052 (_mm256_mask_min_ph): Likewise.
3053 (_mm_maskz_min_ph): Likewise.
3054 (_mm256_maskz_min_ph): Likewise.
3055 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3056 * config/i386/i386-builtin.def: Add corresponding new builtins.
3057 * config/i386/i386-expand.c
3058 (ix86_expand_args_builtin): Handle new builtin types.
3059 * config/i386/sse.md
3060 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3061 support HF vector modes.
3062 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3063 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3065 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3067 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3068 Adjust for HF vector modes.
3070 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3072 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3073 (_mm_mask_add_sh): Likewise.
3074 (_mm_maskz_add_sh): Likewise.
3075 (_mm_sub_sh): Likewise.
3076 (_mm_mask_sub_sh): Likewise.
3077 (_mm_maskz_sub_sh): Likewise.
3078 (_mm_mul_sh): Likewise.
3079 (_mm_mask_mul_sh): Likewise.
3080 (_mm_maskz_mul_sh): Likewise.
3081 (_mm_div_sh): Likewise.
3082 (_mm_mask_div_sh): Likewise.
3083 (_mm_maskz_div_sh): Likewise.
3084 (_mm_add_round_sh): Likewise.
3085 (_mm_mask_add_round_sh): Likewise.
3086 (_mm_maskz_add_round_sh): Likewise.
3087 (_mm_sub_round_sh): Likewise.
3088 (_mm_mask_sub_round_sh): Likewise.
3089 (_mm_maskz_sub_round_sh): Likewise.
3090 (_mm_mul_round_sh): Likewise.
3091 (_mm_mask_mul_round_sh): Likewise.
3092 (_mm_maskz_mul_round_sh): Likewise.
3093 (_mm_div_round_sh): Likewise.
3094 (_mm_mask_div_round_sh): Likewise.
3095 (_mm_maskz_div_round_sh): Likewise.
3096 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3097 * config/i386/i386-builtin.def: Add corresponding new builtins.
3098 * config/i386/i386-expand.c
3099 (ix86_expand_round_builtin): Handle new builtins.
3100 * config/i386/sse.md (VF_128): Change description.
3101 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3102 Adjust to support HF vector modes.
3103 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3106 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3108 * config/i386/i386-expand.c
3109 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3110 * config/i386/i386.c
3111 (ix86_preferred_simd_mode): Handle HF mode.
3112 * config/i386/sse.md (V_256H): New mode iterator.
3113 (avx_vextractf128<mode>): Use it.
3114 (VEC_INIT_MODE): Align vector HFmode condition to vector
3115 HImodes since there're no real HF instruction used.
3116 (VEC_INIT_HALF_MODE): Ditto.
3118 (VIHF_AVX512BW): Ditto.
3119 (*vec_extracthf): Ditto.
3120 (VEC_EXTRACT_MODE): Ditto.
3122 2021-09-10 Richard Biener <rguenther@suse.de>
3125 * config/dbx.h: Remove.
3126 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3127 * config/lynx.h: Likewise.
3129 2021-09-10 liuhongt <hongtao.liu@intel.com>
3131 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3132 right into ANDNOT + AND + IOR, using paradoxical subregs.
3133 (ix86_split_copysign_const): Remove.
3134 (ix86_split_copysign_var): Ditto.
3135 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3136 (ix86_split_copysign_var): Ditto.
3137 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3138 (@copysign<mode>3_var): Ditto.
3140 2021-09-09 qing zhao <qing.zhao@oracle.com>
3142 * builtins.c (expand_builtin_memset): Make external visible.
3143 * builtins.h (expand_builtin_memset): Declare extern.
3144 * common.opt (ftrivial-auto-var-init=): New option.
3145 * doc/extend.texi: Document the uninitialized attribute.
3146 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3147 * flag-types.h (enum auto_init_type): New enumerated type
3149 * gimple-fold.c (clear_padding_type): Add one new parameter.
3150 (clear_padding_union): Likewise.
3151 (clear_padding_emit_loop): Likewise.
3152 (clear_type_padding_in_mask): Likewise.
3153 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3154 * gimplify.c (gimple_add_init_for_auto_var): New function.
3155 (gimple_add_padding_init_for_auto_var): New function.
3156 (is_var_need_auto_init): New function.
3157 (gimplify_decl_expr): Add initialization to automatic variables per
3159 (gimplify_call_expr): Add one new parameter for call to
3160 __builtin_clear_padding.
3161 (gimplify_init_constructor): Add padding initialization in the end.
3162 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3163 (expand_DEFERRED_INIT): New function.
3164 * internal-fn.def (DEFERRED_INIT): New internal function.
3165 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3166 * tree-sra.c (generate_subtree_deferred_init): New function.
3167 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3168 calls to .DEFERRED_INIT.
3169 (sra_modify_deferred_init): New function.
3170 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3171 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3172 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3174 (check_defs): Likewise.
3175 (warn_uninitialized_vars): Likewise.
3176 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3177 * tree.c (build_common_builtin_nodes): Build tree node for
3178 BUILT_IN_CLEAR_PADDING when needed.
3180 2021-09-09 Richard Biener <rguenther@suse.de>
3182 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3185 2021-09-09 Richard Biener <rguenther@suse.de>
3187 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3188 DOM walk from get_loop_body_in_dom_order using a worklist
3191 2021-09-09 liuhongt <hongtao.liu@intel.com>
3193 * config.gcc: Add avx512fp16vlintrin.h.
3194 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3195 (_mm512_mask_add_ph): Likewise.
3196 (_mm512_maskz_add_ph): Likewise.
3197 (_mm512_sub_ph): Likewise.
3198 (_mm512_mask_sub_ph): Likewise.
3199 (_mm512_maskz_sub_ph): Likewise.
3200 (_mm512_mul_ph): Likewise.
3201 (_mm512_mask_mul_ph): Likewise.
3202 (_mm512_maskz_mul_ph): Likewise.
3203 (_mm512_div_ph): Likewise.
3204 (_mm512_mask_div_ph): Likewise.
3205 (_mm512_maskz_div_ph): Likewise.
3206 (_mm512_add_round_ph): Likewise.
3207 (_mm512_mask_add_round_ph): Likewise.
3208 (_mm512_maskz_add_round_ph): Likewise.
3209 (_mm512_sub_round_ph): Likewise.
3210 (_mm512_mask_sub_round_ph): Likewise.
3211 (_mm512_maskz_sub_round_ph): Likewise.
3212 (_mm512_mul_round_ph): Likewise.
3213 (_mm512_mask_mul_round_ph): Likewise.
3214 (_mm512_maskz_mul_round_ph): Likewise.
3215 (_mm512_div_round_ph): Likewise.
3216 (_mm512_mask_div_round_ph): Likewise.
3217 (_mm512_maskz_div_round_ph): Likewise.
3218 * config/i386/avx512fp16vlintrin.h: New header.
3219 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3220 Add new builtin types.
3221 * config/i386/i386-builtin.def: Add corresponding builtins.
3222 * config/i386/i386-expand.c
3223 (ix86_expand_args_builtin): Handle new builtin types.
3224 (ix86_expand_round_builtin): Likewise.
3225 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3226 * config/i386/sse.md (VFH): New mode_iterator.
3228 (avx512fmaskmode): Add HF vector modes.
3229 (avx512fmaskhalfmode): Likewise.
3230 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3232 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3233 (mul<mode>3<mask_name><round_name>): Likewise.
3234 (*mul<mode>3<mask_name><round_name>): Likewise.
3235 (div<mode>3): Likewise.
3236 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3237 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3238 (SUBST_A): Likewise.
3239 (round_mode512bit_condition): Adjust for V32HFmode.
3241 2021-09-09 liuhongt <hongtao.liu@intel.com>
3244 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3245 (reduc_plus_scal_v4sf): .. this, New define_expand.
3246 (reduc_plus_scal_v2df): .. and this, New define_expand.
3248 2021-09-09 liuhongt <hongtao.liu@intel.com>
3251 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3254 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3257 * doc/trouble.texi (Copy Assignment): Fix description of
3258 behaviour and fix code in example.
3260 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3263 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3264 r11 instead of r12 for restoring CR.
3266 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3267 liuhongt <hongtao.liu@intel.com>
3270 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3271 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3272 into AND with mask and XOR, using paradoxical subregs.
3273 (ix86_split_xorsign): Remove.
3274 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3276 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3278 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3280 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3283 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3284 operands[2], emit abs<mode>2 instead.
3285 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3286 first alternative even for avx, add another alternative with
3287 =&Yv <- 0, Yv, Yvm constraints.
3288 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3289 to op1, emit vpandn instead.
3291 2021-09-08 liuhongt <hongtao.liu@intel.com>
3293 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3294 (_mm256_set_ph): Likewise.
3295 (_mm512_set_ph): Likewise.
3296 (_mm_setr_ph): Likewise.
3297 (_mm256_setr_ph): Likewise.
3298 (_mm512_setr_ph): Likewise.
3299 (_mm_set1_ph): Likewise.
3300 (_mm256_set1_ph): Likewise.
3301 (_mm512_set1_ph): Likewise.
3302 (_mm_setzero_ph): Likewise.
3303 (_mm256_setzero_ph): Likewise.
3304 (_mm512_setzero_ph): Likewise.
3305 (_mm_set_sh): Likewise.
3306 (_mm_load_sh): Likewise.
3307 (_mm_store_sh): Likewise.
3308 * config/i386/i386-builtin-types.def (V8HF): New type.
3309 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3310 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3311 Support vector HFmodes.
3312 (ix86_expand_vector_init_one_nonzero): Likewise.
3313 (ix86_expand_vector_init_one_var): Likewise.
3314 (ix86_expand_vector_init_interleave): Likewise.
3315 (ix86_expand_vector_init_general): Likewise.
3316 (ix86_expand_vector_set): Likewise.
3317 (ix86_expand_vector_extract): Likewise.
3318 (ix86_expand_vector_init_concat): Likewise.
3319 (ix86_expand_sse_movcc): Handle vector HFmodes.
3320 (ix86_expand_vector_set_var): Ditto.
3321 * config/i386/i386-modes.def: Add HF vector modes in comment.
3322 * config/i386/i386.c (classify_argument): Add HF vector modes.
3323 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3324 (ix86_vector_mode_supported_p): Likewise.
3325 (ix86_set_reg_reg_cost): Handle vector HFmode.
3326 (ix86_get_ssemov): Handle vector HFmode.
3327 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3329 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3331 (function_arg_advance_32): Ditto.
3332 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
3333 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
3334 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
3335 (VALID_SSE2_REG_VHF_MODE): New.
3336 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
3337 (SSE_REG_MODE_P): Add vector HFmode.
3338 * config/i386/i386.md (mode): Add HF vector modes.
3339 (MODE_SIZE): Likewise.
3340 (ssemodesuffix): Add ph suffix for HF vector modes.
3341 * config/i386/sse.md (VFH_128): New mode iterator.
3342 (VMOVE): Adjust for HF vector modes.
3344 (V_256_512): Likewise.
3346 (avx512fmaskmode): Likewise.
3347 (shuffletype): Likewise.
3348 (sseinsnmode): Likewise.
3349 (ssedoublevecmode): Likewise.
3350 (ssehalfvecmode): Likewise.
3351 (ssehalfvecmodelower): Likewise.
3352 (ssePScmode): Likewise.
3353 (ssescalarmode): Likewise.
3354 (ssescalarmodelower): Likewise.
3355 (sseintprefix): Likewise.
3357 (bcstscalarsuff): Likewise.
3358 (xtg_mode): Likewise.
3359 (VI12HF_AVX512VL): New mode_iterator.
3360 (VF_AVX512FP16): Likewise.
3362 (VIHF_256): Likewise.
3363 (VIHF_AVX512BW): Likewise.
3364 (V16_256): Likewise.
3365 (V32_512): Likewise.
3366 (sseintmodesuffix): New mode_attr.
3367 (sse): Add scalar and vector HFmodes.
3368 (ssescalarmode): Add vector HFmode mapping.
3369 (ssescalarmodesuffix): Add sh suffix for HFmode.
3370 (*<sse>_vm<insn><mode>3): Use VFH_128.
3371 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
3372 (*ieee_<ieee_maxmin><mode>3): Likewise.
3373 (<avx512>_blendm<mode>): New define_insn.
3374 (vec_setv8hf): New define_expand.
3375 (vec_set<mode>_0): New define_insn for HF vector set.
3376 (*avx512fp16_movsh): Likewise.
3377 (avx512fp16_movsh): Likewise.
3378 (vec_extract_lo_v32hi): Rename to ...
3379 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
3381 (vec_extract_hi_v32hi): Likewise.
3382 (vec_extract_hi_<mode>): Likewise.
3383 (vec_extract_lo_v16hi): Likewise.
3384 (vec_extract_lo_<mode>): Likewise.
3385 (vec_extract_hi_v16hi): Likewise.
3386 (vec_extract_hi_<mode>): Likewise.
3387 (vec_set_hi_v16hi): Likewise.
3388 (vec_set_hi_<mode>): Likewise.
3389 (vec_set_lo_v16hi): Likewise.
3390 (vec_set_lo_<mode>): Likewise.
3391 (*vec_extract<mode>_0): New define_insn_and_split for HF
3393 (*vec_extracthf): New define_insn.
3394 (VEC_EXTRACT_MODE): Add HF vector modes.
3395 (PINSR_MODE): Add V8HF.
3396 (sse2p4_1): Likewise.
3397 (pinsr_evex_isa): Likewise.
3398 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
3399 insert for V8HFmode.
3400 (pbroadcast_evex_isa): Add HF vector modes.
3401 (AVX2_VEC_DUP_MODE): Likewise.
3402 (VEC_INIT_MODE): Likewise.
3403 (VEC_INIT_HALF_MODE): Likewise.
3404 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
3406 (avx2_pbroadcast<mode>_1): Likewise.
3407 (<avx512>_vec_dup<mode>_1): Likewise.
3408 (<avx512>_vec_dup<mode><mask_name>): Likewise.
3409 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
3412 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
3413 H.J. Lu <hongjiu.lu@intel.com>
3414 Liu Hongtao <hongtao.liu@intel.com>
3415 Wang Hongyu <hongyu.wang@intel.com>
3416 Xu Dianhong <dianhong.xu@intel.com>
3418 * common/config/i386/cpuinfo.h (get_available_features):
3419 Detect FEATURE_AVX512FP16.
3420 * common/config/i386/i386-common.c
3421 (OPTION_MASK_ISA_AVX512FP16_SET,
3422 OPTION_MASK_ISA_AVX512FP16_UNSET,
3423 OPTION_MASK_ISA2_AVX512FP16_SET,
3424 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
3425 (OPTION_MASK_ISA2_AVX512BW_UNSET,
3426 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
3427 (ix86_handle_option): Handle -mavx512fp16.
3428 * common/config/i386/i386-cpuinfo.h (enum processor_features):
3429 Add FEATURE_AVX512FP16.
3430 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
3431 * config.gcc: Add avx512fp16intrin.h.
3432 * config/i386/avx512fp16intrin.h: New intrinsic header.
3433 * config/i386/cpuid.h: Add bit_AVX512FP16.
3434 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
3435 * config/i386/i386-builtins.c: Support _Float16 type for i386
3437 (ix86_register_float16_builtin_type): New function.
3438 (ix86_float16_type_node): New.
3439 * config/i386/i386-c.c (ix86_target_macros_internal): Define
3441 * config/i386/i386-expand.c (ix86_expand_branch): Support
3443 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
3444 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3445 (ix86_expand_fp_movcc): Ditto.
3446 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
3447 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
3448 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
3449 * config/i386/i386.c (ix86_get_ssemov): Use
3450 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
3451 (ix86_get_excess_precision): Use
3452 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
3454 (sse_store_index): Use SFmode cost for HFmode cost.
3455 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
3456 GPR cost for HFmode.
3457 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
3458 (ix86_mangle_type): Add manlging for _Float16 type.
3459 (inline_secondary_memory_needed): No memory is needed for
3460 16bit movement between gpr and sse reg under
3462 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
3463 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3464 (ix86_division_cost): Ditto.
3465 (ix86_rtx_costs): Ditto.
3466 (ix86_add_stmt_cost): Ditto.
3467 (ix86_optab_supported_p): Ditto.
3468 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
3469 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
3470 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
3471 * config/i386/i386.md (mode): Add HFmode.
3472 (MODE_SIZE): Add HFmode.
3473 (isa): Add avx512fp16.
3474 (enabled): Handle avx512fp16.
3475 (ssemodesuffix): Add sh suffix for HFmode.
3476 (comm): Add mult, div.
3477 (plusminusmultdiv): New code iterator.
3478 (insn): Add mult, div.
3479 (*movhf_internal): Adjust for avx512fp16 instruction.
3480 (*movhi_internal): Ditto.
3481 (*cmpi<unord>hf): New define_insn for HFmode.
3482 (*ieee_s<ieee_maxmin>hf3): Likewise.
3483 (extendhf<mode>2): Likewise.
3484 (trunc<mode>hf2): Likewise.
3485 (float<floatunssuffix><mode>hf2): Likewise.
3486 (*<insn>hf): Likewise.
3487 (cbranchhf4): New expander.
3488 (movhfcc): Likewise.
3489 (<insn>hf3): Likewise.
3492 * config/i386/i386.opt: Add mavx512fp16.
3493 * config/i386/immintrin.h: Include avx512fp16intrin.h.
3494 * doc/invoke.texi: Add mavx512fp16.
3495 * doc/extend.texi: Add avx512fp16 Usage Notes.
3497 2021-09-08 liuhongt <hongtao.liu@intel.com>
3499 * common.opt: Support -fexcess-precision=16.
3500 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
3501 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
3502 EXCESS_PRECISION_TYPE_FLOAT16.
3503 * config/arm/arm.c (arm_excess_precision): Ditto.
3504 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
3505 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
3506 when EXCESS_PRECISION_TYPE_FLOAT16.
3507 * config/s390/s390.c (s390_excess_precision): Ditto.
3508 * coretypes.h (enum excess_precision_type): Add
3509 EXCESS_PRECISION_TYPE_FLOAT16.
3510 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
3511 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
3512 * doc/extend.texi (Half-Precision): Document
3513 -fexcess-precision=16.
3514 * flag-types.h (enum excess_precision): Add
3515 EXCESS_PRECISION_FLOAT16.
3516 * target.def (excess_precision): Update document.
3517 * tree.c (excess_precision_type): Set excess_precision_type to
3518 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
3520 2021-09-08 liuhongt <hongtao.liu@intel.com>
3522 * doc/extend.texi: (@node Floating Types): Adjust the wording.
3523 (@node Half-Precision): Ditto.
3525 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
3528 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
3529 'CONST_INT_P (src)' to the condition of the block that tries to
3530 eliminate literal when loading integer contant.
3532 2021-09-07 David Faust <david.faust@oracle.com>
3534 * doc/extend.texi (BPF Type Attributes) New node.
3535 Document new preserve_access_index attribute.
3536 Document new preserve_access_index builtin.
3537 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
3539 2021-09-07 David Faust <david.faust@oracle.com>
3541 * config/bpf/bpf.c: Adjust includes.
3542 (bpf_handle_preserve_access_index_attribute): New function.
3543 (bpf_attribute_table): Use it here.
3544 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
3545 (bpf_option_override): Handle "-mco-re" option.
3546 (bpf_asm_init_sections): New.
3547 (TARGET_ASM_INIT_SECTIONS): Redefine.
3548 (bpf_file_end): New.
3549 (TARGET_ASM_FILE_END): Redefine.
3550 (bpf_init_builtins): Add "__builtin_preserve_access_index".
3551 (bpf_core_compute, bpf_core_get_index): New.
3552 (is_attr_preserve_access): New.
3553 (bpf_expand_builtin): Handle new builtins.
3554 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
3555 (bpf_core_walk): New.
3556 (bpf_resolve_overloaded_builtin): New.
3557 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
3559 (pass_bpf_core_attr): New RTL pass.
3560 * config/bpf/bpf-passes.def: New file.
3561 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
3562 * config/bpf/coreout.c: New file.
3563 * config/bpf/coreout.h: Likewise.
3564 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
3565 (coreout.o): New rule.
3566 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
3567 * config.gcc (bpf): Add coreout.h to extra_headers.
3568 Add coreout.o to extra_objs.
3569 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
3571 2021-09-07 David Faust <david.faust@oracle.com>
3573 * btfout.c (get_btf_id): Function is no longer static.
3574 * ctfc.h: Expose it here.
3576 2021-09-07 David Faust <david.faust@oracle.com>
3578 * ctfc.c (ctf_lookup_tree_type): New function.
3581 2021-09-07 David Faust <david.faust@oracle.com>
3583 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
3584 * ctfc.h: Analogous change.
3586 2021-09-07 David Faust <david.faust@oracle.com>
3588 * dwarf2out.c (lookup_type_die): Function is no longer static.
3589 * dwarf2out.h: Expose it here.
3591 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3593 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
3594 (ctf_debug_early_finish): New definition.
3595 (ctf_debug_finish): Likewise.
3596 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
3597 (ctf_debug_early_finish): New declaration.
3598 (ctf_debug_finish): Likewise.
3599 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
3600 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
3602 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3604 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
3605 support when compiling for CO-RE.
3606 * config/bpf/bpf.opt: Add new command line option -mco-re.
3608 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3610 * flag-types.h (enum debug_info_type): Add new enum
3611 DINFO_TYPE_BTF_WITH_CORE.
3612 (BTF_WITH_CORE_DEBUG): New bitmask.
3613 * flags.h (btf_with_core_debuginfo_p): New declaration.
3614 * opts.c (btf_with_core_debuginfo_p): New definition.
3616 2021-09-07 Jason Merrill <jason@redhat.com>
3618 * tree.h (error_operand_p): Change to inline function.
3620 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
3622 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
3623 (empty_block_with_phis_p): ...this.
3624 (potentially_threadable_block): Same.
3625 (jump_threader::thread_through_normal_block): Same.
3627 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
3630 * dwarf2out.c (mark_base_types): New overloaded function.
3631 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
3632 as the compilation unit, and call move_marked_base_types afterward.
3634 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
3637 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
3639 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
3641 2021-09-07 Richard Biener <rguenther@suse.de>
3643 PR tree-optimization/102226
3644 * tree-vect-loop.c (vect_transform_cycle_phi): Record
3645 the converted value for the epilogue PHI use.
3647 2021-09-07 Martin Liska <mliska@suse.cz>
3649 PR gcov-profile/80223
3650 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
3651 options, do not inline when no_profile_instrument_function
3652 attributes are different in early inliner. It's fine to inline
3653 it after PGO instrumentation.
3655 2021-09-07 Richard Biener <rguenther@suse.de>
3657 PR tree-optimization/101555
3658 * tree-ssa-pre.c (translate_vuse_through_block): Do not
3659 perform an alias walk to determine the validity of the
3660 mem at the start of the block which is already guaranteed
3661 by means of prune_clobbered_mems.
3662 (phi_translate_1): Pass edge to translate_vuse_through_block.
3664 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
3667 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
3668 (remainder<mode>3): Likewise.
3670 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
3672 * config/mips/mips.c (mips_file_start): add .module for
3675 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
3677 * wide-int.cc (wi::clz): Reorder tests to ensure the result
3678 is zero for all negative values.
3680 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
3682 * doc/invoke.texi (-foffload-options): Fix @opindex.
3684 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
3687 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
3688 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
3689 source alternative for AVX.
3691 2021-09-06 liuhongt <hongtao.liu@intel.com>
3693 PR middle-end/102182
3694 * optabs.c (expand_fix): Add from1 to avoid from being
3697 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
3699 * dwarf2out.c (modified_type_die): Deal with all array types earlier
3700 and use local variable consistently throughout the function.
3702 2021-09-06 Jakub Jelinek <jakub@redhat.com>
3704 PR tree-optimization/102207
3705 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
3706 were promoted from signed to wider unsigned type.
3708 2021-09-06 Andrew Pinski <apinski@marvell.com>
3710 PR tree-optimization/63184
3711 * match.pd: Add simplification of pointer_diff of two pointer_plus
3712 with addr_expr in the first operand of each pointer_plus.
3713 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
3714 in the first operand of each pointer_plus.
3716 2021-09-06 Richard Biener <rguenther@suse.de>
3718 PR tree-optimization/102176
3719 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
3721 (vect_bb_slp_scalar_cost): Use the computed set of
3722 vectorized scalar stmts instead of relying on the out-of-date
3723 and not accurate PURE_SLP_STMT.
3724 (vect_bb_vectorization_profitable_p): Compute the set
3725 of vectorized scalar stmts.
3727 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
3729 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
3730 GIMPLE_COND special casing.
3731 (path_range_query::range_defined_in_block): Use range_of_stmt
3732 instead of calling fold_range directly.
3734 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
3736 * gimple-range-path.cc (path_range_query::range_of_expr): Set
3737 m_undefined_path when appropriate.
3738 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
3739 (path_range_query::unreachable_path_p): New.
3740 (path_range_query::precompute_ranges): Set m_undefined_path.
3741 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
3742 (path_range_query::internal_range_of_expr): New.
3743 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
3744 Use unreachable_path_p.
3746 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
3748 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
3749 Remove argument and call find_taken_edge.
3750 (back_threader::resolve_phi): Do not calculate taken edge before
3751 calling maybe_register_path.
3752 (back_threader::find_paths_to_names): Same.
3754 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
3756 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
3757 * config/h8300/testcompare.md (store_c): Update name, use new
3759 (store_neg_c, store_shifted_c): New patterns.
3761 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
3764 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
3765 instead of r12 for CR save, in all cases.
3767 2021-09-03 Andrew Pinski <apinski@marvell.com>
3769 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
3770 Handle error_mark_node as the type of the type_decl.
3772 2021-09-03 Andrew Pinski <apinski@marvell.com>
3774 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
3776 (aarch64_simd_types): Likewise.
3777 (aarch64_simd_intOI_type_node): Likewise.
3778 (aarch64_simd_intCI_type_node): Likewise.
3779 (aarch64_simd_intXI_type_node): Likewise.
3780 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
3781 (aarch64_fp16_ptr_type_node): Likewise.
3782 (aarch64_bf16_type_node): Likewise.
3783 (aarch64_bf16_ptr_type_node): Likewise.
3785 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3787 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
3789 (minus_op1_op2_relation_effect): ...here.
3790 (class operator_pointer_diff): New.
3791 (operator_pointer_diff::op1_op2_relation_effect): Call
3792 minus_op1_op2_relation_effect.
3793 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
3795 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3797 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
3798 Add may_peel_loop_headers.
3799 (back_threader_registry::thread_through_all_blocks): Same.
3800 (try_thread_blocks): Pass may_peel_loop_headers argument.
3801 (pass_early_thread_jumps::execute): Same.
3803 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3805 * tree-ssa-threadedge.c (has_phis_p): New.
3806 (forwarder_block_p): New.
3807 (potentially_threadable_block): Call forwarder_block_p.
3808 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
3809 (jump_threader::thread_through_normal_block): Call
3812 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3814 * tree-ssa-threadbackward.c (back_threader::dump): New.
3815 (back_threader::debug): New.
3816 (back_threader_profitability::profitable_path_p): Dump blocks
3817 even if we are bailing early.
3819 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3821 * tree-ssa-threadupdate.c (cancel_thread): New.
3822 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
3823 (jump_thread_path_registry::mark_threaded_blocks): Same.
3824 (jump_thread_path_registry::register_jump_thread): Same.
3826 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3828 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
3829 which objects are available.
3830 (jt_state::pop): Same.
3831 (jt_state::register_equiv): Same.
3832 (jt_state::register_equivs_on_edge): Same.
3834 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3836 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
3837 Move pop until after a thread is registered.
3839 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3841 * tree-ssa-threadupdate.c (debug): New.
3843 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3845 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
3846 (push_dump_file::~push_dump_file): New.
3847 (dump_ranger): Change dump_file temporarily while dumping
3849 * gimple-range-trace.h (class push_dump_file): New.
3851 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3853 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
3854 (dump_ranger): Dump function name.
3856 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3858 * gimple-range-path.cc (path_range_query::range_defined_in_block):
3859 Adjust for non-null.
3860 (path_range_query::adjust_for_non_null_uses): New.
3861 (path_range_query::precompute_ranges): Call
3862 adjust_for_non_null_uses.
3863 * gimple-range-path.h: Add m_non_null and
3864 adjust_for_non_null_uses.
3866 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3868 * gimple-range-path.cc (path_range_query::dump): Dump path
3870 (path_range_query::precompute_ranges): Dump entire path.
3872 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3874 * value-relation.cc (relation_oracle::debug): New.
3875 * value-relation.h (relation_oracle::debug): New.
3877 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3879 * tree-ssa-loop-ch.c: Remove unnecessary include file.
3881 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3883 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
3884 Skip statements with no defining BB.
3885 * gimple-range-path.cc (path_range_query::range_defined_in_block):
3886 Do not get confused by statements with no defining BB.
3888 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3890 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
3891 gimple_range_adjustment. Add support for constants.
3892 (adjust_realpart_expr): New.
3893 (gimple_range_adjustment): Move IMAGPART_EXPR code to
3894 adjust_imagpart_expr.
3895 * range-op.cc (integral_table::integral_table): Add entry for
3898 2021-09-03 Jakub Jelinek <jakub@redhat.com>
3900 * omp-expand.c (expand_omp_atomic_pipeline): Use
3901 IFN_ATOMIC_COMPARE_EXCHANGE instead of
3902 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
3905 2021-09-03 Jakub Jelinek <jakub@redhat.com>
3908 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
3909 that is false if DECL_BIT_FIELD.
3910 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
3911 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
3912 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
3913 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
3914 of DECL_FIELD_ABI_IGNORED.
3915 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
3916 SET_DECL_FIELD_ABI_IGNORED instead of writing to
3917 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
3918 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
3919 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
3920 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
3922 2021-09-03 liuhongt <hongtao.liu@intel.com>
3925 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
3926 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
3927 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
3929 2021-09-02 Martin Sebor <msebor@redhat.com>
3931 PR tree-optimization/17506
3933 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
3935 2021-09-02 Richard Biener <rguenther@suse.de>
3937 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
3938 fix for PR78185 and continue processing when leaving
3941 2021-09-02 Jakub Jelinek <jakub@redhat.com>
3943 PR tree-optimization/99591
3944 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
3947 2021-09-02 Richard Biener <rguenther@suse.de>
3950 2021-09-02 Richard Biener <rguenther@suse.de>
3952 PR tree-optimization/102155
3953 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
3954 over a part of the RPO array and do not recurse here.
3955 Dump blocks marked as always executed.
3956 (fill_always_executed_in): Walk over the RPO array and
3957 process loops whose header we run into.
3958 (loop_invariant_motion_in_fun): Compute the first RPO
3959 using rev_post_order_and_mark_dfs_back_seme in iteration
3960 order and pass that to fill_always_executed_in.
3962 2021-09-02 liuhongt <hongtao.liu@intel.com>
3964 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
3965 * config/i386/i386.c (enum x86_64_reg_class): Add
3967 (merge_classes): Handle X86_64_SSEHF_CLASS.
3968 (examine_argument): Ditto.
3969 (construct_container): Ditto.
3970 (classify_argument): Ditto, and set HFmode/HCmode to
3972 (function_value_32): Return _FLoat16/Complex Float16 by
3974 (function_value_64): Return _Float16/Complex Float16 by SSE
3976 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
3977 (ix86_secondary_reload): Require gpr as intermediate register
3978 to store _Float16 from sse register when sse4 is not
3980 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
3982 (ix86_scalar_mode_supported_p): Ditto.
3983 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
3984 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
3985 (VALID_INT_MODE_P): Add HFmode and HCmode.
3986 * config/i386/i386.md (*pushhf_rex64): New define_insn.
3988 (*movhf_internal): Ditto.
3989 * doc/extend.texi (Half-Precision Floating Point): Documemt
3992 2021-09-02 Richard Biener <rguenther@suse.de>
3994 PR tree-optimization/102155
3995 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
3996 over a part of the RPO array and do not recurse here.
3997 Dump blocks marked as always executed.
3998 (fill_always_executed_in): Walk over the RPO array and
3999 process loops whose header we run into.
4000 (loop_invariant_motion_in_fun): Compute the first RPO
4001 using rev_post_order_and_mark_dfs_back_seme in iteration
4002 order and pass that to fill_always_executed_in.
4004 2021-09-02 YunQiang Su <syq@debian.org>
4007 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4009 * config/mips/mips.c (mips_module_isa_name): New.
4010 mips_file_start: add .module mipsREV to all asm output
4012 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4014 PR tree-optimization/102152
4015 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4016 comparison to a scalar comparison before calling
4017 update_stmt_if_modified.
4019 2021-09-01 Andrew Pinski <apinski@marvell.com>
4022 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4023 Check STRICT_ALIGNMENT before creating an overlapping
4026 2021-09-01 Martin Sebor <msebor@redhat.com>
4028 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4029 (check_access): Pass additional argument.
4030 (check_memop_access): Remove template and make a member function.
4031 (maybe_check_dealloc_call): Make a pass_waccess member function.
4032 (class pass_waccess): Add, rename, and remove members.
4033 (pass_waccess::pass_waccess): Adjust to name change.
4034 (pass_waccess::~pass_waccess): Same.
4035 (check_alloca): Make a member function.
4036 (check_alloc_size_call): Same.
4037 (check_strcat): Same.
4038 (check_strncat): Same.
4039 (check_stxcpy): Same.
4040 (check_stxncpy): Same.
4041 (check_strncmp): Same.
4042 (maybe_warn_rdwr_sizes): Rename...
4043 (pass_waccess::maybe_check_access_sizes): ...to this.
4044 (pass_waccess::check_call): Adjust to name changes.
4045 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4047 (pass_waccess::execute): Adjust to name changes.
4048 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4049 * pointer-query.cc (access_ref::phi): Handle null pointer.
4050 (access_ref::inform_access): Same.
4051 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4052 (pointer_query::dump): New function.
4053 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4055 * pointer-query.h (pointer_query::dump): Declare.
4056 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4057 (dump_strlen_info): Use function query.
4058 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4060 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4062 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4064 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4066 * doc/extend.texi: Document unavailable attribute.
4067 * print-tree.c (print_node): Handle unavailable attribute.
4068 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4069 * tree.c (error_unavailable_use): New.
4070 * tree.h (TREE_UNAVAILABLE): New.
4071 (error_unavailable_use): New.
4073 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4075 PR tree-optimization/102124
4076 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4077 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4078 add an extra cast to signed variant of itype to ensure sign-extension.
4080 2021-09-01 Martin Liska <mliska@suse.cz>
4082 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4083 edges and rather use colors for TRUE and FALSE edges.
4085 2021-09-01 Richard Biener <rguenther@suse.de>
4087 PR tree-optimization/93491
4088 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4089 after processing the stmt itself. Do not consider
4090 pure functions possibly not returning. Properly avoid
4091 adding possibly trapping calls to EXP_GEN when there's
4092 a preceeding possibly not returning call.
4093 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4096 2021-09-01 Richard Biener <rguenther@suse.de>
4098 PR tree-optimization/102139
4099 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4100 type to record a std::pair of the stmt-info and the innermost
4102 (dr_vec_info::group): New member.
4103 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4104 (vect_compute_data_ref_alignment): Verify the recorded
4105 base alignment can be used.
4106 (data_ref_pair): Remove.
4107 (dr_group_sort_cmp): Adjust.
4108 (vect_analyze_data_ref_accesses): Store the group-ID in the
4109 dr_vec_info and operate on a vector of dr_vec_infos.
4111 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4113 * read-md.c (md_reader::handle_enum): support value assignation.
4114 * doc/md.texi: record define_c_enum value assignation support.
4116 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4118 PR tree-optimization/102141
4119 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4120 argument. If false, emit stmts after gsi instead of before, and
4122 (bswap_replace): Adjust callers. When converting output of bswap,
4123 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4126 2021-09-01 liuhongt <hongtao.liu@intel.com>
4128 * emit-rtl.c (validate_subreg): Get rid of all float-int
4131 2021-09-01 liuhongt <hongtao.liu@intel.com>
4134 2021-08-30 liuhongt <hongtao.liu@intel.com>
4136 * expmed.c (extract_bit_field_1): Make sure we're playing with
4137 integral modes before call extract_integral_bit_field.
4138 (extract_integral_bit_field): Add a parameter of type
4139 scalar_int_mode which corresponds to of tmode.
4140 And call extract_and_convert_fixed_bit_field instead of
4141 extract_fixed_bit_field and convert_extracted_bit_field.
4142 (extract_and_convert_fixed_bit_field): New function, it's a
4143 combination of extract_fixed_bit_field and
4144 convert_extracted_bit_field.
4146 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4148 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4150 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4152 * omp-general.h (omp_is_reference): Rename to...
4153 (omp_privatize_by_reference): ... this. Adjust all users...
4154 * omp-general.c: ... here, ...
4155 * gimplify.c: ... here, ...
4156 * omp-expand.c: ... here, ...
4157 * omp-low.c: ... here.
4159 2021-08-31 Martin Sebor <msebor@redhat.com>
4161 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4162 pointer element for equality to zero, not that of the cotaining
4165 2021-08-31 Martin Sebor <msebor@redhat.com>
4167 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4169 2021-08-31 Martin Sebor <msebor@redhat.com>
4171 * function.h (function): Add comments.
4172 (get_range_query): Same. Add attribute returns nonnull.
4174 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4176 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4177 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4178 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4180 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4182 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4184 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4185 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4186 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4187 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4188 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4189 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4190 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4191 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4192 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4193 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4195 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4197 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4198 allowed on target constructs and only with particular other clauses.
4199 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4201 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4202 structs when 'ancestor' is used.
4203 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4205 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4206 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4208 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4210 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4211 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4213 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4214 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4215 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4216 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4217 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4218 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4219 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4220 would be paradoxical.
4222 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4224 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4225 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4226 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4228 2021-08-31 Richard Biener <rguenther@suse.de>
4230 PR tree-optimization/102142
4231 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4232 condition under which to unset the visited flag.
4234 2021-08-31 Richard Biener <rguenther@suse.de>
4236 PR middle-end/102129
4237 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4238 possibly trapping expressions across calls.
4240 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4242 PR tree-optimization/102134
4243 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4244 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4247 2021-08-31 Andrew Pinski <apinski@marvell.com>
4250 * collect-utils.c (setup_signals): New declaration.
4251 * collect-utils.h (setup_signals): New function.
4252 * collect2.c (handler): Delete.
4253 (main): Instead of manually setting up the signals,
4254 just call setup_signals.
4255 * lto-wrapper.c (main): Likewise.
4257 2021-08-31 Andrew Pinski <apinski@marvell.com>
4260 * config/i386/i386-protos.h (x86_output_aligned_bss):
4261 Change align argument to unsigned type.
4262 (x86_elf_aligned_decl_common): Likewise.
4263 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4264 (x86_output_aligned_bss): Likewise.
4266 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4268 * config/mips/mips.c (mips_module_isa_name): New.
4269 mips_file_start: add .module mipsREV to all asm output
4271 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4273 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4274 use enum instead of int for 'isa' member.
4275 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4276 config/mips/netbsd.h: replace hardcoded numbers with enum.
4278 2021-08-31 liuhongt <hongtao.liu@intel.com>
4280 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4281 define_split to define_insn_and_split.
4282 (*avx2_eq<mode>3): Removed.
4283 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4284 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4285 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4287 (*avx2_gt<mode>3): Removed.
4288 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4289 define_insn to define_expand, and adjust pattern.
4290 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4292 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4295 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4297 2021-08-30 Jason Merrill <jason@redhat.com>
4299 * doc/invoke.texi: Document -Wmissing-requires.
4301 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4303 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4304 TARGET_EXTRA_BUILTINS guard.
4306 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4308 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4309 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4311 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4313 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4314 decl when new_builtins_are_live.
4315 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4318 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4320 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4321 OPTION_MASK_P10_FUSION_2STORE.
4322 (POWERPC_MASKS): Likewise.
4323 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4324 store fusion for Power10.
4325 (is_fusable_store): New.
4326 (power10_sched_reorder): Likewise.
4327 (rs6000_sched_reorder): Do Power10 specific reordering.
4328 (rs6000_sched_reorder2): Likewise.
4329 * config/rs6000/rs6000.opt: Add new option.
4331 2021-08-30 Richard Biener <rguenther@suse.de>
4333 PR tree-optimization/102128
4334 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4335 Move scanning for if-converted scalar code to the caller
4336 and instead delay clearing the visited flag for profitable
4338 (vect_slp_region): Cost all subgraphs before scheduling.
4339 For if-converted BB vectorization scan for scalar COND_EXPRs
4340 and do not vectorize if any found and the cost model is
4343 2021-08-30 Richard Biener <rguenther@suse.de>
4345 * common.opt (fexceptions): Mark
4346 EnabledBy(fnon-call-exceptions).
4347 * doc/invoke.texi (fnon-call-exceptions): Document this
4348 enables -fexceptions.
4350 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
4352 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
4353 is not already defined.
4355 2021-08-30 liuhongt <hongtao.liu@intel.com>
4357 * expmed.c (extract_bit_field_1): Make sure we're playing with
4358 integral modes before call extract_integral_bit_field.
4359 (extract_integral_bit_field): Add a parameter of type
4360 scalar_int_mode which corresponds to of tmode.
4361 And call extract_and_convert_fixed_bit_field instead of
4362 extract_fixed_bit_field and convert_extracted_bit_field.
4363 (extract_and_convert_fixed_bit_field): New function, it's a
4364 combination of extract_fixed_bit_field and
4365 convert_extracted_bit_field.
4367 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4369 * config/darwin.c (darwin_libc_has_function): Do not run
4370 the checks for x86 or modern Darwin. Make sure that there
4371 is a value set for darwin_macosx_version_min before testing.
4373 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4375 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
4377 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
4379 * ipa-modref-tree.h (modref_access_node::merge): Break out
4380 logic combining offsets and logic merging ranges to ...
4381 (modref_access_node::combined_offsets): ... here
4382 (modref_access_node::update2): ... here
4383 (modref_access_node::closer_pair_p): New member function.
4384 (modref_access_node::forced_merge): New member function.
4385 (modre_ref_node::insert): Do merging when table is full.
4387 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
4390 * config.gcc: MIPS: use N64 ABI by default if the triple end
4391 with -gnuabi64, which is used by Debian since 2013.
4393 2021-08-28 Alexandre Oliva <oliva@adacore.com>
4395 * ipa-modref.c (analyze_function): Skip debug stmts.
4396 * tree-inline.c (estimate_num_insn): Consider builtins even
4397 without a cgraph_node.
4399 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4401 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
4402 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
4403 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
4404 * config/h8300/h8300.md (eqne): New code iterator.
4405 (geultu, geultu_to_c): Similarly.
4406 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
4407 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
4408 (cmp<mode>_c): New pattern
4410 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4412 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
4414 (dom_opt_dom_walker::optimize_stmt): Use it.
4416 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4418 * config/darwin.c (finalize_ctors): Add a section-start linker-
4420 (finalize_dtors): Likewise.
4421 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
4423 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
4425 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
4426 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
4427 old initialization logic when new builtins are enabled.
4428 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
4429 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
4430 (write_init_file): Likewise.
4432 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4434 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
4435 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
4436 (amdgcn-* | gcn-*) Likewise.
4438 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
4440 * symbol-summary.h: Added missing template keyword.
4442 2021-08-27 Richard Biener <rguenther@suse.de>
4444 PR tree-optimization/45178
4445 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
4446 infinite loops without exit do not mark control dependent
4447 edges of the latch necessary.
4449 2021-08-27 konglin1 <lingling.kong@intel.com>
4452 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
4454 (<avx512>scattersi<mode>): Likewise.
4455 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
4456 (*avx512f_scatterdi<VI48F:mode>): Likewise
4458 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4460 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
4461 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
4462 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
4463 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
4464 P10_BUILTIN_PEXTD on Power10.
4466 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4468 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
4469 signedness for some Power10 bifs.
4471 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
4474 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
4475 computed alignment if the entire struct has attribute packed.
4477 2021-08-27 liuhongt <hongtao.liu@intel.com>
4481 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
4482 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
4483 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
4484 IX86_BUILTIN_SHUFPS256.
4485 (ix86_masked_all_ones): New function.
4487 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4489 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
4490 (conditional moves with memory inputs splitters): Ditto.
4491 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
4493 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4495 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
4496 search after merging.
4498 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4500 * config/rs6000/rs6000-overload.def: Add remaining overloads.
4502 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4504 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
4506 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4508 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
4509 crypto, and htm stanzas.
4511 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4513 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
4515 2021-08-26 Martin Sebor <msebor@redhat.com>
4517 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
4518 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
4519 (warn_uninitialized_vars): Same. Reduce visibility of locals.
4520 (warn_uninitialized_phi): Same.
4522 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4524 * tree-ssa-ccp.c (get_individual_bits): Helper function to
4525 extract the individual bits from a widest_int constant (mask).
4526 (gray_code_bit_flips): New read-only table for effiently
4527 enumerating permutations/combinations of bits.
4528 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
4529 by unknown counts that are guaranteed less than the target
4530 precision and four or fewer unknown bits by enumeration.
4531 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
4532 enumeration under the same conditions. Handle remaining
4533 shifts as a mask based upon the minimum possible shift value.
4535 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4536 Richard Biener <rguenther@suse.de>
4538 * match.pd (shift transformations): Remove a redundant
4539 !POINTER_TYPE_P check.
4541 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4544 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
4545 Set all_regs to true in the call to replace_rtx.
4547 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4549 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
4550 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
4551 hiting --param modref-max-refs limit.
4552 (modref_tree:insert_base): Be smarter when hitting
4553 --param modref-max-bases limit. Add new parameter REF.
4554 (modref_tree:insert): Update.
4555 (modref_tree:merge): Update.
4556 * ipa-modref.c (read_modref_records): Update.
4558 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4560 * params.opt: (modref-max-adjustments): Add full stop.
4562 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4564 * ipa-modref-tree.h (modref_ref_node::verify): New member
4566 (modref_ref_node::insert): Use it.
4567 (modref_ref_node::try_mere_with): Fix off by one error.
4569 2021-08-26 Martin Liska <mliska@suse.cz>
4570 Stefan Kneifel <stefan.kneifel@bluewin.ch>
4572 * cgraph.h (create_version_clone_with_body): Add new parameter.
4573 * cgraphclones.c: Likewise.
4574 * multiple_target.c (create_dispatcher_calls): Do not use
4576 (create_target_clone): Likewise here.
4578 2021-08-26 Jonathan Yong <10walls@gmail.com>
4580 * doc/extend.texi: Add note about reserved priorities
4581 to the constructor attribute.
4583 2021-08-25 Martin Sebor <msebor@redhat.com>
4585 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
4586 range table header alone.
4587 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
4589 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
4591 * doc/invoke.texi: Document --param modref-max-adjustments.
4592 * ipa-modref-tree.c (test_insert_search_collapse): Update.
4593 (test_merge): Update.
4594 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
4595 (modref_access_node::operator==): Fix handling of access ranges.
4596 (modref_access_node::contains): Constify parameter; handle also
4597 mismatched parm offsets.
4598 (modref_access_node::update): New function.
4599 (modref_access_node::merge): New function.
4600 (unspecified_modref_access_node): Update constructor.
4601 (modref_ref_node::insert_access): Add record_adjustments parameter;
4603 (modref_ref_node::try_merge_with): New private function.
4604 (modref_tree::insert): New record_adjustments parameter.
4605 (modref_tree::merge): New record_adjustments parameter.
4606 (modref_tree::copy_from): Update.
4607 * ipa-modref.c (dump_access): Dump adjustments field.
4608 (get_access): Update constructor.
4609 (record_access): Update call of insert.
4610 (record_access_lto): Update call of insert.
4611 (merge_call_side_effects): Add record_adjustments parameter.
4612 (get_access_for_fnspec): Update.
4613 (process_fnspec): Update.
4614 (analyze_call): Update.
4615 (analyze_function): Update.
4616 (read_modref_records): Update.
4617 (ipa_merge_modref_summary_after_inlining): Update.
4618 (propagate_unknown_call): Update.
4619 (modref_propagate_in_scc): Update.
4620 * params.opt (param-max-modref-adjustments=): New.
4622 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
4624 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
4626 (xxspltiw_v4si): Use vecperm type attribute.
4627 (xxspltiw_v4si_inst): Use vecperm type attribute.
4628 (xxspltiw_v4sf_inst): Likewise.
4629 (xxspltidp_v2df): Use vecperm type attribute. Use
4630 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
4631 (xxspltidp_v2df_inst): Likewise.
4632 (xxsplti32dx_v4si): Use vecperm type attribute.
4633 (xxsplti32dx_v4si_inst): Likewise.
4634 (xxsplti32dx_v4sf_inst): Likewise.
4635 (xxblend_<mode>): Likewise.
4636 (xxpermx): Likewise.
4637 (xxpermx_inst): Likewise.
4640 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
4643 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
4644 * diagnostic.c (diagnostic_initialize_input_context): New function.
4645 * diagnostic.h (diagnostic_initialize_input_context): Declare.
4646 * input.c (default_charset_callback): New function.
4647 (file_cache::initialize_input_context): New function.
4648 (file_cache_slot::create): Added ability to convert the input
4649 according to the input context.
4650 (file_cache::file_cache): Initialize the new input context.
4651 (class file_cache_slot): Added new m_alloc_offset member.
4652 (file_cache_slot::file_cache_slot): Initialize the new member.
4653 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
4654 (file_cache_slot::maybe_grow): Likewise.
4655 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
4656 (file_cache_slot::get_next_line): Likewise.
4657 * input.h (class file_cache): Added input context member.
4659 2021-08-25 Richard Biener <rguenther@suse.de>
4661 PR tree-optimization/102046
4662 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
4663 update ->any_pattern when swapping operands.
4665 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
4668 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
4669 (ix86_decompose_address): Remove retval check for ASHIFT,
4670 allow non-canonical zero extend if AND mask covers ASHIFT
4672 (ix86_legitimate_address_p): Adjust condition for decompose.
4673 (ix86_rtx_costs): Adjust cost for lea with non-canonical
4675 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
4677 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
4679 PR tree-optimization/101145
4680 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
4682 (number_of_iterations_lt): Invoke above function.
4683 (adjust_cond_for_loop_until_wrap):
4684 Merge to number_of_iterations_until_wrap.
4685 (number_of_iterations_cond): Update invokes for
4686 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
4688 2021-08-25 konglin1 <lingling.kong@intel.com>
4691 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
4693 (_mm512_mask_fpclass_ps_mask): Ditto.
4695 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
4697 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
4698 (vec_unpacku_hi_v8hi): Likewise.
4699 (vec_unpacku_lo_v16qi): Likewise.
4700 (vec_unpacku_lo_v8hi): Likewise.
4701 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
4702 (vec_unpacku_lo_<VP_small_lc>): Likewise.
4704 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
4706 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
4707 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
4708 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
4709 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
4711 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
4713 PR middle-end/102031
4714 * simplify-rtx.c (simplify_truncation): When comparing precisions
4715 use "subreg_prec" variable, not "subreg_mode".
4717 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4719 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
4722 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4724 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
4725 various pointer type nodes.
4726 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
4727 values for various pointer types.
4728 (ptr_V16QI_type_node): New macro.
4729 (ptr_V1TI_type_node): New macro.
4730 (ptr_V2DI_type_node): New macro.
4731 (ptr_V2DF_type_node): New macro.
4732 (ptr_V4SI_type_node): New macro.
4733 (ptr_V4SF_type_node): New macro.
4734 (ptr_V8HI_type_node): New macro.
4735 (ptr_unsigned_V16QI_type_node): New macro.
4736 (ptr_unsigned_V1TI_type_node): New macro.
4737 (ptr_unsigned_V8HI_type_node): New macro.
4738 (ptr_unsigned_V4SI_type_node): New macro.
4739 (ptr_unsigned_V2DI_type_node): New macro.
4740 (ptr_bool_V16QI_type_node): New macro.
4741 (ptr_bool_V8HI_type_node): New macro.
4742 (ptr_bool_V4SI_type_node): New macro.
4743 (ptr_bool_V2DI_type_node): New macro.
4744 (ptr_bool_V1TI_type_node): New macro.
4745 (ptr_pixel_type_node): New macro.
4746 (ptr_intQI_type_node): New macro.
4747 (ptr_uintQI_type_node): New macro.
4748 (ptr_intHI_type_node): New macro.
4749 (ptr_uintHI_type_node): New macro.
4750 (ptr_intSI_type_node): New macro.
4751 (ptr_uintSI_type_node): New macro.
4752 (ptr_intDI_type_node): New macro.
4753 (ptr_uintDI_type_node): New macro.
4754 (ptr_intTI_type_node): New macro.
4755 (ptr_uintTI_type_node): New macro.
4756 (ptr_long_integer_type_node): New macro.
4757 (ptr_long_unsigned_type_node): New macro.
4758 (ptr_float_type_node): New macro.
4759 (ptr_double_type_node): New macro.
4760 (ptr_long_double_type_node): New macro.
4761 (ptr_dfloat64_type_node): New macro.
4762 (ptr_dfloat128_type_node): New macro.
4763 (ptr_ieee128_type_node): New macro.
4764 (ptr_ibm128_type_node): New macro.
4765 (ptr_vector_pair_type_node): New macro.
4766 (ptr_vector_quad_type_node): New macro.
4767 (ptr_long_long_integer_type_node): New macro.
4768 (ptr_long_long_unsigned_type_node): New macro.
4770 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4772 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
4773 and power9-64 stanzas.
4775 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
4776 Tom de Vries <tdevries@suse.de>
4778 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
4779 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
4780 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
4781 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
4782 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
4783 * config/nvptx/nvptx-c.c: New source file.
4784 (nvptx_cpu_cpp_builtins): Move implementation here.
4786 2021-08-24 Martin Sebor <msebor@redhat.com>
4788 PR middle-end/101600
4789 PR middle-end/101977
4790 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
4791 the phrasing of a warning.
4792 (check_access): Use the remaining size after subtracting any offset
4793 rather than the whole object size.
4794 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
4795 clear for any nonnull PHI argument.
4796 (compute_objsize): Clear argument.
4798 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4800 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
4802 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4804 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
4807 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
4809 * value-relation.cc (rr_transitive_table): New.
4810 (relation_transitive): New.
4811 (value_relation::swap): Remove.
4812 (value_relation::apply_transitive): New.
4813 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
4814 (relation_oracle::register_relation): Call register_transitives.
4815 (relation_oracle::register_transitives): New.
4816 * value-relation.h (relation_oracle): Add new temporary bitmap and
4819 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
4822 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
4823 from integer to a pseudo vector register.
4825 2021-08-24 Richard Biener <rguenther@suse.de>
4827 PR tree-optimization/100089
4828 * tree-vectorizer.h (vect_slp_bb): Rename to ...
4829 (vect_slp_if_converted_bb): ... this and get the original
4830 loop as new argument.
4831 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
4832 pass original loop to vect_slp_if_converted_bb.
4833 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4834 If orig_loop was passed scan the not vectorized stmts
4835 for COND_EXPRs and force not profitable if found.
4836 (vect_slp_region): Pass down all SLP instances to costing
4837 if orig_loop was specified.
4838 (vect_slp_bbs): Pass through orig_loop.
4839 (vect_slp_bb): Rename to ...
4840 (vect_slp_if_converted_bb): ... this and get the original
4841 loop as new argument.
4842 (vect_slp_function): Adjust.
4844 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
4847 * config/arm/arm.md (attribute arch): Add fix_vlldm.
4848 (arch_enabled): Use it.
4849 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
4850 use when erratum mitigation is needed.
4852 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
4855 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
4856 * doc/invoke.texi (Arm Options): Document it.
4857 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
4858 (ALL_QUIRKS): Add quirk_vlldm.
4859 (cortex-m33): Add quirk_vlldm.
4860 (cortex-m35p, cortex-m55): Likewise.
4861 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
4862 targetting an affected CPU and not explicitly controlled on
4865 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
4867 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
4868 (lazy_load_multiple_insn): Likewise.
4870 2021-08-24 liuhongt <hongtao.liu@intel.com>
4873 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
4874 Enable avx512 embedded broadcast.
4875 (*<avx512>_vternlog<mode>_all): Ditto.
4876 (<avx512>_vternlog<mode>_mask): Ditto.
4878 2021-08-24 liuhongt <hongtao.liu@intel.com>
4881 * config/i386/i386.c (ix86_rtx_costs): Define cost for
4883 * config/i386/i386.h (STRIP_UNARY): New macro.
4884 * config/i386/predicates.md (reg_or_notreg_operand): New
4886 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
4887 (*<avx512>_vternlog<mode>_1): New pre_reload
4888 define_insn_and_split.
4889 (*<avx512>_vternlog<mode>_2): Ditto.
4890 (*<avx512>_vternlog<mode>_3): Ditto.
4891 (any_logic1,any_logic2): New code iterator.
4892 (logic_op): New code attribute.
4893 (ternlogsuffix): Extend to VNxDF and VNxSF.
4895 2021-08-24 Richard Biener <rguenther@suse.de>
4897 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
4898 * params.opt (--param vect-inner-loop-cost-factor): Adjust
4900 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
4901 inner_loop_cost_factor to the minimum of the estimated number
4902 of iterations of the inner loop and vect-inner-loop-cost-factor.
4904 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
4905 Richard Biener <rguenther@suse.de>
4907 * config/i386/i386-features.c (compute_convert_gain): Provide
4908 more accurate values for CONST_INT, when optimizing for size.
4909 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
4910 * config/i386/i386.h (COSTS_N_BYTES): to here.
4912 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
4913 Jakub Jelinek <jakub@redhat.com>
4915 PR middle-end/102029
4916 * match.pd (shift transformations): Add an additional check for
4917 !POINTER_TYPE_P in the recently added left shift transformation.
4919 2021-08-24 liuhongt <hongtao.liu@intel.com>
4921 PR tree-optimization/100089
4922 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
4923 loop vectorizer when cost model is very-cheap.
4925 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
4927 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
4928 asprintf, which is not available on AIX.
4930 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
4932 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
4933 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
4935 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
4937 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
4938 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
4940 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
4941 (rbtree.o): Likewise.
4942 (rs6000-gen-builtins): Likewise.
4943 (rs6000-builtins.c): Likewise.
4944 (rs6000-builtins.h): Likewise.
4945 (rs6000.o): Add dependency.
4946 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
4947 (rs6000-vecdefines.h): New target.
4948 (rs6000-builtins.o): Likewise.
4949 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
4950 (rs6000-c.o): Likewise.
4952 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
4955 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
4956 Diagnose buffer overrun.
4957 (safe_inc_pos): Fix overrun detection.
4958 (match_identifier): Diagnose buffer overrun.
4959 (match_integer): Likewise.
4960 (match_to_right_bracket): Likewise.
4962 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
4964 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
4965 Improve range compare.
4966 (modref_access_node::contains): New member function.
4967 (modref_access_node::search): Remove.
4968 (modref_access_node::insert): Be smarter about subaccesses.
4970 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
4972 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
4973 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
4974 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
4977 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
4979 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
4981 * config/h8300/jumpcall.md (call, call_value): Restrict to
4982 !SIBLING_CALL_P cases.
4983 (subcall, sibcall_value): New patterns & expanders.
4984 * config/h8300/proepi.md (epilogue): Pass new argument to
4985 h8300_expand_epilogue.
4986 (sibcall_epilogue): New expander.
4987 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
4989 (h8300_ok_for_sibcall_p): New function.
4990 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
4992 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
4994 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
4995 Handle case where the operand is already the desired mode.
4997 2021-08-23 Richard Biener <rguenther@suse.de>
5000 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5001 in addition to has_gimple_body.
5003 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5005 PR middle-end/101949
5006 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5009 2021-08-23 Martin Liska <mliska@suse.cz>
5011 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5014 2021-08-23 Richard Biener <rguenther@suse.de>
5016 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5017 Properly scale the inner loop cost only once.
5019 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5021 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5022 Provide bounds for unsigned (and signed with non-negative operands)
5023 division and modulus.
5025 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5027 * simplify-rtx.c (simplify_truncation): Generalize simplification
5028 of (truncate:A (subreg:B X)).
5029 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5030 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5031 already has the desired machine mode.
5032 (test_scalar_int_ops): Add tests that useless extensions and
5033 truncations are optimized away.
5034 (test_scalar_int_ext_ops): New self-test function to confirm
5035 that truncations of extensions are correctly simplified.
5036 (test_scalar_int_ext_ops2): New self-test function to check
5037 truncations of truncations, extensions of extensions, and
5038 truncations of extensions.
5039 (test_scalar_ops): Call the above two functions with a
5040 representative sampling of integer machine modes.
5042 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5044 * match.pd (shift transformations): Change the sign of an
5045 LSHIFT_EXPR if it reduces the number of explicit conversions.
5047 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5049 PR tree-optimization/86723
5050 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5051 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5052 non-memory permutation of 64-bit src which only has bytes of
5053 0 or [5..8] and n->range is 4.
5054 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5055 find_bswap_or_nop_finalize caller, support bswap with some bytes
5056 zeroed, as long as at least two bytes are not zeroed.
5057 (bswap_replace): Add mask argument and handle masking of bswap
5059 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5060 caller, punt if cast64_to_32 or mask is not all ones.
5061 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5062 caller, for now punt if cast64_to_32.
5064 2021-08-23 Richard Biener <rguenther@suse.de>
5066 PR tree-optimization/79334
5067 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5068 a type also for COMPONENT_REFs.
5069 (vn_reference_may_trap): Check ARRAY_REF with constant index
5070 against the array domain.
5072 2021-08-23 liuhongt <hongtao.liu@intel.com>
5075 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5076 TARGET_AVX512BW to condition.
5078 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5081 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5082 register variables already during early_dwarf if possible.
5084 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5086 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5088 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5091 * config/arm/arm.opt: Fix typo.
5092 * config/arm/t-rmprofile: Fix typo.
5094 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5096 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5097 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5098 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5099 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5100 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5101 if either grainsize or num_tasks clause has the strict modifier.
5103 2021-08-23 Martin Liska <mliska@suse.cz>
5105 * dbgcnt.def (DEBUG_COUNTER): New counter.
5106 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5108 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5110 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5112 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5115 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5117 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5119 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5120 vlshr<mode>3): Call it.
5122 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5123 Martin Liska <mliska@suse.cz>
5125 PR middle-end/101949
5126 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5129 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5131 * config/mips/mips.c (mips_function_rodata_section,
5132 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5134 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5136 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5138 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5140 * configure.ac (thread-local storage support): Remove tls_first_major
5141 and tls_first_minor. Use "$conftest_s" to check support.
5142 * configure: Regenerate.
5144 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5146 * configure.ac: Fixup formatting.
5148 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5150 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5151 * configure.ac: ... update all callers.
5153 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5156 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5157 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5158 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5159 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5160 Remove remaining checks for in-tree assembler.
5161 * configure: Regenerate.
5163 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5165 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5166 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5168 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5169 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5170 Improve logical shifts by 27 bits for H8/S.
5171 (get_shift_alg): Corresponding changes.
5172 (h8300_option_override): Revert to loops for -Os when profitable.
5174 2021-08-20 Richard Biener <rguenther@suse.de>
5176 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5178 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5179 index as group_id when dataref_groups were not computed.
5180 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5181 we advace to the next BB.
5183 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5185 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5188 2021-08-20 Martin Liska <mliska@suse.cz>
5190 PR gcov-profile/89961
5191 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5192 (mangle_name): Simplify, do not used the second argument.
5193 (strip_extention): New function.
5194 (get_md5sum): Likewise.
5195 (get_gcov_intermediate_filename): Handle properly -p and -x
5197 (output_gcov_file): Use string type.
5198 (generate_results): Likewise.
5199 (md5sum_to_hex): Remove.
5201 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5203 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5204 (UNSPEC_XXSPLTIW): Move to vsx.md.
5205 (UNSPEC_XXSPLTID): Move to vsx.md.
5206 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5207 (UNSPEC_XXBLEND): Move to vsx.md.
5208 (UNSPEC_XXPERMX): Move to vsx.md.
5209 (VM3): Move to vsx.md.
5210 (VM3_char): Move to vsx.md.
5211 (xxspltiw_v4si): Move to vsx.md.
5212 (xxspltiw_v4sf): Move to vsx.md.
5213 (xxspltiw_v4sf_inst): Move to vsx.md.
5214 (xxspltidp_v2df): Move to vsx.md.
5215 (xxspltidp_v2df_inst): Move to vsx.md.
5216 (xxsplti32dx_v4si_inst): Move to vsx.md.
5217 (xxsplti32dx_v4sf): Move to vsx.md.
5218 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5219 (xxblend_<mode>): Move to vsx.md.
5220 (xxpermx): Move to vsx.md.
5221 (xxpermx_inst): Move to vsx.md.
5222 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5223 (UNSPEC_XXSPLTIW): Move from altivec.md.
5224 (UNSPEC_XXSPLTID): Move from altivec.md.
5225 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5226 (UNSPEC_XXBLEND): Move from altivec.md.
5227 (UNSPEC_XXPERMX): Move from altivec.md.
5228 (VM3): Move from altivec.md.
5229 (VM3_char): Move from altivec.md.
5230 (xxspltiw_v4si): Move from altivec.md.
5231 (xxspltiw_v4sf): Move from altivec.md.
5232 (xxspltiw_v4sf_inst): Move from altivec.md.
5233 (xxspltidp_v2df): Move from altivec.md.
5234 (xxspltidp_v2df_inst): Move from altivec.md.
5235 (xxsplti32dx_v4si_inst): Move from altivec.md.
5236 (xxsplti32dx_v4sf): Move from altivec.md.
5237 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5238 (xxblend_<mode>): Move from altivec.md.
5239 (xxpermx): Move from altivec.md.
5240 (xxpermx_inst): Move from altivec.md.
5242 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5244 * tree-vect-generic.c (expand_vector_operations_1): Use either
5245 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5246 when constructing scalar splat expressions.
5248 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5251 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5252 pointer to __vector_pair *.
5254 2021-08-19 Martin Sebor <msebor@redhat.com>
5256 * gimple-range.cc: Add comments.
5257 * gimple-range.h: Same.
5259 2021-08-19 Martin Sebor <msebor@redhat.com>
5261 PR middle-end/101984
5262 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5265 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5267 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5268 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5269 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5271 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5273 PR middle-end/101950
5274 * optabs.c (expand_clrsb_using_clz): New function.
5275 (expand_unop): Use it as another clrsb expansion fallback.
5277 2021-08-19 liuhongt <hongtao.liu@intel.com>
5280 2021-07-28 liuhongt <hongtao.liu@intel.com>
5283 * config/i386/i386.h (processor_costs): Add new member
5285 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5286 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5287 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5288 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5289 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5290 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5291 generic_cost, core_cost): Initialize integer_to_sse same value
5293 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5294 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5295 Use integer_to_sse instead of sse_op to calculate the cost of
5298 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5300 * config.gcc: Include rpath.opt for Darwin.
5301 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5303 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5306 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5309 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5311 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5312 pattern and type cast.
5313 (vld4_lane_f32): Use float RTL pattern.
5314 (vld4q_lane_f64): Use float type cast.
5316 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5318 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5321 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5323 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5324 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5325 (hash_map_tests_c_tests): Call it.
5327 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5329 * ggc.h (enum ggc_collect): New.
5330 (ggc_collect): Use it.
5331 * ggc-page.c: Adjust.
5332 * ggc-common.c: Likewise.
5333 * ggc-tests.c: Likewise.
5334 * read-rtl-function.c: Likewise.
5335 * selftest-run-tests.c: Likewise.
5336 * doc/gty.texi (Invoking the garbage collector): Likewise.
5338 2021-08-18 liuhongt <hongtao.liu@intel.com>
5341 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
5343 * config/i386/sse.md (*sse3_haddv2df3_low): Add
5344 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
5345 (*sse3_hsubv2df3_low): Ditto.
5346 * config/i386/x86-tune.def
5347 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
5349 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5351 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
5352 (gori_compute::compute_operand_range): Add tracing.
5353 (gori_compute::logical_combine): Ditto.
5354 (gori_compute::compute_logical_operands): Ditto.
5355 (gori_compute::compute_operand1_range): Ditto.
5356 (gori_compute::compute_operand2_range): Ditto.
5357 (gori_compute::outgoing_edge_range_p): Ditto.
5358 * gimple-range-gori.h (class gori_compute): Add range_tracer.
5360 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5362 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
5363 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
5364 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
5365 * params.opt (--param=evrp-mode): Adjust options.
5367 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5369 * Makefile.in (OBJS): Add gimple-range-trace.o.
5370 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
5371 * gimple-range-fold.cc: Adjust headers.
5372 * gimple-range-trace.cc: New.
5373 * gimple-range-trace.h: New.
5374 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
5375 (gimple_ranger::range_of_expr): Add tracing.
5376 (gimple_ranger::range_on_entry): Ditto.
5377 (gimple_ranger::range_on_exit): Ditto.
5378 (gimple_ranger::range_on_edge): Ditto.
5379 (gimple_ranger::fold_range_internal): Ditto.
5380 (gimple_ranger::dump_bb): Do not calculate edge range twice.
5381 (trace_ranger::*): Remove.
5382 (enable_ranger): Never create a trace_ranger.
5383 (debug_seed_ranger): Move to gimple-range-trace.cc.
5384 (dump_ranger): Ditto.
5385 (debug_ranger): Ditto.
5386 * gimple-range.h: Include gimple-range-trace.h.
5387 (range_on_entry, range_on_exit): No longer virtual.
5388 (class trace_ranger): Remove.
5389 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
5391 2021-08-17 Martin Sebor <msebor@redhat.com>
5393 PR middle-end/101854
5394 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
5395 in gimple-ssa-warn-access.cc.
5396 * calls.c (alloc_max_size): Move code to check_alloca.
5397 (get_size_range): Move to pointer-query.cc.
5398 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
5399 (get_attr_nonstring_decl): Move to tree.c.
5400 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
5401 (append_attrname): Same.
5402 (maybe_warn_rdwr_sizes): Same.
5403 (initialize_argument_information): Move code to
5404 gimple-ssa-warn-access.cc.
5405 * calls.h (maybe_warn_alloc_args_overflow): Move to
5406 gimple-ssa-warn-access.h.
5407 (get_attr_nonstring_decl): Move to tree.h.
5408 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
5409 (enum size_range_flags): Move to pointer-query.h.
5410 (get_size_range): Same.
5411 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
5412 to avoid Clang -Wunused-function.
5413 (get_size_range): Declare static.
5414 (maybe_emit_free_warning): Rename...
5415 (maybe_check_dealloc_call): ...to this for consistency.
5416 (class pass_waccess): Add members.
5417 (pass_waccess::~pass_waccess): Defined.
5418 (alloc_max_size): Move here from calls.c.
5419 (maybe_warn_alloc_args_overflow): Same.
5420 (check_alloca): New function.
5421 (check_alloc_size_call): New function.
5422 (check_strncat): Handle another warning flag.
5423 (pass_waccess::check_builtin): Handle alloca.
5424 (fntype_argno_type): Move here from calls.c.
5425 (append_attrname): Same.
5426 (maybe_warn_rdwr_sizes): Same.
5427 (pass_waccess::check_call): Define.
5428 (check_nonstring_args): New function.
5429 (pass_waccess::check): Call new member functions.
5430 (pass_waccess::execute): Enable ranger.
5431 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
5432 (maybe_warn_nonstring_arg): Same.
5433 * gimple-ssa-warn-restrict.c: Remove #include.
5434 * pointer-query.cc (get_size_range): Move here from calls.c.
5435 * pointer-query.h (enum size_range_flags): Same.
5436 (get_size_range): Same.
5437 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
5438 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
5440 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
5442 * ggc.h (ggc_collect): Add 'force_collect' parameter.
5443 * ggc-page.c (ggc_collect): Use that one instead of global
5444 'ggc_force_collect'. Adjust all users.
5445 * doc/gty.texi (Invoking the garbage collector): Update.
5446 * ggc-internal.h (ggc_force_collect): Remove.
5447 * ggc-common.c (ggc_force_collect): Likewise.
5448 * selftest.h (forcibly_ggc_collect): Remove.
5449 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
5450 * read-rtl-function.c (test_loading_labels): Adjust.
5451 * selftest-run-tests.c (run_tests): Likewise.
5453 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5455 * config/darwin.c (darwin_file_end): Reset and reclaim the
5456 section names table at the end of compile.
5458 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5461 * config.in: Regenerate.
5462 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
5463 (ASM_SPEC): Pass options to disable branch shortening where
5465 * configure: Regenerate.
5466 * configure.ac: Detect versions of 'as' that support the
5467 optimisation which has the bug.
5469 2021-08-17 Richard Biener <rguenther@suse.de>
5471 * optabs-query.c (supports_vec_gather_load_p): Also check
5473 (supports_vec_scatter_store_p): Likewise.
5474 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
5475 back to masked variants if non-masked are not supported.
5476 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
5477 When we need to use masked gather/scatter but do not have
5478 a mask set up a constant true one.
5479 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
5482 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5484 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
5485 algorithm as PLUS_EXPR to improve subtraction bit bounds.
5486 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
5488 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5490 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
5491 calculate the mask-value pair result of a multiplication by an
5493 (bit_value_binop) [MULT_EXPR]: Call it from here for
5494 multiplications by (sparse) non-negative constants.
5496 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
5499 * config.gcc (gcc_cv_initfini_array): Leave undefined for
5500 uclinuxfdpiceabi targets.
5502 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5504 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
5505 reverse debug stmts.
5507 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5509 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
5511 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5513 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
5514 (__LD4Q_LANE_FUNC): Likewise.
5515 (vld4_lane_u8): Define without macro.
5516 (vld4_lane_u16): Likewise.
5517 (vld4_lane_u32): Likewise.
5518 (vld4_lane_u64): Likewise.
5519 (vld4_lane_s8): Likewise.
5520 (vld4_lane_s16): Likewise.
5521 (vld4_lane_s32): Likewise.
5522 (vld4_lane_s64): Likewise.
5523 (vld4_lane_f16): Likewise.
5524 (vld4_lane_f32): Likewise.
5525 (vld4_lane_f64): Likewise.
5526 (vld4_lane_p8): Likewise.
5527 (vld4_lane_p16): Likewise.
5528 (vld4_lane_p64): Likewise.
5529 (vld4q_lane_u8): Likewise.
5530 (vld4q_lane_u16): Likewise.
5531 (vld4q_lane_u32): Likewise.
5532 (vld4q_lane_u64): Likewise.
5533 (vld4q_lane_s8): Likewise.
5534 (vld4q_lane_s16): Likewise.
5535 (vld4q_lane_s32): Likewise.
5536 (vld4q_lane_s64): Likewise.
5537 (vld4q_lane_f16): Likewise.
5538 (vld4q_lane_f32): Likewise.
5539 (vld4q_lane_f64): Likewise.
5540 (vld4q_lane_p8): Likewise.
5541 (vld4q_lane_p16): Likewise.
5542 (vld4q_lane_p64): Likewise.
5543 (vld4_lane_bf16): Likewise.
5544 (vld4q_lane_bf16): Likewise.
5546 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5548 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
5549 (__LD3Q_LANE_FUNC): Delete.
5550 (vld3_lane_u8): Define without macro.
5551 (vld3_lane_u16): Likewise.
5552 (vld3_lane_u32): Likewise.
5553 (vld3_lane_u64): Likewise.
5554 (vld3_lane_s8): Likewise.
5555 (vld3_lane_s16): Likewise.
5556 (vld3_lane_s32): Likewise.
5557 (vld3_lane_s64): Likewise.
5558 (vld3_lane_f16): Likewise.
5559 (vld3_lane_f32): Likewise.
5560 (vld3_lane_f64): Likewise.
5561 (vld3_lane_p8): Likewise.
5562 (vld3_lane_p16): Likewise.
5563 (vld3_lane_p64): Likewise.
5564 (vld3q_lane_u8): Likewise.
5565 (vld3q_lane_u16): Likewise.
5566 (vld3q_lane_u32): Likewise.
5567 (vld3q_lane_u64): Likewise.
5568 (vld3q_lane_s8): Likewise.
5569 (vld3q_lane_s16): Likewise.
5570 (vld3q_lane_s32): Likewise.
5571 (vld3q_lane_s64): Likewise.
5572 (vld3q_lane_f16): Likewise.
5573 (vld3q_lane_f32): Likewise.
5574 (vld3q_lane_f64): Likewise.
5575 (vld3q_lane_p8): Likewise.
5576 (vld3q_lane_p16): Likewise.
5577 (vld3q_lane_p64): Likewise.
5578 (vld3_lane_bf16): Likewise.
5579 (vld3q_lane_bf16): Likewise.
5581 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5583 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
5584 (__LD2Q_LANE_FUNC): Likewise.
5585 (vld2_lane_u8): Define without macro.
5586 (vld2_lane_u16): Likewise.
5587 (vld2_lane_u32): Likewise.
5588 (vld2_lane_u64): Likewise.
5589 (vld2_lane_s8): Likewise.
5590 (vld2_lane_s16): Likewise.
5591 (vld2_lane_s32): Likewise.
5592 (vld2_lane_s64): Likewise.
5593 (vld2_lane_f16): Likewise.
5594 (vld2_lane_f32): Likewise.
5595 (vld2_lane_f64): Likewise.
5596 (vld2_lane_p8): Likewise.
5597 (vld2_lane_p16): Likewise.
5598 (vld2_lane_p64): Likewise.
5599 (vld2q_lane_u8): Likewise.
5600 (vld2q_lane_u16): Likewise.
5601 (vld2q_lane_u32): Likewise.
5602 (vld2q_lane_u64): Likewise.
5603 (vld2q_lane_s8): Likewise.
5604 (vld2q_lane_s16): Likewise.
5605 (vld2q_lane_s32): Likewise.
5606 (vld2q_lane_s64): Likewise.
5607 (vld2q_lane_f16): Likewise.
5608 (vld2q_lane_f32): Likewise.
5609 (vld2q_lane_f64): Likewise.
5610 (vld2q_lane_p8): Likewise.
5611 (vld2q_lane_p16): Likewise.
5612 (vld2q_lane_p64): Likewise.
5613 (vld2_lane_bf16): Likewise.
5614 (vld2q_lane_bf16): Likewise.
5616 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5618 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
5621 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5623 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
5624 (rank_for_schedule): Use it.
5626 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5628 PR rtl-optimization/91598
5629 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
5630 insns after memory reads and before memory writes.
5632 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
5634 * rtl.h (CONST_VECTOR_P): New macro.
5635 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
5636 code testing macros.
5637 (aarch64_ptrue_all_mode): Likewise.
5638 (aarch64_expand_mov_immediate): Likewise.
5639 (aarch64_const_vec_all_in_range_p): Likewise.
5640 (aarch64_rtx_costs): Likewise.
5641 (aarch64_legitimate_constant_p): Likewise.
5642 (aarch64_simd_valid_immediate): Likewise.
5643 (aarch64_simd_make_constant): Likewise.
5644 (aarch64_convert_mult_to_shift): Likewise.
5645 (aarch64_expand_sve_vec_perm): Likewise.
5646 (aarch64_vec_fpconst_pow_of_2): Likewise.
5648 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5650 PR tree-optimization/101938
5651 * range-op.cc (operator_abs::op1_range): Special case
5652 -TYPE_MIN_VALUE for flag_wrapv.
5654 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
5656 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
5659 2021-08-17 Jakub Jelinek <jakub@redhat.com>
5661 * tree.def (OMP_SCOPE): New tree code.
5662 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
5663 * tree-nested.c (convert_nonlocal_reference_stmt,
5664 convert_local_reference_stmt, convert_gimple_call): Handle
5666 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
5667 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
5668 * gimple.c (gimple_build_omp_scope): New function.
5669 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
5670 * gimple.h (gimple_build_omp_scope): Declare.
5671 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
5672 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
5673 gimple_omp_scope_set_clauses): New inline functions.
5674 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
5675 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
5676 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
5677 * gimple-walk.c (walk_gimple_stmt): Likewise.
5678 * gimple-low.c (lower_stmt): Likewise.
5679 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
5680 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
5681 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
5682 allowing task reductions. Reject inscan reductions on scope.
5683 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
5684 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
5685 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
5686 (estimate_num_insns): Likewise.
5687 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
5688 contexts if var isn't privatized there.
5689 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
5690 (scan_omp_1_stmt): Likewise.
5691 (maybe_add_implicit_barrier_cancel): Look through outer
5693 (lower_omp_scope): New function.
5694 (lower_omp_task_reductions): Handle OMP_SCOPE.
5695 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
5696 (diagnose_sb_1, diagnose_sb_2): Likewise.
5697 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
5698 (expand_omp): Handle GIMPLE_OMP_SCOPE.
5699 (omp_make_gimple_edges): Likewise.
5700 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
5702 2021-08-17 Richard Biener <rguenther@suse.de>
5704 PR tree-optimization/101925
5705 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
5706 reverse on COMPONENT_REF and ARRAY_REF according to
5707 what reverse_storage_order_for_component_p does.
5708 (vn_reference_eq): Compare reversed on reference ops.
5709 (reverse_storage_order_for_component_p): New overload.
5710 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
5711 on the reference looked up.
5713 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
5715 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
5717 (h8300_option_override): Use loops on H8/S more often when optimizing
5719 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
5720 accordingly. Handle various arithmetic right shifts with special
5721 sequences that we couldn't handle before.
5723 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
5725 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
5727 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
5729 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
5730 * config/sparc/sparc.c (sparc_gcov_type_size): New.
5731 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
5732 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
5733 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
5734 * doc/tm.texi.in: Regenerate.
5735 * target.def (gcov_type_size): New target hook.
5736 * targhooks.c (default_gcov_type_size): New.
5737 * targhooks.h (default_gcov_type_size): Declare.
5738 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
5740 (gimple_gen_time_profiler): Likewise.
5742 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
5744 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
5746 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
5748 PR middle-end/101931
5749 * omp-low.c (omp_runtime_api_call): Update for routines
5750 added in the meanwhile.
5752 2021-08-16 Martin Liska <mliska@suse.cz>
5754 PR tree-optimization/100393
5755 * tree-switch-conversion.c (group_cluster::dump): Use
5756 get_comparison_count.
5757 (jump_table_cluster::find_jump_tables): Pre-compute number of
5758 comparisons and then decrement it. Cache also max_ratio.
5759 (jump_table_cluster::can_be_handled): Change signature.
5760 * tree-switch-conversion.h (get_comparison_count): New.
5762 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
5764 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
5765 encodings only when -fgnat-encodings=all is specified.
5766 (add_bound_info): Likewise.
5767 (add_byte_size_attribute): Likewise.
5768 (gen_member_die): Likewise.
5770 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
5772 * omp-oacc-neuter-broadcast.cc
5773 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
5775 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
5777 * omp-oacc-neuter-broadcast.cc
5778 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
5781 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
5783 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
5784 (execute_omp_oacc_neuter_broadcast): ... here.
5785 (install_var_field, build_receiver_ref, build_sender_ref): Take
5786 'field_map_t *' parameter. Adjust all users.
5787 (worker_single_copy, neuter_worker_single): Take a
5788 'record_field_map_t *' parameter. Adjust all users.
5790 2021-08-16 liuhongt <hongtao.liu@intel.com>
5793 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
5796 2021-08-16 Martin Liska <mliska@suse.cz>
5799 * multiple_target.c (create_dispatcher_calls): Make default
5800 function local only if it is a definition.
5802 2021-08-16 Martin Liska <mliska@suse.cz>
5805 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
5806 consider equal SSA_NAMEs when one is a param.
5808 2021-08-16 liuhongt <hongtao.liu@intel.com>
5811 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
5812 Support vpermi2b for V32QI/V16QImode.
5813 (ix86_extract_perm_from_pool_constant): New function.
5814 (ix86_expand_vec_one_operand_perm_avx512): Support
5815 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
5816 (expand_vec_perm_1): Adjust comments for upper.
5817 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
5819 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
5820 (pshufb_truncv4siv4hi_operand): Ditto.
5821 (pshufb_truncv8hiv8qi_operand): Ditto.
5822 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
5823 New pre_reload define_insn_and_split.
5824 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
5825 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
5826 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
5827 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
5828 (*avx512f_permvar_truncv4div4si_1): Ditto.
5829 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
5830 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
5831 (*avx512f_pshufd_truncv2div2si_1): Ditto.
5833 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
5835 * config/riscv/multilib-generator: Support code model option for
5837 * doc/install.texi: Add document of new option for
5838 --with-multilib-generator.
5840 2021-08-15 Clément Chigot <clement.chigot@atos.net>
5842 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
5843 (rs6000_legitimize_tls_address_aix): Use it.
5844 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
5845 xcoff_tls_exec_model_detected is true.
5847 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
5849 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
5850 to allow a bit more code growth, saving many dozens of cycles.
5851 (h8300_option_override): Adjus shift_alg_si if optimizing for
5853 (get_shift_alg): Use special + inline shifts for residuals
5856 2021-08-14 Stafford Horne <shorne@gmail.com>
5859 * config/or1k/or1k-opts.h: New file.
5860 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
5861 Support generating gotha relocations if -mcmodel=large is
5863 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
5865 * config/or1k/or1k.opt (mcmodel=): New option.
5866 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
5868 2021-08-14 Martin Sebor <msebor@redhat.com>
5870 PR middle-end/101791
5871 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
5872 to valid_new_delete_pair_p.
5873 * tree.c (valid_new_delete_pair_p): Add argument.
5874 * tree.h (valid_new_delete_pair_p): Same.
5876 2021-08-14 Jakub Jelinek <jakub@redhat.com>
5879 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
5880 <case E_V64QImode>: For this mode assert
5881 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
5883 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
5886 * config/rs6000/altivec.md (xxeval): Use register_predicate
5887 instead of altivec_register_predicate.
5889 2021-08-13 Martin Sebor <msebor@redhat.com>
5891 PR middle-end/101734
5892 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
5893 (maybe_warn_operand): Call it.
5895 2021-08-13 Martin Liska <mliska@suse.cz>
5898 * attribs.c (decl_attributes): Make naked functions "noipa"
5901 2021-08-13 Martin Liska <mliska@suse.cz>
5904 * symtab.c (symtab_node::noninterposable_alias): Do not create
5905 local aliases for target_clone functions as the clonning pass
5908 2021-08-13 Martin Liska <mliska@suse.cz>
5910 * opts.c (LIVE_PATCHING_OPTION): Define.
5911 (control_options_for_live_patching): Use it in error messages.
5913 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
5915 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
5916 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
5917 ignore_stores_eaf_flags): New constants.
5918 (remove_useless_eaf_flags): New function.
5919 (eaf_flags_useful_p): Use it.
5920 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
5922 (modref_lattice::init): Add EAF_NOREAD.
5923 (modref_lattice::add_escape_point): Do not reacord escape point if
5925 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
5926 use remove_useless_eaf_flags.
5927 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
5928 (modref_lattice::merge_direct_load): Add EAF_NOREAD
5929 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
5930 (analyze_parms): Use remove_useless_eaf_flags.
5931 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
5932 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
5933 use remove_useless_eaf_flags.
5934 (modref_propagate_flags_in_scc): Update.
5935 * ipa-modref.h: Turn eaf_flags_t back to char.
5936 * tree-core.h (EAF_NOT_RETURNED): Fix.
5937 (EAF_NOREAD): New constant
5938 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
5940 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
5941 (handle_pure_call): Likewise.
5943 2021-08-12 Jakub Jelinek <jakub@redhat.com>
5945 * tree.def (OMP_MASKED): New tree code.
5946 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
5947 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
5948 OMP_CLAUSE_FILTER_EXPR): Define.
5949 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
5950 (omp_clause_code_name): Likewise.
5951 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
5952 * tree-nested.c (convert_nonlocal_omp_clauses,
5953 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
5954 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
5955 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
5956 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
5957 (dump_generic_node): Handle OMP_MASTER.
5958 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
5959 * gimple.c (gimple_build_omp_masked): New function.
5960 (gimple_copy): Handle GIMPLE_OMP_MASKED.
5961 * gimple.h (gimple_build_omp_masked): Declare.
5962 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
5963 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
5964 gimple_omp_masked_set_clauses): New inline functions.
5965 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
5966 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
5967 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
5968 * gimple-walk.c (walk_gimple_stmt): Likewise.
5969 * gimple-low.c (lower_stmt): Likewise.
5970 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
5971 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
5972 that take one expression rather than decl or constant, force
5973 gimplification of that into a SSA_NAME or temporary unless min
5975 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
5976 (gimplify_expr): Handle OMP_MASKED.
5977 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
5978 (estimate_num_insns): Likewise.
5979 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
5980 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
5981 diagnostics for existence of masked construct.
5982 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
5983 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
5984 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
5987 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
5990 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
5991 (ldexp<mode>3): Use avx512f_scalef<mode>2.
5992 (UNSPEC_SCALEF): Move from ...
5993 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
5995 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
5997 * ipa-split.c (consider_split): Fix condition testing void functions.
5999 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6001 * doc/invoke.texi: Remove docs for threader-mode param.
6002 * flag-types.h (enum threader_mode): Remove.
6003 * params.opt: Remove threader-mode param.
6004 * tree-ssa-threadbackward.c (class back_threader): Remove
6005 path_is_unreachable_p.
6006 Make find_paths private.
6007 Add maybe_thread and thread_through_all_blocks.
6008 Remove reference marker for m_registry.
6009 Remove reference marker for m_profit.
6010 (back_threader::back_threader): Adjust for registry and profit not
6012 (dump_path): Move down.
6014 (class thread_jumps): Remove.
6015 (class back_threader_registry): Remove m_all_paths.
6017 (thread_jumps::thread_through_all_blocks): Move to back_threader
6019 (fsm_find_thread_path): Remove
6020 (back_threader::maybe_thread): New.
6021 (back_threader::thread_through_all_blocks): Move from
6023 (back_threader_registry::back_threader_registry): Remove
6025 (back_threader_registry::~back_threader_registry): Remove.
6026 (thread_jumps::find_taken_edge): Remove.
6027 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6028 (thread_jumps::maybe_register_path): Remove.
6029 (thread_jumps::handle_phi): Remove.
6030 (handle_assignment_p): Remove.
6031 (thread_jumps::handle_assignment): Remove.
6032 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6033 (thread_jumps::find_jump_threads_backwards): Remove.
6034 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6035 (try_thread_blocks): Rename find_jump_threads_backwards to
6037 (pass_early_thread_jumps::execute): Same.
6039 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6041 * tree-core.h (omp_clause_proc_bind_kind): Add
6042 OMP_CLAUSE_PROC_BIND_PRIMARY.
6043 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6044 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6046 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6048 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6049 fno-common reference.
6050 * config/arc/arc.c (arc_override_options): Remove overriding of
6053 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6056 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6057 If d->testing_p, return true after performing checks instead of
6058 actually expanding the insn.
6059 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6060 !TARGET_AVX512BW and return false.
6062 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6064 * configure.ac (PE linker --disable-dynamicbase support): New check.
6065 * configure: Regenerate.
6066 * config.in: Likewise.
6067 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6068 (LINK_SPEC): Use it.
6069 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6070 (LINK_SPEC): Likewise.
6072 2021-08-12 liuhongt <hongtao.liu@intel.com>
6075 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6076 post_reload define_insn_and_split.
6077 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6078 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6079 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6080 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6081 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6082 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6083 (*avx2_zero_extendv4siv4di2_2): Ditto.
6084 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6085 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6088 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6090 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6093 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6095 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6097 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6099 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6100 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6101 initialization of pcvoid_type_node here...
6102 (altivec_init_builtins): ...from here.
6103 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6104 RS6000_BTI_const_ptr_void.
6105 (pcvoid_type_node): New macro.
6107 2021-08-11 Richard Biener <rguenther@suse.de>
6110 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6111 hard-register accesses.
6113 2021-08-11 Richard Biener <rguenther@suse.de>
6115 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6116 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6117 to determine has_volatile_ops.
6119 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6121 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6123 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6124 Alexandre Oliva <oliva@adacore.com>
6126 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6128 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6129 (update_escape_summary_1): Likewise.
6130 (update_escape_summary): Likewise.
6131 (ipa_merge_modref_summary_after_inlining): Likewise.
6133 2021-08-11 Richard Biener <rguenther@suse.de>
6135 PR middle-end/101858
6136 * fold-const.c (fold_binary_loc): Guard simplification
6137 of X < (cast) (1 << Y) to integer types.
6139 2021-08-11 Richard Biener <rguenther@suse.de>
6141 PR tree-optimization/101861
6142 * tree-vect-stmts.c (vectorizable_load): Fix error in
6143 previous change with regard to gather vectorization.
6145 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6148 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6150 (vdup_n_s16): Likewise.
6151 (vdup_n_s32): Likewise.
6152 (vdup_n_s64): Likewise.
6153 (vdup_n_u8): Likewise.
6154 (vdup_n_u16): Likewise.
6155 (vdup_n_u32): Likewise.
6156 (vdup_n_u64): Likewise.
6157 (vdup_n_p8): Likewise.
6158 (vdup_n_p16): Likewise.
6159 (vdup_n_p64): Likewise.
6160 (vdup_n_f16): Likewise.
6161 (vdup_n_f32): Likewise.
6162 (vdupq_n_s8): Likewise.
6163 (vdupq_n_s16): Likewise.
6164 (vdupq_n_s32): Likewise.
6165 (vdupq_n_s64): Likewise.
6166 (vdupq_n_u8): Likewise.
6167 (vdupq_n_u16): Likewise.
6168 (vdupq_n_u32): Likewise.
6169 (vdupq_n_u64): Likewise.
6170 (vdupq_n_p8): Likewise.
6171 (vdupq_n_p16): Likewise.
6172 (vdupq_n_p64): Likewise.
6173 (vdupq_n_f16): Likewise.
6174 (vdupq_n_f32): Likewise.
6175 (vmov_n_s8): Replace call to builtin with call to corresponding
6177 (vmov_n_s16): Likewise.
6178 (vmov_n_s32): Likewise.
6179 (vmov_n_s64): Likewise.
6180 (vmov_n_u8): Likewise.
6181 (vmov_n_u16): Likewise.
6182 (vmov_n_u32): Likewise.
6183 (vmov_n_u64): Likewise.
6184 (vmov_n_p8): Likewise.
6185 (vmov_n_p16): Likewise.
6186 (vmov_n_f16): Likewise.
6187 (vmov_n_f32): Likewise.
6188 (vmovq_n_s8): Likewise.
6189 (vmovq_n_s16): Likewise.
6190 (vmovq_n_s32): Likewise.
6191 (vmovq_n_s64): Likewise.
6192 (vmovq_n_u8): Likewise.
6193 (vmovq_n_u16): Likewise.
6194 (vmovq_n_u32): Likewise.
6195 (vmovq_n_u64): Likewise.
6196 (vmovq_n_p8): Likewise.
6197 (vmovq_n_p16): Likewise.
6198 (vmovq_n_f16): Likewise.
6199 (vmovq_n_f32): Likewise.
6200 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6202 2021-08-11 liuhongt <hongtao.liu@intel.com>
6205 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6206 when TARGET_AVX512F and TARGET_SSE_MATH.
6208 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6211 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6212 for V32HImode if !TARGET_AVX512BW.
6213 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6214 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6215 early, but actually check the permutation.
6217 2021-08-10 Richard Biener <rguenther@suse.de>
6219 PR tree-optimization/101809
6220 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6221 gathers with offset vector nunits being a constant multiple
6222 of the data vector nunits.
6223 (vect_get_gather_scatter_ops): Use the appropriate nunits
6224 for the offset vector defs.
6225 (vectorizable_store): Adjust call to
6226 vect_get_gather_scatter_ops.
6227 (vectorizable_load): Likewise. Handle the case of less
6228 offset vectors than data vectors.
6230 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6233 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6234 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6237 2021-08-10 Richard Biener <rguenther@suse.de>
6239 PR tree-optimization/101801
6240 PR tree-optimization/101819
6241 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6242 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6243 (vectorizable_reduction): Re-instantiate a check for emulated
6245 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6246 (vectorizable_operation): Likewise. Cost emulated vector
6247 operations according to the scalar sequence synthesized by
6250 2021-08-10 Richard Biener <rguenther@suse.de>
6252 PR middle-end/101824
6253 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6254 volatile in case the variable was.
6256 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6259 * config/i386/constraints.md (BC): Document for integer SSE
6260 constant all bits set operand.
6261 (BF): New constraint for const floating-point all bits set
6263 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6264 (standard_sse_constant_opcode): Likewise.
6265 * config/i386/sse.md (sseconstm1): New mode attribute.
6266 (mov<mode>_internal): Replace BC with <sseconstm1>.
6268 2021-08-10 liuhongt <hongtao.liu@intel.com>
6270 * config/i386/sse.md (cond_<insn><mode>): New expander.
6271 (VI248_AVX512VLBW): New mode iterator.
6272 * config/i386/predicates.md
6273 (nonimmediate_or_const_vec_dup_operand): New predicate.
6275 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6277 PR tree-optimization/101741
6278 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6279 type of parameter for toupper/tolower.
6281 2021-08-09 Martin Jambor <mjambor@suse.cz>
6284 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6286 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6288 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6290 (is_store_insn1): Verify source is a register.
6292 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6295 * config/i386/mmx.md (<any_logic:code>v2sf3):
6296 Rename from *mmx_<any_logic:code>v2sf3
6298 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6300 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6301 'gcc/omp-oacc-neuter-broadcast.cc'.
6302 * omp-low.c: Likewise.
6303 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6306 2021-08-09 Julian Brown <julian@codesourcery.com>
6307 Kwok Cheung Yeung <kcy@codesourcery.com>
6308 Thomas Schwinge <thomas@codesourcery.com>
6310 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6311 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6312 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6313 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6314 (gcn_fork_join): Update comment.
6315 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6316 (macc_experimental_workers): Remove unused option.
6318 2021-08-09 Julian Brown <julian@codesourcery.com>
6319 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6320 Kwok Cheung Yeung <kcy@codesourcery.com>
6321 Thomas Schwinge <thomas@codesourcery.com>
6323 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6324 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6325 Add documentation hook.
6326 * doc/tm.texi: Regenerate.
6327 * omp-oacc-neuter-broadcast.cc: New file.
6328 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6329 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6330 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6331 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
6332 * target.def (goacc.create_worker_broadcast_record): Add target
6334 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
6336 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
6337 Rename prototype to...
6338 (gcn_goacc_create_worker_broadcast_record): ... this.
6339 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
6341 (gcn_goacc_create_worker_broadcast_record): ... this.
6342 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
6344 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
6346 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
6349 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
6352 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6354 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
6356 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6358 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
6359 consider '-foffload-abi'.
6360 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
6361 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
6364 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6366 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
6369 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6371 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
6373 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6375 * doc/gty.texi (Files): Update.
6377 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6379 * doc/gty.texi (Files): Fix GTY header file example.
6381 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
6383 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
6384 determine the upper and lower bounds from a mask-value pair.
6385 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
6386 absolute value and unsigned absolute value expressions.
6387 (bit_value_binop): Initialize *VAL's precision.
6388 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
6389 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
6390 support when the operands are unknown but potentially equal.
6391 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
6393 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
6395 * config/aarch64/aarch64.md
6396 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
6398 2021-08-08 Sergei Trofimovich <siarheit@google.com>
6400 * lra-constraints.c: Fix s/otput/output/ typo.
6402 2021-08-06 Martin Sebor <msebor@redhat.com>
6404 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
6405 (expand_builtin_strcat): Same.
6406 (expand_builtin_stpncpy): Same.
6407 (expand_builtin_strncat): Same.
6408 (check_read_access): Same.
6409 (check_memop_access): Same.
6410 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
6411 (expand_builtin_strnlen): Same.
6412 (expand_builtin_memcpy): Same.
6413 (expand_builtin_memmove): Same.
6414 (expand_builtin_mempcpy): Same.
6415 (expand_builtin_strcpy): Same.
6416 (expand_builtin_strcpy_args): Same.
6417 (expand_builtin_stpcpy_1): Same.
6418 (expand_builtin_strncpy): Same.
6419 (expand_builtin_memset): Same.
6420 (expand_builtin_bzero): Same.
6421 (expand_builtin_strcmp): Same.
6422 (expand_builtin_strncmp): Same.
6423 (expand_builtin): Remove handlers.
6424 (fold_builtin_strlen): Add a comment.
6425 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
6426 * calls.c (maybe_warn_nonstring_arg): Same.
6427 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
6428 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
6429 (gimple_fold_builtin_stpcpy): Same.
6430 * gimple-ssa-warn-access.cc (has_location): New function.
6431 (get_location): Same.
6432 (get_callee_fndecl): Same.
6435 (warn_string_no_nul): Define.
6436 (unterminated_array): Same.
6437 (check_nul_terminated_array): Same.
6438 (maybe_warn_nonstring_arg): Same.
6439 (maybe_warn_for_bound): Same.
6440 (warn_for_access): Same.
6441 (check_access): Same.
6442 (check_memop_access): Same.
6443 (check_read_access): Same.
6444 (warn_dealloc_offset): Use helper functions.
6445 (maybe_emit_free_warning): Same.
6446 (class pass_waccess): Add members.
6447 (check_strcat): New function.
6448 (check_strncat): New function.
6449 (check_stxcpy): New function.
6450 (check_stxncpy): New function.
6451 (check_strncmp): New function.
6452 (pass_waccess::check_builtin): New function.
6453 (pass_waccess::check): Call it.
6454 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
6456 (maybe_warn_for_bound): Same.
6457 (check_access): Same.
6458 (check_memop_access): Same.
6459 (check_read_access): Same.
6460 * pointer-query.h (struct access_data): Define a ctor overload.
6462 2021-08-06 Richard Biener <rguenther@suse.de>
6464 PR tree-optimization/101801
6465 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
6466 (vect_can_vectorize_without_simd_p): ... to this.
6467 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
6468 (vect_can_vectorize_without_simd_p): ... to this and fold
6469 in vect_min_worthwhile_factor.
6470 (vect_min_worthwhile_factor): Remove.
6471 (vectorizable_reduction): Adjust and remove the cost part.
6472 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6473 (vectorizable_operation): Likewise.
6475 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
6478 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
6479 Add general_gr_operand predicate to operand 3.
6481 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
6483 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
6484 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
6486 2021-08-06 Tamar Christina <tamar.christina@arm.com>
6488 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
6489 register_svprfop): Pass vec<> by pointer.
6490 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
6491 * langhooks.c (lhd_simulate_enum_decl): Likewise.
6492 * langhooks.h (struct lang_hooks_for_types): Likewise.
6494 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6496 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
6497 __builtin_memcpy instead of constructing an additional
6498 __builtin_aarch64_simd_oi one vector at a time.
6499 (vst1q_bf16_x2): Likewise.
6500 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
6501 an additional __builtin_aarch64_simd_ci one vector at a time.
6502 (vst1q_bf16_x3): Likewise.
6503 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
6504 (vst1q_bf16_x4): Likewise.
6505 (vst2_bf16): Use __builtin_memcpy instead of constructing an
6506 additional __builtin_aarch64_simd_oi one vector at a time.
6507 (vst2q_bf16): Likewise.
6508 (vst3_bf16): Use __builtin_memcpy instead of constructing an
6509 additional __builtin_aarch64_simd_ci mode one vector at a
6511 (vst3q_bf16): Likewise.
6512 (vst4_bf16): Use __builtin_memcpy instead of constructing an
6513 additional __builtin_aarch64_simd_xi one vector at a time.
6514 (vst4q_bf16): Likewise.
6516 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6518 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
6519 (__ST2Q_LANE_FUNC): Delete.
6520 (vst2_lane_f16): Use __builtin_memcpy to copy vector
6521 structure instead of constructing __builtin_aarch64_simd_oi
6522 one vector at a time.
6523 (vst2_lane_f32): Likewise.
6524 (vst2_lane_f64): Likewise.
6525 (vst2_lane_p8): Likewise.
6526 (vst2_lane_p16): Likewise.
6527 (vst2_lane_p64): Likewise.
6528 (vst2_lane_s8): Likewise.
6529 (vst2_lane_s16): Likewise.
6530 (vst2_lane_s32): Likewise.
6531 (vst2_lane_s64): Likewise.
6532 (vst2_lane_u8): Likewise.
6533 (vst2_lane_u16): Likewise.
6534 (vst2_lane_u32): Likewise.
6535 (vst2_lane_u64): Likewise.
6536 (vst2_lane_bf16): Likewise.
6537 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
6538 structure instead of using a union.
6539 (vst2q_lane_f32): Likewise.
6540 (vst2q_lane_f64): Likewise.
6541 (vst2q_lane_p8): Likewise.
6542 (vst2q_lane_p16): Likewise.
6543 (vst2q_lane_p64): Likewise.
6544 (vst2q_lane_s8): Likewise.
6545 (vst2q_lane_s16): Likewise.
6546 (vst2q_lane_s32): Likewise.
6547 (vst2q_lane_s64): Likewise.
6548 (vst2q_lane_u8): Likewise.
6549 (vst2q_lane_u16): Likewise.
6550 (vst2q_lane_u32): Likewise.
6551 (vst2q_lane_u64): Likewise.
6552 (vst2q_lane_bf16): Likewise.
6554 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6556 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
6557 (__ST3Q_LANE_FUNC): Delete.
6558 (vst3_lane_f16): Use __builtin_memcpy to copy vector
6559 structure instead of constructing __builtin_aarch64_simd_ci
6560 one vector at a time.
6561 (vst3_lane_f32): Likewise.
6562 (vst3_lane_f64): Likewise.
6563 (vst3_lane_p8): Likewise.
6564 (vst3_lane_p16): Likewise.
6565 (vst3_lane_p64): Likewise.
6566 (vst3_lane_s8): Likewise.
6567 (vst3_lane_s16): Likewise.
6568 (vst3_lane_s32): Likewise.
6569 (vst3_lane_s64): Likewise.
6570 (vst3_lane_u8): Likewise.
6571 (vst3_lane_u16): Likewise.
6572 (vst3_lane_u32): Likewise.
6573 (vst3_lane_u64): Likewise.
6574 (vst3_lane_bf16): Likewise.
6575 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
6576 structure instead of using a union.
6577 (vst3q_lane_f32): Likewise.
6578 (vst3q_lane_f64): Likewise.
6579 (vst3q_lane_p8): Likewise.
6580 (vst3q_lane_p16): Likewise.
6581 (vst3q_lane_p64): Likewise.
6582 (vst3q_lane_s8): Likewise.
6583 (vst3q_lane_s16): Likewise.
6584 (vst3q_lane_s32): Likewise.
6585 (vst3q_lane_s64): Likewise.
6586 (vst3q_lane_u8): Likewise.
6587 (vst3q_lane_u16): Likewise.
6588 (vst3q_lane_u32): Likewise.
6589 (vst3q_lane_u64): Likewise.
6590 (vst3q_lane_bf16): Likewise.
6592 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6594 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
6595 (__ST4Q_LANE_FUNC): Delete.
6596 (vst4_lane_f16): Use __builtin_memcpy to copy vector
6597 structure instead of constructing __builtin_aarch64_simd_xi
6598 one vector at a time.
6599 (vst4_lane_f32): Likewise.
6600 (vst4_lane_f64): Likewise.
6601 (vst4_lane_p8): Likewise.
6602 (vst4_lane_p16): Likewise.
6603 (vst4_lane_p64): Likewise.
6604 (vst4_lane_s8): Likewise.
6605 (vst4_lane_s16): Likewise.
6606 (vst4_lane_s32): Likewise.
6607 (vst4_lane_s64): Likewise.
6608 (vst4_lane_u8): Likewise.
6609 (vst4_lane_u16): Likewise.
6610 (vst4_lane_u32): Likewise.
6611 (vst4_lane_u64): Likewise.
6612 (vst4_lane_bf16): Likewise.
6613 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
6614 structure instead of using a union.
6615 (vst4q_lane_f32): Likewise.
6616 (vst4q_lane_f64): Likewise.
6617 (vst4q_lane_p8): Likewise.
6618 (vst4q_lane_p16): Likewise.
6619 (vst4q_lane_p64): Likewise.
6620 (vst4q_lane_s8): Likewise.
6621 (vst4q_lane_s16): Likewise.
6622 (vst4q_lane_s32): Likewise.
6623 (vst4q_lane_s64): Likewise.
6624 (vst4q_lane_u8): Likewise.
6625 (vst4q_lane_u16): Likewise.
6626 (vst4q_lane_u32): Likewise.
6627 (vst4q_lane_u64): Likewise.
6628 (vst4q_lane_bf16): Likewise.
6630 2021-08-06 Martin Liska <mliska@suse.cz>
6632 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
6633 a target option is restored, it can have
6634 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
6635 and error should not be emitted.
6637 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
6639 * gcov-io.h (gcov_write): Declare.
6640 * gcov-io.c (gcov_write): New.
6641 (gcov_write_counter): Remove.
6642 (gcov_write_tag_length): Likewise.
6643 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
6644 gcov_write_unsigned().
6645 * doc/invoke.texi (fprofile-info-section): Mention
6646 __gcov_info_to_gdca().
6648 2021-08-06 Martin Sebor <msebor@redhat.com>
6650 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
6651 arguments to by-reference.
6652 (iterate_fix_dominators): Same.
6653 * dominance.h (iterate_fix_dominators): Same.
6654 * ipa-prop.h: Call auto_vec::to_vec_legacy.
6655 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
6656 arguments to by-reference.
6657 (debug_data_dependence_relation): Same.
6658 (dump_data_dependence_relations): Same.
6659 * tree-data-ref.h (debug_data_dependence_relation): Same.
6660 (dump_data_dependence_relations): Same.
6661 * tree-predcom.c (dump_chains): Same.
6662 (initialize_root_vars_lm): Same.
6663 (determine_unroll_factor): Same.
6664 (replace_phis_by_defined_names): Same.
6665 (insert_init_seqs): Same.
6666 (pcom_worker::tree_predictive_commoning_loop): Call
6667 auto_vec::to_vec_legacy.
6668 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
6669 arguments to by-reference.
6670 * tree-ssa-threadbackward.c (populate_worklist): Same.
6671 (back_threader::resolve_def): Same.
6672 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
6673 (vect_enhance_data_refs_alignment): Same.
6674 (vect_check_lower_bound): Same.
6675 (vect_prune_runtime_alias_test_list): Same.
6676 (vect_permute_store_chain): Same.
6677 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
6678 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
6679 * tree-vectorizer.h (vect_permute_store_chain): Same.
6680 * vec.c (test_init): New function.
6681 (vec_c_tests): Call new function.
6682 * vec.h (vec): Declare ctors, dtor, and assignment.
6683 (auto_vec::vec_to_legacy): New function.
6684 (vec::copy): Adjust initialization.
6686 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
6689 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
6690 callee only uses GPRs.
6691 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
6692 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
6693 * config/i386/x86gprintrin.h: Add
6694 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
6695 to disable non-GPR ISAs.
6697 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
6699 PR middle-end/101787
6700 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
6702 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
6704 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
6705 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
6706 (vect_is_extending_load, vect_is_integer_truncation): New functions,
6707 moved from aarch64.c but given different names.
6708 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
6709 (aarch64_is_reduction, aarch64_reduc_type)
6710 (aarch64_embedded_comparison_type, aarch64_comparison_type)
6711 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
6712 in favor of the above. Update callers accordingly.
6714 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
6717 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
6718 writing .cpu directive in asm output.
6719 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
6720 (arm_last_printed_arch_string): Delete.
6721 (arm_last-printed_fpu_string): Delete.
6722 (arm_configure_build_target): If use of floating-point/SIMD is
6723 disabled, remove all fp/simd related features from the target ISA.
6724 (last_arm_targ_options): New variable.
6725 (arm_print_asm_arch_directives): Add new parameters. Change order
6726 of emitted directives and handle all cases here.
6727 (arm_file_start): Always call arm_print_asm_arch_directives, move
6728 all generation of .arch/.arch_extension here.
6729 (arm_file_end): Call arm_print_asm_arch.
6730 (arm_declare_function_name): Call arm_print_asm_arch_directives
6731 instead of printing .arch/.fpu directives directly.
6733 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
6735 * config/arm/arm.c (arm_configure_build_target): Don't call
6736 arm_option_reconfigure_globals.
6737 (arm_option_restore): Call arm_option_reconfigure_globals after
6738 reconfiguring the target.
6739 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
6741 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
6743 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
6744 arch_name is always set.
6746 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
6748 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
6749 of vec_select high-half from being added into Neon subtract
6752 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
6754 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
6755 of vec_select high-half from being added into Neon add cost.
6757 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
6759 * cfgloop.h (loops_list::loops_list): Add one optional argument
6760 root and adjust accordingly, update loop tree walking and factor
6762 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
6764 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
6766 PR tree-optimization/101626
6767 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
6768 reverse scalar storage order on a pointer or vector component.
6770 2021-08-05 liuhongt <hongtao.liu@intel.com>
6772 * config/i386/sse.md (cond_<code><mode>): New expander.
6774 2021-08-05 liuhongt <hongtao.liu@intel.com>
6776 * config/i386/sse.md (cond_<code><mode>): New expander.
6778 2021-08-05 liuhongt <hongtao.liu@intel.com>
6780 * config/i386/sse.md (cond_<code><mode>): New expander.
6782 2021-08-04 David Malcolm <dmalcolm@redhat.com>
6785 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
6787 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
6790 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
6791 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
6793 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
6796 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
6797 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
6798 data with SSE register from one memory location to another.
6800 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6802 * config/s390/s390.c (expand_perm_with_vpdi): New function.
6803 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
6804 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
6805 parameterized expander.
6806 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
6808 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6810 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
6811 (struct expand_vec_perm_d): Define struct.
6812 (expand_perm_with_merge): New function.
6813 (vectorize_vec_perm_const_1): New function.
6814 (s390_vectorize_vec_perm_const): New function.
6815 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
6817 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6819 * config/s390/vector.md (V_HW_64): Remove mode iterator.
6820 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
6821 * config/s390/vx-builtins.md
6822 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
6825 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6827 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
6829 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
6831 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
6832 instead of an unspec.
6834 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6836 * config/s390/s390-modes.def: Add more vector modes to support
6837 concatenation of two vectors.
6838 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
6840 (s390_expand_merge): Likewise.
6841 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
6842 (s390_expand_merge): New function.
6843 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
6844 Remove constant definitions.
6845 * config/s390/vector.md (V_HW_2): Add mode iterators.
6846 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
6847 (vec_2x_nelts, vec_2x_wide): New mode attributes.
6848 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
6849 New pattern definitions.
6850 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
6851 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
6852 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
6853 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
6855 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
6857 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
6858 emit vec merge pattern.
6860 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
6862 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
6864 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
6865 vec_select high-half from being added into Neon multiply
6867 * rtlanal.c (vec_series_highpart_p): Define.
6868 * rtlanal.h (vec_series_highpart_p): Declare.
6870 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
6872 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
6874 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
6875 vec_select cost from being added into Neon multiply cost.
6877 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
6879 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
6880 which old_loop_vinfo is an epilogue loop that handles a constant
6881 number of iterations.
6883 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
6885 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
6886 when a reanalyzed loop fails to be cheaper than the current
6889 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
6891 * config/aarch64/aarch64.c: Fix a typo.
6893 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
6895 PR gcov-profile/101773
6896 * gcov-io.c (gcov_close): Check return code of a fclose.
6898 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
6901 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
6902 .file statement when needed.
6904 2021-08-04 Richard Biener <rguenther@suse.de>
6906 * tree-vect-data-refs.c (vect_check_gather_scatter):
6907 Include widening conversions only when the result is
6908 still handed by native gather or the current offset
6909 size not already matches the data size.
6910 Also succeed analysis in case there's no native support,
6911 noted by a IFN_LAST ifn and a NULL decl.
6912 (vect_analyze_data_refs): Always consider gathers.
6913 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
6914 Test for no IFN gather rather than decl gather.
6915 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
6916 gather-scatter info and cost emulated gathers accordingly.
6917 (vect_truncate_gather_scatter_offset): Properly test for
6919 (vect_use_strided_gather_scatters_p): Likewise.
6920 (get_load_store_type): Handle emulated gathers and its
6922 (vectorizable_load): Likewise. Emulate them by extracting
6923 scalar offsets, doing scalar loads and a vector construct.
6925 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
6928 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
6929 argument to set m_max_size.
6930 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
6931 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
6932 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
6934 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
6935 Marc Glisse <marc.glisse@inria.fr>
6937 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
6938 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
6939 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
6941 2021-08-04 Richard Biener <rguenther@suse.de>
6943 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
6944 out code to decompose vector loads ...
6945 (optimize_vector_load): ... here. Generalize it to
6946 handle intermediate widening and TARGET_MEM_REF loads
6947 and apply it to loads with a supported vector mode as well.
6949 2021-08-04 Richard Biener <rguenther@suse.de>
6951 PR tree-optimization/101756
6952 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
6953 the result of the reduction epilogue is compatible to the original
6956 2021-08-04 liuhongt <hongtao.liu@intel.com>
6959 * config/i386/i386.md (peephole2): Refine predicate from
6960 register_operand to general_reg_operand.
6962 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
6964 * gimple-range-path.h (path_range_query::dump): Mark override.
6966 2021-08-04 Richard Biener <rguenther@suse.de>
6968 PR tree-optimization/101769
6969 * tree-tailcall.c (eliminate_tail_call): Add the created loop
6970 for the first recursion and return it via the new output parameter.
6971 (optimize_tail_call): Pass through new output param.
6972 (tree_optimize_tail_calls_1): After creating all latches,
6973 add the created loop to the loop tree. Do not mark loops for fixup.
6975 2021-08-04 Martin Liska <mliska@suse.cz>
6977 * doc/invoke.texi: Document threader-mode param.
6979 2021-08-04 liuhongt <hongtao.liu@intel.com>
6981 * config/i386/sse.md (cond_fma<mode>): New expander.
6982 (cond_fms<mode>): Ditto.
6983 (cond_fnma<mode>): Ditto.
6984 (cond_fnms<mode>): Ditto.
6986 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
6988 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
6990 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
6992 * config/rs6000/constraints.md: Remove "e" from the list of available
6993 constraint characters.
6995 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
6997 PR gcov-profile/71672
6998 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7000 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7002 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7004 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7006 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7007 try XMM31 to avoid vzeroupper.
7009 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7011 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7012 -mcpu=neoverse-512tvb.
7013 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7014 * config/aarch64/aarch64-tune.md: Regenerate.
7015 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7016 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7017 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7018 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7019 (aarch64_adjust_body_cost): Likewise.
7021 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7023 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7024 record issue information for operations that occur in the
7027 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7029 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7030 parameter. Detect cases in which an Advanced SIMD MLA would almost
7031 certainly require a MOV.
7032 (aarch64_count_ops): Update accordingly.
7034 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7036 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7037 function, split out from...
7038 (aarch64_detect_vector_stmt_subtype): ...here.
7039 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7041 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7043 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7044 Add gather_load_x32_cost and gather_load_x64_cost.
7045 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7046 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7047 accordingly, using the values given by the scalar_load * number
7048 of elements calculation that we used previously.
7049 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7051 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7053 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7054 function, split out from...
7055 (aarch64_adjust_body_cost): ...here.
7057 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7059 * config/aarch64/fractional-cost.h: New file.
7060 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7061 and cost_fraction.h.
7062 (vec_cost_fraction): New typedef.
7063 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7064 (aarch64_detect_vector_stmt_subtype): Likewise.
7065 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7066 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7068 (aarch64_adjust_body_cost): Likewise.
7069 (aarch64_test_cost_fraction): New function.
7070 (aarch64_run_selftests): Call it.
7072 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7074 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7076 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7078 (aarch64_estimated_poly_value): Likewise. Use the least significant
7079 set bit for the minimum and likely values. Use the most significant
7080 set bit for the maximum value.
7082 2021-08-03 liuhongt <hongtao.liu@intel.com>
7084 * config/i386/sse.md (cond_<insn><mode>): New expander.
7085 (cond_mul<mode>): Ditto.
7087 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7089 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7091 2021-08-03 liuhongt <hongtao.liu@intel.com>
7093 * config/i386/sse.md (cond_<insn><mode>):New expander.
7094 (cond_mul<mode>): Ditto.
7095 (cond_div<mode>): Ditto.
7097 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7099 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7100 check stack_realign_needed for stack realignment.
7101 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7102 than the largest integer supported by vector register.
7103 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7104 (MOVE_MAX): Set to bytes of the largest integer supported by
7106 (STORE_MAX_PIECES): New.
7108 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7110 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7111 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7112 data from one memory location to another.
7114 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7117 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7119 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7121 PR tree-optimization/101724
7122 * params.opt: Remove --param=threader-iterative.
7123 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7126 2021-08-02 Tom de Vries <tdevries@suse.de>
7128 PR middle-end/101665
7129 * doc/extend.texi (nonnull attribute): Improve documentation.
7131 2021-08-02 Andrew Pinski <apinski@marvell.com>
7133 PR rtl-optimization/101683
7134 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7136 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7138 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7139 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7140 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7141 (tree_ssa_phiop_worker): Update call to function above.
7143 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7146 * config/i386/i386.md (bsr_rex64_1_zext): New.
7147 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7148 with gen_bsr_rex64_1_zext.
7150 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7153 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7154 define_insn patterns.
7155 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7156 Add combine splitters for constant - clz.
7157 (clz<mode>2): Use a temporary pseudo for bsr result.
7159 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7161 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7162 _mm_floor_sd, _mm_floor_ss): New.
7164 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7166 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7167 _mm_ceil_sd, _mm_ceil_ss): New.
7169 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7171 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7172 _mm_blend_ps, _mm_blendv_ps): New.
7174 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7175 Uroš Bizjak <ubizjak@gmail.com>
7177 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7178 to generate a conditional move using the carry flag after sub $1.
7179 (peephole2): Eliminate a register-to-register move by inverting
7180 the condition of a conditional move.
7182 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7184 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7185 ("*call_value_real"): Don't generate rtx mentioning the generic
7186 operands 1 and 2 to "call", and similarly for "call_value".
7187 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7188 (mmix_print_operand): Use '!' instead of 'p'.
7190 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7192 * doc/md.texi (call): Correct information about operand 2.
7193 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7195 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7197 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7199 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7201 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7203 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7205 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7207 (*::get_bb_range): Ditto.
7208 (*::bb_range_p): Ditto.
7209 * gimple-range-cache.h: Change prototypes.
7211 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7214 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7215 from scalar_int_mode to fixed_size_mode.
7216 (builtin_strncpy_read_str): Likewise.
7217 (gen_memset_value_from_prev): New function.
7218 (builtin_memset_read_str): Change the mode argument from
7219 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7220 and support CONST_VECTOR.
7221 (builtin_memset_gen_str): Likewise.
7222 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7224 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7225 with fixed_size_mode.
7226 (builtin_memset_read_str): Likewise.
7227 * expr.c (widest_int_mode_for_size): Renamed to ...
7228 (widest_fixed_size_mode_for_size): Add a bool argument to
7229 indicate if QI vector mode can be used.
7230 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7231 instead of widest_int_mode_for_size.
7232 (pieces_addr::adjust): Change the mode argument from
7233 scalar_int_mode to fixed_size_mode.
7234 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7235 m_qi_vector_mode, to indicate that QI vector mode can be used.
7236 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7237 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7238 instead of widest_int_mode_for_size.
7239 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7240 scalar_int_mode to fixed_size_mode. Call
7241 widest_fixed_size_mode_for_size instead of
7242 widest_int_mode_for_size.
7243 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7244 function to return the smallest integer or QI vector mode.
7245 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7246 instead of widest_int_mode_for_size. Call
7247 smallest_fixed_size_mode_for_size instead of
7248 smallest_int_mode_for_size.
7249 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7250 indicate that QI vector mode can be used and pass it to
7251 op_by_pieces_d::op_by_pieces_d.
7252 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7253 instead of widest_int_mode_for_size. Pass memsetp to
7254 widest_fixed_size_mode_for_size to support QI vector mode.
7255 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7256 (store_by_pieces): Pass memsetp to
7257 store_by_pieces_d::store_by_pieces_d.
7258 (clear_by_pieces_1): Removed.
7259 (clear_by_pieces): Replace clear_by_pieces_1 with
7260 builtin_memset_read_str and pass true to store_by_pieces_d to
7261 support vector mode broadcast.
7262 (string_cst_read_str): Change the mode argument from
7263 scalar_int_mode to fixed_size_mode.
7264 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7266 (by_pieces_prev): Likewise.
7267 * rtl.h (lowpart_subreg_regno): New.
7268 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7269 simplify_subreg_regno.
7270 * target.def (gen_memset_scratch_rtx): New hook.
7271 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7272 * doc/tm.texi: Regenerated.
7274 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7277 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7278 TARGET_EXPR instead of MODIFY_EXPR.
7280 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7283 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7284 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7285 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7287 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7289 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7292 * config/i386/i386-options.c (ix86_option_override_internal):
7293 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7295 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7298 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7299 explicitly dereferencing __a.
7300 (vld1_s64): Likewise.
7301 (vld1_u64): Likewise.
7302 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7303 and change to VAR13.
7305 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7307 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7308 use of m_range_analyzer.
7309 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7310 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7311 (loop_versioning::prune_loop_conditions): Replace vr_values use
7312 with range_query interface.
7313 (pass_loop_versioning::execute): Use ranger.
7315 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7318 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7319 enum values, and emit a warning if they mismatch.
7321 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7323 * cfgloop.h (as_const): New function.
7324 (class loop_iterator): Rename to ...
7325 (class loops_list): ... this.
7326 (loop_iterator::next): Rename to ...
7327 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7328 (loop_iterator::loop_iterator): Rename to ...
7329 (loops_list::loops_list): ... this and adjust.
7330 (loops_list::Iter): New class.
7331 (loops_list::iterator): New type.
7332 (loops_list::const_iterator): New type.
7333 (loops_list::begin): New function.
7334 (loops_list::end): Likewise.
7335 (loops_list::begin const): Likewise.
7336 (loops_list::end const): Likewise.
7337 (FOR_EACH_LOOP): Remove.
7338 (FOR_EACH_LOOP_FN): Remove.
7339 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
7340 for loop with loops_list instance.
7341 (sort_sibling_loops): Likewise.
7342 (disambiguate_loops_with_multiple_latches): Likewise.
7343 (verify_loop_structure): Likewise.
7344 * cfgloopmanip.c (create_preheaders): Likewise.
7345 (force_single_succ_latches): Likewise.
7346 * config/aarch64/falkor-tag-collision-avoidance.c
7347 (execute_tag_collision_avoidance): Likewise.
7348 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
7349 * config/s390/s390.c (s390_adjust_loops): Likewise.
7350 * doc/loop.texi: Likewise.
7351 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
7352 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
7353 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
7354 (loop_versioning::make_versioning_decisions): Likewise.
7355 * gimple-ssa-split-paths.c (split_paths): Likewise.
7356 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
7357 * graphite.c (canonicalize_loop_form): Likewise.
7358 (graphite_transform_loops): Likewise.
7359 * ipa-fnsummary.c (analyze_function_body): Likewise.
7360 * ipa-pure-const.c (analyze_function): Likewise.
7361 * loop-doloop.c (doloop_optimize_loops): Likewise.
7362 * loop-init.c (loop_optimizer_finalize): Likewise.
7363 (fix_loop_structure): Likewise.
7364 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
7365 (move_loop_invariants): Likewise.
7366 * loop-unroll.c (decide_unrolling): Likewise.
7367 (unroll_loops): Likewise.
7368 * modulo-sched.c (sms_schedule): Likewise.
7369 * predict.c (predict_loops): Likewise.
7370 (pass_profile::execute): Likewise.
7371 * profile.c (branch_prob): Likewise.
7372 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
7373 (sel_find_rgns): Likewise.
7374 * tree-cfg.c (replace_loop_annotate): Likewise.
7375 (replace_uses_by): Likewise.
7376 (move_sese_region_to_fn): Likewise.
7377 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
7378 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
7379 * tree-parloops.c (parallelize_loops): Likewise.
7380 * tree-predcom.c (tree_predictive_commoning): Likewise.
7381 * tree-scalar-evolution.c (scev_initialize): Likewise.
7382 (scev_reset): Likewise.
7383 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
7384 * tree-ssa-live.c (remove_unused_locals): Likewise.
7385 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
7386 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
7387 (tree_ssa_lim_initialize): Likewise.
7388 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
7389 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
7390 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
7391 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
7392 (free_numbers_of_iterations_estimates): Likewise.
7393 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
7394 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
7395 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
7396 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
7397 (pass_scev_cprop::execute): Likewise.
7398 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
7399 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
7400 * tree-ssa-threadupdate.c
7401 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
7402 * tree-vectorizer.c (vectorize_loops): Likewise.
7403 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
7405 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
7407 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
7408 generating a VOIDmode register for e.g the
7409 function_arg_info::end_marker.
7411 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
7413 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
7414 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
7415 (cc mode_attr): Similarly.
7416 (ccz subst_attr): Similarly.
7417 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
7418 * config/h8300/testcompare.md: Remove various cc0 based patterns
7419 that had been commented out. Add pattern to set CCZ from a bit
7422 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
7423 Julian Brown <julian@codesourcery.com>
7424 Kwok Cheung Yeung <kcy@codesourcery.com>
7426 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
7427 'update_stmt' after modification.
7428 (pass_oacc_loop_designation): New function, extracted out of...
7429 (pass_oacc_device_lower): ... this.
7430 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
7431 (make_pass_oacc_loop_designation): New
7432 * passes.def: Add it.
7433 * tree-parloops.c (create_parallel_loop): Adjust.
7434 * tree-pass.h (make_pass_oacc_loop_designation): New.
7436 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
7438 * flag-types.h (enum threader_mode): New.
7439 * params.opt: Add entry for --param=threader-mode.
7440 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
7441 (class back_threader): New.
7442 (back_threader::back_threader): New.
7443 (back_threader::~back_threader): New.
7444 (back_threader::maybe_register_path): New.
7445 (back_threader::find_taken_edge): New.
7446 (back_threader::find_taken_edge_switch): New.
7447 (back_threader::find_taken_edge_cond): New.
7448 (back_threader::resolve_def): New.
7449 (back_threader::resolve_phi): New.
7450 (back_threader::find_paths_to_names): New.
7451 (back_threader::find_paths): New.
7454 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
7455 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
7456 (pass_thread_jumps::execute): Abstract out code...
7457 (try_thread_blocks): ...here.
7458 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
7459 Abstract out threading candidate code to...
7460 (single_succ_to_potentially_threadable_block): ...here.
7461 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
7463 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
7464 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7465 Return bool from register_jump_thread.
7467 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
7469 * target.def: in0 and in1 do not need to be registers.
7470 * doc/tm.texi: Regenerate.
7472 2021-07-29 liuhongt <hongtao.liu@intel.com>
7475 * config/i386/i386.c (ix86_widen_mult_cost): New function.
7476 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
7479 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
7482 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
7483 (rs6000_preferred_doloop_mode): New hook.
7484 * doc/tm.texi: Regenerate.
7485 * doc/tm.texi.in: Add hook preferred_doloop_mode.
7486 * target.def (preferred_doloop_mode): New hook.
7487 * targhooks.c (default_preferred_doloop_mode): New hook.
7488 * targhooks.h (default_preferred_doloop_mode): New hook.
7489 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
7490 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
7491 and compute_doloop_base_on_mode.
7493 2021-07-28 Martin Sebor <msebor@redhat.com>
7495 PR middle-end/101494
7496 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
7497 and size computation.
7499 2021-07-28 Martin Sebor <msebor@redhat.com>
7501 PR middle-end/101601
7502 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
7504 Handle pointers to functions.
7506 2021-07-28 Martin Sebor <msebor@redhat.com>
7508 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
7509 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
7510 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
7511 (access_ref::access_ref): Same.
7512 (access_ref::phi): Same.
7513 (access_ref::get_ref): Same.
7514 (access_ref::size_remaining): Same.
7515 (access_ref::offset_in_range): Same.
7516 (access_ref::add_offset): Same.
7517 (access_ref::inform_access): Same.
7518 (ssa_name_limit_t::visit_phi): Same.
7519 (ssa_name_limit_t::leave_phi): Same.
7520 (ssa_name_limit_t::next): Same.
7521 (ssa_name_limit_t::next_phi): Same.
7522 (ssa_name_limit_t::~ssa_name_limit_t): Same.
7523 (pointer_query::pointer_query): Same.
7524 (pointer_query::get_ref): Same.
7525 (pointer_query::put_ref): Same.
7526 (pointer_query::flush_cache): Same.
7527 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
7528 (check_nul_terminated_array): Same.
7529 (unterminated_array): Same.
7530 (maybe_warn_for_bound): Same.
7531 (check_read_access): Same.
7532 (warn_for_access): Same.
7533 (get_size_range): Same.
7534 (check_access): Same.
7535 (gimple_call_alloc_size): Move to tree.c.
7536 (gimple_parm_array_size): Move to pointer-query.cc.
7537 (get_offset_range): Same.
7538 (gimple_call_return_array): Same.
7539 (handle_min_max_size): Same.
7540 (handle_array_ref): Same.
7541 (handle_mem_ref): Same.
7542 (compute_objsize): Same.
7543 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
7544 (call_dealloc_argno): Same.
7545 (fndecl_dealloc_argno): Same.
7546 (new_delete_mismatch_p): Same.
7547 (matching_alloc_calls_p): Same.
7548 (warn_dealloc_offset): Same.
7549 (maybe_emit_free_warning): Same.
7550 * builtins.h (check_nul_terminated_array): Move to
7551 gimple-ssa-warn-access.h.
7552 (check_nul_terminated_array): Same.
7553 (warn_string_no_nul): Same.
7554 (unterminated_array): Same.
7555 (class ssa_name_limit_t): Same.
7556 (class pointer_query): Same.
7557 (struct access_ref): Same.
7558 (class range_query): Same.
7559 (struct access_data): Same.
7560 (gimple_call_alloc_size): Same.
7561 (gimple_parm_array_size): Same.
7562 (compute_objsize): Same.
7563 (class access_data): Same.
7564 (maybe_emit_free_warning): Same.
7565 * calls.c (initialize_argument_information): Remove call to
7566 maybe_emit_free_warning.
7567 * gimple-array-bounds.cc: Include new header..
7568 * gimple-fold.c: Same.
7569 * gimple-ssa-sprintf.c: Same.
7570 * gimple-ssa-warn-restrict.c: Same.
7571 * passes.def: Add pass_warn_access.
7572 * tree-pass.h (make_pass_warn_access): Declare.
7573 * tree-ssa-strlen.c: Include new headers.
7574 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
7575 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
7576 * gimple-ssa-warn-access.cc: New file.
7577 * gimple-ssa-warn-access.h: New file.
7578 * pointer-query.cc: New file.
7579 * pointer-query.h: New file.
7581 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7583 PR middle-end/101624
7584 * ubsan.c (maybe_instrument_pointer_overflow,
7585 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
7586 PARM_DECLs or RESULT_DECLs.
7587 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
7589 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7591 PR middle-end/101642
7592 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
7593 to type of bswap16 for comparison.
7594 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
7596 2021-07-28 Richard Biener <rguenther@suse.de>
7598 PR tree-optimization/101615
7599 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
7600 at CTOR SLP graph entries.
7602 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7604 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
7605 Add "r,w" alternative.
7607 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
7610 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
7611 AVX_U128_DIRTY when all bits are zero.
7613 2021-07-28 Richard Biener <rguenther@suse.de>
7615 PR tree-optimization/101615
7616 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
7617 external nodes cannot be permuted so make them perm_out 0.
7619 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
7622 * config.in: Regenerate.
7623 * config/gcn/gcn-hsa.h (A_FIJI): New define.
7624 (A_900): New define.
7625 (A_906): New define.
7626 (A_908): New define.
7627 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
7628 * config/gcn/gcn.c (output_file_start): Adjust attributes according
7629 to the assembler capabilities.
7630 * config/gcn/mkoffload.c (main): Likewise.
7631 * configure: Regenerate.
7632 * configure.ac: Add tests for LLVM assembler attribute features.
7634 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
7636 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
7637 cond_false and cond_true on branches.
7639 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
7641 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
7644 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
7646 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
7648 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7651 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
7652 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
7653 and has special !TARGET_AVX512VL expansion.
7654 (vashrv2di3<mask_name>): Rename to ...
7655 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
7656 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
7658 2021-07-28 Martin Uecker <muecker@gwdg.de>
7660 * calls.c (maybe_warn_rdwr_sizes): Correct argument
7661 numbers in warning that were switched.
7663 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
7665 PR tree-optimization/101596
7666 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
7667 by using new_type's precision instead.
7669 2021-07-28 liuhongt <hongtao.liu@intel.com>
7672 * config/i386/i386.h (processor_costs): Add new member
7674 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
7675 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
7676 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
7677 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
7678 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
7679 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
7680 generic_cost, core_cost): Initialize integer_to_sse same value
7682 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
7683 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7684 Use integer_to_sse instead of sse_op to calculate the cost of
7687 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7689 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
7691 (write_init_file): Call write_ovld_static_init.
7693 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7695 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
7697 (write_init_file): Call write_bif_static_init.
7699 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7701 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
7702 (TYPE_MAP_SIZE): New macro.
7703 (type_map): New initialized variable.
7704 (typemap_cmp): New function.
7705 (write_type_node): Likewise.
7706 (write_fntype_init): Implement.
7708 2021-07-27 Martin Sebor <msebor@redhat.com>
7710 PR tree-optimization/101584
7711 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
7712 (check_defs): Call it.
7714 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
7716 * tree-ssa-dom.c (dom_jump_threader_simplifier):
7717 Put avail_exprs_stack in the class, instead of passing it to
7718 jump_threader_simplifier.
7719 (dom_jump_threader_simplifier::simplify): Add state argument.
7720 (dom_opt_dom_walker): Add state.
7721 (pass_dominator::execute): Pass state to threader.
7722 (dom_opt_dom_walker::before_dom_children): Use state.
7723 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
7725 (jump_threader::record_temporary_equivalences_from_phis):
7726 Register equivalences through the state variable.
7727 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
7728 Record ranges in a statement through the state variable.
7729 (jump_threader::simplify_control_stmt_condition): Pass state to
7731 (jump_threader::simplify_control_stmt_condition_1): Same.
7732 (jump_threader::thread_around_empty_blocks): Remove obsolete
7734 (jump_threader::thread_through_normal_block): Record equivalences
7735 on edge through the state variable.
7736 (jump_threader::thread_across_edge): Abstract state pushing.
7737 (jt_state::jt_state): New.
7738 (jt_state::push): New.
7739 (jt_state::pop): New.
7740 (jt_state::register_equiv): New.
7741 (jt_state::record_ranges_from_stmt): New.
7742 (jt_state::register_equivs_on_edge): New.
7743 (jump_threader_simplifier::jump_threader_simplifier): Move from
7745 (jump_threader_simplifier::simplify): Add state argument.
7746 * tree-ssa-threadedge.h (class jt_state): New.
7747 (class jump_threader): Add state to constructor.
7748 (class jump_threader_simplifier): Add state to simplify. Remove
7749 avail_exprs_stack from class.
7750 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
7752 (vrp_jump_threader::vrp_jump_threader): Add state.
7753 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
7755 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
7757 * Makefile.in (OBJS): Add gimple-range-path.o.
7758 * gimple-range-path.cc: New file.
7759 * gimple-range-path.h: New file.
7761 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
7763 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
7764 inside vec_duplicate for all patterns.
7765 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
7766 Push sign/zero-extension inside vec_duplicate.
7768 2021-07-27 Richard Biener <rguenther@suse.de>
7770 PR tree-optimization/101573
7771 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
7772 looking at uninitialized PHI arg defs in some constrained cases.
7773 (warn_uninitialized_vars): Call it.
7774 (execute_early_warn_uninitialized): Calculate dominators.
7776 2021-07-27 Richard Biener <rguenther@suse.de>
7778 PR tree-optimization/39821
7779 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
7780 vector_stmt for widening arithmetic.
7781 (vectorizable_conversion): Adjust.
7783 2021-07-27 Martin Jambor <mjambor@suse.cz>
7785 * cgraph.h (ipa_replace_map): New field force_load_ref.
7786 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
7787 aded new flag load_dereferenced, adjusted comments.
7788 (ipa_get_param_dereferenced): New function.
7789 (ipa_set_param_dereferenced): Likewise.
7790 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
7791 * ipa-cp.c: Include gimple.h.
7792 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
7793 (get_replacement_map): New parameter force_load_ref, set the
7794 appropriate flag in ipa_replace_map if set.
7795 (struct symbol_and_index_together): New type.
7796 (adjust_refs_in_act_callers): New function.
7797 (adjust_references_in_caller): Likewise.
7798 (create_specialized_node): When appropriate, call
7799 adjust_references_in_caller and force only load references.
7800 * ipa-prop.c (load_from_dereferenced_name): New function.
7801 (ipa_analyze_controlled_uses): Also detect loads from a
7802 dereference, harden testing of call statements.
7803 (ipa_write_node_info): Stream the dereferenced flag.
7804 (ipa_read_node_info): Likewise.
7805 (ipa_set_jf_constant): Also create refdesc when jump function
7806 references a variable.
7807 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
7808 also on references of variables and return a symtab_node. Adjust
7810 (propagate_controlled_uses): Also remove references to VAR_DECLs.
7812 2021-07-27 Jakub Jelinek <jakub@redhat.com>
7814 PR middle-end/101586
7815 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
7816 positions above or equal to sz except for diagnostics of flexible
7819 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
7821 PR tree-optimization/78888
7822 * gimple-range-fold.cc (get_letter_range): New.
7823 (fold_using_range::range_of_builtin_call): Call get_letter_range.
7825 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
7827 PR tree-optimization/78888
7828 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
7829 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
7831 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
7832 Marc Glisse <marc.glisse@inria.fr>
7834 * match.pd (rotate): Simplify equality/inequality of rotations.
7835 (bswap): Simplify equality/inequality tests of byte swapping.
7837 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
7839 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
7842 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
7844 * range-op.cc (operator_lshift::fold_range): Pass rel to
7845 base class fold_range.
7846 (operator_rshift::fold_range): Same.
7848 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
7851 * toplev.h (min_align_loops_log): Remove declaration.
7852 (min_align_jumps_log, min_align_labels_log): Likewise.
7853 (min_align_functions_log): Likewise.
7855 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
7857 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
7859 (execute_vrp): Abstract out simplification of conditionals...
7860 (simplify_casted_conds): ...here.
7862 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
7864 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
7865 Add gimple argument.
7866 (array_bounds_checker::check_array_ref): Same.
7867 (array_bounds_checker::check_addr_expr): Same.
7868 (array_bounds_checker::check_array_bounds): Pass statement to
7869 check_array_bounds and check_addr_expr.
7870 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
7871 (check_addr_expr): Same.
7872 (get_value_range): Same.
7874 2021-07-26 Tamar Christina <tamar.christina@arm.com>
7876 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
7877 (sdot_prod, udot_prod): ... This.
7878 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
7880 (<sur>dot_prod<vsi2qi>): ... this.
7881 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
7882 Change operands order.
7883 (<sur>sadv16qi): Use new operands order.
7884 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
7885 vdotq_s32): Use new RTL ordering.
7887 2021-07-26 Tamar Christina <tamar.christina@arm.com>
7889 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
7890 aarch64_types_ternop_suss_qualifiers): New.
7891 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
7892 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
7893 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
7895 2021-07-23 Jakub Jelinek <jakub@redhat.com>
7897 PR rtl-optimization/101562
7898 * expmed.c (store_integral_bit_field): Only use movstrict_optab
7899 if the operand isn't paradoxical.
7901 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
7903 * gimple-array-bounds.h (class array_bounds_checker): Change
7904 ranges type to range_query.
7906 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
7908 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
7909 __builtin_memcpy instead of constructing
7910 __builtin_aarch64_simd_oi one vector at a time.
7911 (vst1_u64_x2): Likewise.
7912 (vst1_f64_x2): Likewise.
7913 (vst1_s8_x2): Likewise.
7914 (vst1_p8_x2): Likewise.
7915 (vst1_s16_x2): Likewise.
7916 (vst1_p16_x2): Likewise.
7917 (vst1_s32_x2): Likewise.
7918 (vst1_u8_x2): Likewise.
7919 (vst1_u16_x2): Likewise.
7920 (vst1_u32_x2): Likewise.
7921 (vst1_f16_x2): Likewise.
7922 (vst1_f32_x2): Likewise.
7923 (vst1_p64_x2): Likewise.
7924 (vst1q_s8_x2): Likewise.
7925 (vst1q_p8_x2): Likewise.
7926 (vst1q_s16_x2): Likewise.
7927 (vst1q_p16_x2): Likewise.
7928 (vst1q_s32_x2): Likewise.
7929 (vst1q_s64_x2): Likewise.
7930 (vst1q_u8_x2): Likewise.
7931 (vst1q_u16_x2): Likewise.
7932 (vst1q_u32_x2): Likewise.
7933 (vst1q_u64_x2): Likewise.
7934 (vst1q_f16_x2): Likewise.
7935 (vst1q_f32_x2): Likewise.
7936 (vst1q_f64_x2): Likewise.
7937 (vst1q_p64_x2): Likewise.
7939 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
7941 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
7942 __builtin_memcpy instead of constructing
7943 __builtin_aarch64_simd_ci one vector at a time.
7944 (vst1_u64_x3): Likewise.
7945 (vst1_f64_x3): Likewise.
7946 (vst1_s8_x3): Likewise.
7947 (vst1_p8_x3): Likewise.
7948 (vst1_s16_x3): Likewise.
7949 (vst1_p16_x3): Likewise.
7950 (vst1_s32_x3): Likewise.
7951 (vst1_u8_x3): Likewise.
7952 (vst1_u16_x3): Likewise.
7953 (vst1_u32_x3): Likewise.
7954 (vst1_f16_x3): Likewise.
7955 (vst1_f32_x3): Likewise.
7956 (vst1_p64_x3): Likewise.
7957 (vst1q_s8_x3): Likewise.
7958 (vst1q_p8_x3): Likewise.
7959 (vst1q_s16_x3): Likewise.
7960 (vst1q_p16_x3): Likewise.
7961 (vst1q_s32_x3): Likewise.
7962 (vst1q_s64_x3): Likewise.
7963 (vst1q_u8_x3): Likewise.
7964 (vst1q_u16_x3): Likewise.
7965 (vst1q_u32_x3): Likewise.
7966 (vst1q_u64_x3): Likewise.
7967 (vst1q_f16_x3): Likewise.
7968 (vst1q_f32_x3): Likewise.
7969 (vst1q_f64_x3): Likewise.
7970 (vst1q_p64_x3): Likewise.
7972 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
7975 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
7976 hard register when LRA is in progress.
7978 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
7980 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
7981 __builtin_memcpy instead of using a union.
7982 (vst1q_s8_x4): Likewise.
7983 (vst1_s16_x4): Likewise.
7984 (vst1q_s16_x4): Likewise.
7985 (vst1_s32_x4): Likewise.
7986 (vst1q_s32_x4): Likewise.
7987 (vst1_u8_x4): Likewise.
7988 (vst1q_u8_x4): Likewise.
7989 (vst1_u16_x4): Likewise.
7990 (vst1q_u16_x4): Likewise.
7991 (vst1_u32_x4): Likewise.
7992 (vst1q_u32_x4): Likewise.
7993 (vst1_f16_x4): Likewise.
7994 (vst1q_f16_x4): Likewise.
7995 (vst1_f32_x4): Likewise.
7996 (vst1q_f32_x4): Likewise.
7997 (vst1_p8_x4): Likewise.
7998 (vst1q_p8_x4): Likewise.
7999 (vst1_p16_x4): Likewise.
8000 (vst1q_p16_x4): Likewise.
8001 (vst1_s64_x4): Likewise.
8002 (vst1_u64_x4): Likewise.
8003 (vst1_p64_x4): Likewise.
8004 (vst1q_s64_x4): Likewise.
8005 (vst1q_u64_x4): Likewise.
8006 (vst1q_p64_x4): Likewise.
8007 (vst1_f64_x4): Likewise.
8008 (vst1q_f64_x4): Likewise.
8010 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8012 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8013 instead of constructing __builtin_aarch64_simd_oi one vector
8015 (vst2_u64): Likewise.
8016 (vst2_f64): Likewise.
8017 (vst2_s8): Likewise.
8018 (vst2_p8): Likewise.
8019 (vst2_s16): Likewise.
8020 (vst2_p16): Likewise.
8021 (vst2_s32): Likewise.
8022 (vst2_u8): Likewise.
8023 (vst2_u16): Likewise.
8024 (vst2_u32): Likewise.
8025 (vst2_f16): Likewise.
8026 (vst2_f32): Likewise.
8027 (vst2_p64): Likewise.
8028 (vst2q_s8): Likewise.
8029 (vst2q_p8): Likewise.
8030 (vst2q_s16): Likewise.
8031 (vst2q_p16): Likewise.
8032 (vst2q_s32): Likewise.
8033 (vst2q_s64): Likewise.
8034 (vst2q_u8): Likewise.
8035 (vst2q_u16): Likewise.
8036 (vst2q_u32): Likewise.
8037 (vst2q_u64): Likewise.
8038 (vst2q_f16): Likewise.
8039 (vst2q_f32): Likewise.
8040 (vst2q_f64): Likewise.
8041 (vst2q_p64): Likewise.
8043 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8045 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8046 instead of constructing __builtin_aarch64_simd_ci one vector
8048 (vst3_u64): Likewise.
8049 (vst3_f64): Likewise.
8050 (vst3_s8): Likewise.
8051 (vst3_p8): Likewise.
8052 (vst3_s16): Likewise.
8053 (vst3_p16): Likewise.
8054 (vst3_s32): Likewise.
8055 (vst3_u8): Likewise.
8056 (vst3_u16): Likewise.
8057 (vst3_u32): Likewise.
8058 (vst3_f16): Likewise.
8059 (vst3_f32): Likewise.
8060 (vst3_p64): Likewise.
8061 (vst3q_s8): Likewise.
8062 (vst3q_p8): Likewise.
8063 (vst3q_s16): Likewise.
8064 (vst3q_p16): Likewise.
8065 (vst3q_s32): Likewise.
8066 (vst3q_s64): Likewise.
8067 (vst3q_u8): Likewise.
8068 (vst3q_u16): Likewise.
8069 (vst3q_u32): Likewise.
8070 (vst3q_u64): Likewise.
8071 (vst3q_f16): Likewise.
8072 (vst3q_f32): Likewise.
8073 (vst3q_f64): Likewise.
8074 (vst3q_p64): Likewise.
8076 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8078 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8079 instead of constructing __builtin_aarch64_simd_xi one vector
8081 (vst4_u64): Likewise.
8082 (vst4_f64): Likewise.
8083 (vst4_s8): Likewise.
8084 (vst4_p8): Likewise.
8085 (vst4_s16): Likewise.
8086 (vst4_p16): Likewise.
8087 (vst4_s32): Likewise.
8088 (vst4_u8): Likewise.
8089 (vst4_u16): Likewise.
8090 (vst4_u32): Likewise.
8091 (vst4_f16): Likewise.
8092 (vst4_f32): Likewise.
8093 (vst4_p64): Likewise.
8094 (vst4q_s8): Likewise.
8095 (vst4q_p8): Likewise.
8096 (vst4q_s16): Likewise.
8097 (vst4q_p16): Likewise.
8098 (vst4q_s32): Likewise.
8099 (vst4q_s64): Likewise.
8100 (vst4q_u8): Likewise.
8101 (vst4q_u16): Likewise.
8102 (vst4q_u32): Likewise.
8103 (vst4q_u64): Likewise.
8104 (vst4q_f16): Likewise.
8105 (vst4q_f32): Likewise.
8106 (vst4q_f64): Likewise.
8107 (vst4q_p64): Likewise.
8109 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8111 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8112 instead of constructing __builtin_aarch64_simd_oi one vector
8114 (vtbx4_u8): Likewise.
8115 (vtbx4_p8): Likewise.
8117 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8119 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8120 instead of constructing __builtin_aarch64_simd_oi one vector
8122 (vtbl3_u8): Likewise.
8123 (vtbl3_p8): Likewise.
8124 (vtbl4_s8): Likewise.
8125 (vtbl4_u8): Likewise.
8126 (vtbl4_p8): Likewise.
8128 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8130 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8131 instead of constructing __builtin_aarch64_simd_oi one vector
8133 (vqtbx2_u8): Likewise.
8134 (vqtbx2_p8): Likewise.
8135 (vqtbx2q_s8): Likewise.
8136 (vqtbx2q_u8): Likewise.
8137 (vqtbx2q_p8): Likewise.
8138 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8139 __builtin_aarch64_simd_ci one vector at a time.
8140 (vqtbx3_u8): Likewise.
8141 (vqtbx3_p8): Likewise.
8142 (vqtbx3q_s8): Likewise.
8143 (vqtbx3q_u8): Likewise.
8144 (vqtbx3q_p8): Likewise.
8145 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8146 __builtin_aarch64_simd_xi one vector at a time.
8147 (vqtbx4_u8): Likewise.
8148 (vqtbx4_p8): Likewise.
8149 (vqtbx4q_s8): Likewise.
8150 (vqtbx4q_u8): Likewise.
8151 (vqtbx4q_p8): Likewise.
8153 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8155 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8156 instead of constructing __builtin_aarch64_simd_oi one vector
8158 (vqtbl2_u8): Likewise.
8159 (vqtbl2_p8): Likewise.
8160 (vqtbl2q_s8): Likewise.
8161 (vqtbl2q_u8): Likewise.
8162 (vqtbl2q_p8): Likewise.
8163 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8164 __builtin_aarch64_simd_ci one vector at a time.
8165 (vqtbl3_u8): Likewise.
8166 (vqtbl3_p8): Likewise.
8167 (vqtbl3q_s8): Likewise.
8168 (vqtbl3q_u8): Likewise.
8169 (vqtbl3q_p8): Likewise.
8170 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8171 __builtin_aarch64_simd_xi one vector at a time.
8172 (vqtbl4_u8): Likewise.
8173 (vqtbl4_p8): Likewise.
8174 (vqtbl4q_s8): Likewise.
8175 (vqtbl4q_u8): Likewise.
8176 (vqtbl4q_p8): Likewise.
8178 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8181 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8183 2021-07-22 Andrew Pinski <apinski@marvell.com>
8185 PR tree-optimization/10153
8186 * tree-tailcall.c (create_tailcall_accumulator):
8187 Don't call fold_convert as the type should be correct already.
8188 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8189 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8191 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8193 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8194 varying_p check for null/non-null check.
8196 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8198 PR tree-optimization/101511
8199 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8200 is in ssa2's equiv set, and don't trap if so.
8202 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8204 PR tree-optimization/101497
8205 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8208 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8210 PR tree-optimization/101496
8211 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8212 first, then vrp_visit_cond_Stmt.
8214 2021-07-22 liuhongt <hongtao.liu@intel.com>
8216 * config/i386/i386-expand.c
8217 (ix86_broadcast_from_integer_constant): Rename to ..
8218 (ix86_broadcast_from_constant): .. this, and extend it to
8220 (ix86_expand_vector_move): Extend to float mode.
8221 * config/i386/i386-features.c
8222 (replace_constant_pool_with_broadcast): Remove.
8223 (remove_partial_avx_dependency_gate): Ditto.
8224 (constant_pool_broadcast): Ditto.
8225 (class pass_constant_pool_broadcast): Ditto.
8226 (make_pass_constant_pool_broadcast): Ditto.
8227 (remove_partial_avx_dependency): Adjust gate.
8228 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8229 * config/i386/i386-protos.h
8230 (make_pass_constant_pool_broadcast): Remove.
8232 2021-07-22 liuhongt <hongtao.liu@intel.com>
8234 * config/i386/constraints.md (Wb): New constraint.
8236 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8238 (*ashlqi3_1): Ditto.
8239 (*<insn><mode>3_1): Split to ..
8240 (*ashr<mode>3_1): this, ...
8241 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8243 (*<insn><mode>3_1): Split to ..
8244 (*ashr<mode>3_1): this, ...
8245 (*lshrqi3_1): and this, also extend this pattern to avx512
8247 (*lshrhi3_1): And this, also extend this pattern to avx512
8249 * config/i386/sse.md (k<code><mode>): New define_split after
8250 it to convert generic shift pattern to mask shift ones.
8252 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8253 Joseph Myers <joseph@codesourcery.com>
8254 Cesar Philippidis <cesar@codesourcery.com>
8256 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8257 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8259 * tree-pretty-print.c (dump_omp_clause): Likewise.
8260 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8261 * gimplify.c (gimplify_scan_omp_clauses)
8262 (gimplify_adjust_omp_clauses): Likewise.
8263 * tree-nested.c (convert_nonlocal_omp_clauses)
8264 (convert_local_omp_clauses): Likewise.
8265 * omp-low.c (scan_sharing_clauses): Likewise.
8266 * omp-offload.c (execute_oacc_device_lower): Update.
8268 2021-07-21 Martin Sebor <msebor@redhat.com>
8270 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8272 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8274 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8277 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8279 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8281 (write_fntype_init): New stub function.
8282 (write_init_bif_table): Likewise.
8283 (write_init_ovld_table): New function.
8284 (write_init_file): Implement.
8286 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8288 * config/rs6000/rs6000-gen-builtins.c
8289 (write_autogenerated_header): New function.
8290 (write_decls): Likewise.
8291 (write_extern_fntype): New callback function.
8292 (write_header_file): Implement.
8294 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8296 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8299 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8301 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8303 (complete_base_type): Likewise.
8304 (construct_fntype_id): Likewise.
8305 (parse_bif_entry): Call contruct_fntype_id.
8306 (parse_ovld_entry): Likewise.
8308 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8310 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8311 (MAXOVLDSTANZAS): New macro.
8312 (ovld_stanzas): New variable.
8313 (curr_ovld_stanza): Likewise.
8314 (MAXOVLDS): New macro.
8315 (ovlddata): New struct.
8316 (ovlds): New variable.
8317 (curr_ovld): Likewise.
8318 (max_ovld_args): Likewise.
8319 (parse_ovld_entry): New function.
8320 (parse_ovld_stanza): Likewise.
8321 (parse_ovld): Implement.
8323 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8325 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8328 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8330 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8331 (parse_prototype): Implement.
8333 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8335 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
8336 (curr_bif_stanza): New variable.
8337 (stanza_entry): New struct.
8338 (stanza_map): New initialized variable.
8339 (enable_string): Likewise.
8340 (fnkinds): New enum.
8341 (typelist): New struct.
8342 (attrinfo): Likewise.
8343 (MAXRESTROPNDS): New macro.
8344 (prototype): New struct.
8345 (MAXBIFS): New macro.
8346 (bifdata): New struct.
8347 (bifs): New variable.
8348 (curr_bif): Likewise.
8349 (bif_order): Likewise.
8350 (bif_index): Likewise.
8351 (fatal): New function.
8352 (stanza_name_to_stanza): Likewise.
8353 (parse_bif_attrs): New stub function.
8354 (parse_prototype): Likewise.
8355 (parse_bif_entry): New function.
8356 (parse_bif_stanza): Likewise.
8357 (parse_bif): Implement.
8358 (set_bif_order): New function.
8359 (create_bif_order): Implement.
8361 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8363 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
8364 (num_bifs): New variable.
8365 (num_ovld_stanzas): Likewise.
8366 (num_ovlds): Likewise.
8367 (parse_codes): New enum.
8368 (bif_rbt): New variable.
8369 (ovld_rbt): Likewise.
8370 (fntype_rbt): Likewise.
8371 (bifo_rbt): Likewise.
8372 (parse_bif): New stub function.
8373 (create_bif_order): Likewise.
8374 (parse_ovld): Likewise.
8375 (write_header_file): Likewise.
8376 (write_init_file): Likewise.
8377 (write_defines_file): Likewise.
8378 (delete_output_files): New function.
8381 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
8384 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
8385 from CRC32 _builtin functions.
8387 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
8389 * coverage.c (build_gcov_info_var_registration): Mark the object placed
8390 in the linker set as referenced so that it does not get optimized away.
8392 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
8395 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8397 * config.gcc (riscv*-*-*): Detect which python is available.
8399 2021-07-21 Jakub Jelinek <jakub@redhat.com>
8401 PR middle-end/101535
8402 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
8403 contexts in which decl isn't privatized and for ORT_TARGET return
8404 false if decl is mapped.
8406 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8408 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
8409 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
8410 has been successfully unrolled.
8412 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8414 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
8415 exit condition if no tail loop is needed, and if the original exit
8416 condition should therefore be kept as-is.
8417 (tree_transform_and_unroll_loop): Handle that case here too.
8419 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
8421 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
8423 (free_data_refs): Likewise.
8424 * tree-data-ref.h (free_dependence_relations): Likewise.
8425 (free_data_refs): Likewise.
8426 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
8428 (struct component): Likewise.
8429 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
8430 (pcom_worker::~pcom_worker): Likewise.
8431 (pcom_worker::release_chain): Adjust as auto_vec changes.
8432 (pcom_worker::loop): Rename to ...
8433 (pcom_worker::m_loop): ... this.
8434 (pcom_worker::datarefs): Rename to ...
8435 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
8436 (pcom_worker::dependences): Rename to ...
8437 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
8438 (pcom_worker::chains): Rename to ...
8439 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
8440 (pcom_worker::looparound_phis): Rename to ...
8441 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
8443 (pcom_worker::cache): Rename to ...
8444 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
8445 (pcom_worker::release_chain): Adjust for auto_vec changes.
8446 (pcom_worker::release_chains): Adjust for auto_vec and renaming
8448 (release_component): Remove.
8449 (release_components): Adjust for release_component removal.
8450 (component_of): Adjust to use vec.
8451 (merge_comps): Likewise.
8452 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
8453 (pcom_worker::determine_offset): Likewise.
8454 (class comp_ptrs): Remove.
8455 (pcom_worker::split_data_refs_to_components): Adjust for renaming
8456 changes, for comp_ptrs removal with auto_vec.
8457 (pcom_worker::suitable_component_p): Adjust for renaming changes.
8458 (pcom_worker::filter_suitable_components): Adjust for release_component
8460 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
8461 (pcom_worker::find_looparound_phi): Likewise.
8462 (pcom_worker::add_looparound_copies): Likewise.
8463 (pcom_worker::determine_roots_comp): Likewise.
8464 (pcom_worker::single_nonlooparound_use): Likewise.
8465 (pcom_worker::execute_pred_commoning_chain): Likewise.
8466 (pcom_worker::execute_pred_commoning): Likewise.
8467 (pcom_worker::try_combine_chains): Likewise.
8468 (pcom_worker::prepare_initializers_chain): Likewise.
8469 (pcom_worker::prepare_initializers): Likewise.
8470 (pcom_worker::prepare_finalizers_chain): Likewise.
8471 (pcom_worker::prepare_finalizers): Likewise.
8472 (pcom_worker::tree_predictive_commoning_loop): Likewise.
8474 2021-07-20 Martin Sebor <msebor@redhat.com>
8476 PR middle-end/101397
8477 * builtins.c (gimple_call_return_array): Add argument. Correct
8478 offsets for memchr, mempcpy, stpcpy, and stpncpy.
8479 (compute_objsize_r): Adjust offset computation for argument returning
8482 2021-07-20 Martin Sebor <msebor@redhat.com>
8484 PR middle-end/101300
8485 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
8487 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
8489 * function.c (assign_parm_setup_block): Use adjust_address instead
8490 of change_address to preserve MEM_EXPR and friends.
8492 2021-07-20 Martin Sebor <msebor@redhat.com>
8494 * cfgloop.h (single_likely_exit): Adjust by-value argument to
8496 * cfgloopanal.c (single_likely_exit): Same.
8497 * cgraph.h (struct cgraph_node): Same.
8498 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
8499 * genautomata.c (merge_states): Same.
8500 * genextract.c (VEC_char_to_string): Same.
8501 * genmatch.c (dt_node::gen_kids_1): Same.
8502 (walk_captures): Adjust by-value argument to by-reference.
8503 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
8504 to by-const-reference.
8505 * gimple.c (gimple_build_call_vec): Same.
8506 (gimple_build_call_internal_vec): Same.
8507 (gimple_build_switch): Same.
8508 (sort_case_labels): Same.
8509 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8511 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
8513 (gimple_build_call_internal_vec): Same.
8514 (gimple_build_switch): Same.
8515 (sort_case_labels): Same.
8516 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8518 * haifa-sched.c (calc_priorities): Adjust by-value argument to
8520 (sched_init_luids): Same.
8521 (haifa_init_h_i_d): Same.
8522 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
8523 (adjust_callers_for_value_intersection): Adjust by-value argument to
8525 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
8527 (find_more_contexts_for_caller_subset): Same.
8528 (find_aggregate_values_for_callers_subset): Same.
8529 (copy_useful_known_contexts): Same.
8530 * ipa-fnsummary.c (remap_edge_summaries): Same.
8531 (remap_freqcounting_predicate): Same.
8532 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
8534 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
8535 to by-const-reference.
8536 * ipa-predicate.h (predicate::remap_after_inlining): Same.
8537 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
8538 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
8539 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
8540 * read-rtl.c (add_overload_instance): Same.
8541 * rtl.h (native_decode_rtx): Same.
8542 (native_decode_vector_rtx): Same.
8543 * sched-int.h (sched_init_luids): Same.
8544 (haifa_init_h_i_d): Same.
8545 * simplify-rtx.c (native_decode_vector_rtx): Same.
8546 (native_decode_rtx): Same.
8547 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
8548 (shrink_wrap_one_built_in_call_with_conds): Same.
8549 (shrink_wrap_conditional_dead_built_in_calls): Same.
8550 * tree-data-ref.c (create_runtime_alias_checks): Same.
8551 (compute_all_dependences): Same.
8552 * tree-data-ref.h (compute_all_dependences): Same.
8553 (create_runtime_alias_checks): Same.
8554 (index_in_loop_nest): Same.
8555 * tree-if-conv.c (mask_exists): Same.
8556 * tree-loop-distribution.c (class loop_distribution): Same.
8557 (loop_distribution::create_rdg_vertices): Same.
8558 (dump_rdg_partitions): Same.
8559 (debug_rdg_partitions): Same.
8560 (partition_contains_all_rw): Same.
8561 (loop_distribution::distribute_loop): Same.
8562 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
8563 (oacc_entry_exit_single_gang): Same.
8564 * tree-ssa-loop-im.c (hoist_memory_references): Same.
8565 (loop_suitable_for_sm): Same.
8566 * tree-ssa-loop-niter.c (bound_index): Same.
8567 * tree-ssa-reassoc.c (update_ops): Same.
8568 (swap_ops_for_binary_stmt): Same.
8569 (rewrite_expr_tree): Same.
8570 (rewrite_expr_tree_parallel): Same.
8571 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
8572 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
8573 * tree-ssa-structalias.c (process_all_all_constraints): Same.
8574 (make_constraints_to): Same.
8575 (handle_lhs_call): Same.
8576 (find_func_aliases_for_builtin_call): Same.
8577 (sort_fieldstack): Same.
8578 (check_for_overlaps): Same.
8579 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
8580 (vect_create_cond_for_unequal_addrs): Same.
8581 (vect_create_cond_for_lower_bounds): Same.
8582 (vect_create_cond_for_alias_checks): Same.
8583 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
8584 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
8585 (vect_make_slp_decision): Same.
8586 (vect_slp_bbs): Same.
8587 (duplicate_and_interleave): Same.
8588 (vect_transform_slp_perm_load): Same.
8589 (vect_schedule_slp): Same.
8590 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
8591 (vect_schedule_slp): Same.
8592 (duplicate_and_interleave): Same.
8593 * tree.c (build_vector_from_ctor): Same.
8594 (build_vector): Same.
8595 (check_vector_cst): Same.
8596 (check_vector_cst_duplicate): Same.
8597 (check_vector_cst_fill): Same.
8598 (check_vector_cst_stepped): Same.
8599 * tree.h (build_vector_from_ctor): Same.
8601 2021-07-20 Jakub Jelinek <jakub@redhat.com>
8604 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
8605 type from bool to int.
8606 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
8607 EASY_VECTOR_MSB case if either step or copies is not 1.
8608 (vspltis_shifted): Fix comment typo.
8609 (easy_altivec_constant): Change return type from bool to int, instead
8610 of returning true return byte size of the element mode that should be
8611 used to synthetize the constant.
8612 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
8613 that vspltis_shifted is 0, handle the case where easy_altivec_constant
8614 assumes using different vector mode from CONST_VECTOR's mode.
8615 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
8616 easy_altivec_constant to determine mode in which -1 >> -1 should be
8617 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
8619 2021-07-20 Richard Biener <rguenther@suse.de>
8622 * dwarf2out.h (dwarf_file_data): Add key member.
8623 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
8624 (dwarf_file_hasher::hash): Hash key.
8625 (lookup_filename): Remap the filename and store it in the
8626 filename member of dwarf_file_data when creating a new
8628 (file_name_acquire): Do not remap the filename again.
8629 (maybe_emit_file): Likewise.
8631 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
8633 * config/aarch64/aarch64-simd-builtins.def: Use two variant
8634 generators for all TBL/TBX intrinsics and rename to
8635 consistent forms: qtbl[1234] or qtbx[1234].
8636 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
8638 (aarch64_qtbl1<mode>): This.
8639 (aarch64_tbx1<mode>): Rename to...
8640 (aarch64_qtbx1<mode>): This.
8641 (aarch64_tbl2v16qi): Delete.
8642 (aarch64_tbl3<mode>): Rename to...
8643 (aarch64_qtbl2<mode>): This.
8644 (aarch64_tbx4<mode>): Rename to...
8645 (aarch64_qtbx2<mode>): This.
8646 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
8647 renamed qtbl1 and qtbl2 RTL patterns.
8648 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
8650 (vqtbl1_s8): Likewise.
8651 (vqtbl1_u8): Likewise.
8652 (vqtbl1q_p8): Likewise.
8653 (vqtbl1q_s8): Likewise.
8654 (vqtbl1q_u8): Likewise.
8655 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
8656 (vqtbx1_u8): Likewise.
8657 (vqtbx1_p8): Likewise.
8658 (vqtbx1q_s8): Likewise.
8659 (vqtbx1q_u8): Likewise.
8660 (vqtbx1q_p8): Likewise.
8661 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
8662 (vtbl1_u8): Likewise.
8663 (vtbl1_p8): Likewise.
8664 (vtbl2_s8): Likewise
8665 (vtbl2_u8): Likewise.
8666 (vtbl2_p8): Likewise.
8667 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
8668 (vtbl3_u8): Likewise.
8669 (vtbl3_p8): Likewise.
8670 (vtbl4_s8): Likewise.
8671 (vtbl4_u8): Likewise.
8672 (vtbl4_p8): Likewise.
8673 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
8674 (vtbx2_u8): Likewise.
8675 (vtbx2_p8): Likewise.
8676 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
8677 (vqtbl2_u8): Likewise.
8678 (vqtbl2_p8): Likewise.
8679 (vqtbl2q_s8): Likewise.
8680 (vqtbl2q_u8): Likewise.
8681 (vqtbl2q_p8): Likewise.
8682 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
8683 (vqtbx2_u8): Likewise.
8684 (vqtbx2_p8): Likewise.
8685 (vqtbx2q_s8): Likewise.
8686 (vqtbx2q_u8): Likewise.
8687 (vqtbx2q_p8): Likewise.
8688 (vtbx4_s8): Likewise.
8689 (vtbx4_u8): Likewise.
8690 (vtbx4_p8): Likewise.
8692 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
8695 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
8697 (define_peephole2 atomic_loaddi_fpu): Ditto.
8699 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8701 * config.gcc (riscv*-*-*): Detect which python is available.
8703 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
8705 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
8706 (smul<mode>3_highpart): ... this.
8707 (mulhu_<mode>): Rename to...
8708 (umul<mode>3_highpart): ... this.
8709 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
8710 MULHU_V2DI, MULHU_V4SI): Adjust.
8712 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
8714 PR tree-optimization/100696
8715 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
8716 * internal-fn.def (IFN_MULH): New internal function.
8717 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
8718 recog normal multiply highpart as IFN_MULH.
8719 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
8722 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
8724 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
8725 (BTF_DEBUGGING_INFO): Likewise.
8726 * doc/tm.texi.in: Document the new macros.
8727 * doc/tm.texi: Regenerated.
8728 * toplev.c: Guard initialization of debug hooks.
8730 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
8732 * flags.h (ctf_debuginfo_p): New function declaration.
8733 * opts.c (ctf_debuginfo_p): New function definition.
8735 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
8738 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
8739 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
8740 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
8741 * config/gcn/gcn-valu.md: Add a warning comment.
8742 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
8743 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
8744 * config/gcn/gcn.md: Add a warning comment.
8745 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
8746 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
8748 (EF_AMDGPU_XNACK): New.
8749 (EF_AMDGPU_SRAM_ECC): New.
8751 (copy_early_debug_info): Use elf_flags.
8752 (main): Handle -mxnack and -msram-ecc options.
8753 * doc/invoke.texi: Document -mxnack and -msram-ecc.
8755 2021-07-19 Andrew Pinski <apinski@marvell.com>
8758 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
8759 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
8761 2021-07-19 Richard Biener <rguenther@suse.de>
8763 PR tree-optimization/101505
8764 * tree-vect-patterns.c (vect_determine_precisions): Walk
8765 PHIs also for loop vectorization.
8767 2021-07-19 Richard Biener <rguenther@suse.de>
8769 * gimple.h (gimple_expr_type): Remove.
8770 * doc/gimple.texi: Remove gimple_expr_type documentation.
8772 2021-07-19 Richard Biener <rguenther@suse.de>
8774 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
8775 (ao_ref_init_from_vn_reference): Likewise.
8776 (fully_constant_reference): Likewise.
8777 (vn_reference_lookup_call): Do not set vr->type to random
8779 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
8781 * tree-vect-generic.c (expand_vector_piecewise): Pass in
8782 whether we expanded parallel.
8783 (expand_vector_parallel): Adjust.
8784 (expand_vector_addition): Likewise.
8785 (expand_vector_comparison): Likewise.
8786 (expand_vector_operation): Likewise.
8787 (expand_vector_scalar_condition): Likewise.
8788 (expand_vector_conversion): Likewise.
8790 2021-07-19 Richard Biener <rguenther@suse.de>
8792 * tree-vrp.c (register_edge_assert_for_2): Use the
8794 (vrp_folder::fold_predicate_in): Likewise.
8795 * vr-values.c (gimple_assign_nonzero_p): Likewise.
8796 (vr_values::extract_range_from_comparison): Likewise.
8797 (vr_values::extract_range_from_ubsan_builtin): Use the
8798 type of the first operand.
8799 (vr_values::extract_range_basic): Push down type
8800 computation, use the appropriate LHS.
8801 (vr_values::extract_range_from_assignment): Use the
8804 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
8807 * common/config/i386/i386-common.c (ix86_handle_option): For
8808 -mgeneral-regs-only, enable the GPR only instructions which are
8809 enabled implicitly by SSE ISAs unless they have been disabled
8812 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
8815 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
8816 ix86_avx_u128_mode_needed.
8817 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
8818 returns AVX register.
8820 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
8822 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
8823 (handle_const_call): Liekise
8824 (handle_pure_call): Liekise
8826 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
8828 PR tree-optimization/96542
8829 * range-op.cc (range_operator::wi_fold_in_parts): New.
8830 (range_operator::fold_range): Call wi_fold_in_parts.
8831 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
8832 * range-op.h (wi_fold_in_parts): Add prototype.
8834 2021-07-16 David Malcolm <dmalcolm@redhat.com>
8836 * doc/analyzer.texi: Add __analyzer_dump_state.
8838 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8840 * config/rs6000/rbtree.c: New file.
8841 * config/rs6000/rbtree.h: New file.
8843 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8845 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
8846 (typeinfo): Add restr field.
8847 (match_bracketed_pair): New function.
8848 (match_const_restriction): Implement.
8850 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8852 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
8854 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8856 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
8857 (basetype): Likewise.
8858 (typeinfo): Likewise.
8859 (handle_pointer): New function.
8860 (match_basetype): New stub function.
8861 (match_const_restriction): Likewise.
8862 (match_type): New function.
8864 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8866 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
8868 (advance_line): Likewise.
8869 (safe_inc_pos): Likewise.
8870 (match_identifier): Likewise.
8871 (match_integer): Likewise.
8872 (match_to_right_bracket): Likewise.
8874 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8876 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
8877 (ovld_file): Likewise.
8878 (header_file): Likewise.
8879 (init_file): Likewise.
8880 (defines_file): Likewise.
8881 (pgm_path): Likewise.
8882 (bif_path): Likewise.
8883 (ovld_path): Likewise.
8884 (header_path): Likewise.
8885 (init_path): Likewise.
8886 (defines_path): Likewise.
8887 (LINELEN): New macro.
8888 (linebuf): New variable.
8892 (bif_diag): New function.
8893 (ovld_diag): Likewise.
8895 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8897 * config/rs6000/rs6000-builtin-new.def: New.
8898 * config/rs6000/rs6000-overload.def: New.
8900 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8902 * config/rs6000/rs6000-gen-builtins.c: New.
8904 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8906 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
8907 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
8908 * gengtype-state.c (state_writer::write_state_file_list): Add a
8909 parameter to the fileslist expression for the number of build
8911 (read_state_files_list): Detect build headers and strip the
8912 initial "./" or ".\" from their names.
8913 * gengtype.c (build_headers): New global variable.
8914 (num_build_headers): Likewise.
8915 (open_base_files): Emit #include for each build header.
8916 (main): Detect and count build headers.
8917 * gengtype.h (build_headers): New extern variable.
8918 (num_build_headers): Likewise.
8920 2021-07-16 Richard Biener <rguenther@suse.de>
8922 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
8923 the type of the LHS.
8924 (find_bswap_or_nop_1): Likewise.
8925 (find_bswap_or_nop): Likewise.
8926 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
8928 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
8929 Remove unused parameters, pass in the scalar type. Fix
8930 internal store function handling.
8931 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
8932 (vect_get_vector_types_for_stmt): Move down check for
8933 existing vector stmt after we've determined a scalar type.
8934 Pass down the used scalar type to vect_get_smallest_scalar_type.
8935 * tree-vect-generic.c (expand_vector_condition): Use
8936 the type of the LHS.
8937 (expand_vector_scalar_condition): Likewise.
8938 (expand_vector_operations_1): Likewise.
8939 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
8940 (vect_recog_dot_prod_pattern): Likewise.
8941 (vect_recog_sad_pattern): Likewise.
8942 (vect_recog_widen_op_pattern): Likewise.
8943 (vect_recog_widen_sum_pattern): Likewise.
8944 (vect_recog_mixed_size_cond_pattern): Likewise.
8946 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
8948 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
8949 (dump_eaf_flags): Dump EAF_NOT_RETURNED
8950 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
8951 and EAF_NOT_RETURNED.
8952 (modref_summary::useful_p): Likewise.
8953 (modref_summary_lto::useful_p): Likewise.
8954 (struct) modref_summary_lto: Use eaf_fleags_t.
8955 (deref_flags): Handle EAF_NOT_RETURNED.
8956 (struct escape_point): Use min_flags.
8957 (modref_lattice::init): Add EAF_NOT_RETURNED.
8958 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
8959 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
8961 (analyze_parms): Also analyze const functions; update conition on
8963 (modref_write): Update streaming.
8964 (read_section): Update streaming.
8965 (remap_arg_flags): Use eaf_flags_t.
8966 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
8967 * ipa-modref.h: (eaf_flags_t): New typedef.
8968 (struct modref_summary): Use eaf_flags_t.
8969 * tree-core.h (EAF_NOT_RETURNED): New constant.
8971 2021-07-16 Richard Biener <rguenther@suse.de>
8973 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
8974 the type of the LHS.
8975 (gimple_assign_nonnegative_warnv_p): Likewise.
8976 (gimple_call_nonnegative_warnv_p): Likewise. Return false
8977 if the call has no LHS.
8978 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
8979 * tree-eh.c (stmt_could_throw_1_p): Likewise.
8980 * tree-inline.c (insert_init_stmt): Likewise.
8981 * tree-ssa-loop-niter.c (get_val_for): Likewise.
8982 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
8984 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
8985 gassign *. Use the type of the lhs.
8986 (vn_nary_op_lookup_stmt): Adjust.
8987 (vn_nary_op_insert_stmt): Likewise.
8989 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8991 * config/s390/predicates.md (bras_sym_operand): Accept all
8992 functions in 64-bit mode, use UNSPEC_PLT31.
8993 (larl_operand): Use UNSPEC_PLT31.
8994 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
8995 (legitimize_pic_address): Likewise.
8996 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
8998 (s390_delegitimize_address): Use UNSPEC_PLT31.
8999 (s390_output_addr_const_extra): Likewise.
9000 (print_operand): Add @PLT to TLS calls, handle %K.
9001 (s390_function_profiler): Mark __fentry__/_mcount as function,
9002 use %K, use UNSPEC_PLT31.
9003 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9004 (s390_emit_call): Use UNSPEC_PLT31.
9005 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9006 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9007 (*movdi_64): Use %K.
9008 (reload_base_64): Likewise.
9009 (*sibcall_brc): Likewise.
9010 (*sibcall_brcl): Likewise.
9011 (*sibcall_value_brc): Likewise.
9012 (*sibcall_value_brcl): Likewise.
9015 (*bras_r): Likewise.
9016 (*brasl_r): Likewise.
9017 (*bras_tls): Likewise.
9018 (*brasl_tls): Likewise.
9019 (main_base_64): Likewise.
9020 (reload_base_64): Likewise.
9021 (@split_stack_call<mode>): Likewise.
9023 2021-07-16 Richard Biener <rguenther@suse.de>
9025 PR tree-optimization/101467
9026 * tree-vect-stmts.c (vect_gen_while): Properly guard
9027 make_temp_ssa_name usage.
9029 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9031 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9032 t-sysroot-suffix for C-SKY.
9033 * config/csky/print-sysroot-suffix.sh: Delete.
9034 * config/csky/t-csky-linux: Delete.
9035 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9036 instead of CSKY_MULTILIB_DIRNAMES.
9038 2021-07-16 Richard Biener <rguenther@suse.de>
9040 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9041 conversion issues with the partial reduction of the reused
9044 2021-07-16 Richard Biener <rguenther@suse.de>
9046 * config/i386/i386-options.c (ix86_option_override_internal): Set
9047 param_vect_partial_vector_usage to zero if not set.
9049 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9052 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9053 (VALID_INT_MODE_P): Add SDmode and DDmode.
9054 Add TDmode for TARGET_64BIT.
9055 (VALID_DFP_MODE_P): Remove.
9056 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9057 Do not use VALID_DFP_MODE_P.
9059 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9061 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9063 (fold_using_range::fold_stmt): Ditto.
9064 (fold_using_range::range_of_range_op): Ditto.
9065 (fold_using_range::range_of_phi): Ditto.
9066 (fold_using_range::range_of_call): Ditto.
9067 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9068 (fold_using_range::range_of_builtin_call): Ditto.
9069 (fold_using_range::range_of_cond_expr): Ditto.
9070 * gimple-range-fold.h (gimple_range_type): New.
9072 2021-07-15 Martin Sebor <msebor@redhat.com>
9075 * tree-ssa-strlen.c (handle_assign): New function.
9076 (maybe_warn_overflow): Add argument.
9077 (nonzero_bytes_for_type): New function.
9078 (count_nonzero_bytes): Handle more tree types. Call
9079 nonzero_bytes_for_tye.
9080 (count_nonzero_bytes): Handle types.
9081 (handle_store): Handle stores from function calls.
9082 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9083 it for assignments from function calls.
9085 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9090 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9092 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9094 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9096 2021-07-15 Martin Sebor <msebor@redhat.com>
9100 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9101 (operand_compare::verify_hash_value): Same.
9102 * tree-core.h (OEP_DECL_NAME): New.
9104 2021-07-15 Martin Jambor <mjambor@suse.cz>
9106 * profile-count.h (profile_count::value): Change the return type to
9108 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9110 * tree-cfg.c (dump_function_to_file): Likewise.
9112 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9115 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9116 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9119 2021-07-15 Richard Biener <rguenther@suse.de>
9121 * tree-vectorizer.h (vect_gen_while): Match up with
9123 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9124 of vect_gen_while_not.
9125 (vect_gen_while_not): Adjust.
9126 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9128 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9130 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9131 (ranger_cache::range_of_def): Call adjust_range.
9132 (ranger_cache::entry_range): Same.
9133 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9134 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9136 (gimple_ranger::range_on_entry): Same.
9138 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9141 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9143 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9145 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9148 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9150 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9151 (sdot_prod, udot_prod): ...These.
9152 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9153 (aarch64_<sur>dot<vsi2qi>): Rename to...
9154 (<sur>dot_prod<vsi2qi>): ...This.
9155 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9158 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9160 PR middle-end/101437
9161 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9162 types even if they have non-BLKmode TYPE_MODE.
9164 2021-07-15 Richard Biener <rguenther@suse.de>
9167 * gcc.c (process_command): Process -gtoggle like process_options
9168 would after parsing options.
9170 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9172 * cfgexpand.c (expand_asm_loc): Adjust.
9173 (expand_asm_stmt): Likewise.
9174 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9175 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9176 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9177 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9178 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9179 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9180 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9181 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9182 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9183 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9184 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9185 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9186 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9187 * doc/tm.texi: Regenerate.
9188 * target.def: Add location argument to md_asm_adjust.
9190 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9192 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9193 diagnostic's location, not input_location.
9195 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9197 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9199 (expand_asm_stmt): Likewise.
9201 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9203 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9204 addressed memory rtx, if any.
9205 (rs6000_split_multireg_move): Fix code formatting.
9206 Handle MMA build built-ins with operands in adjacent memory locations.
9208 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9210 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9213 2021-07-14 Jason Merrill <jason@redhat.com>
9215 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9217 2021-07-14 Jason Merrill <jason@redhat.com>
9219 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9222 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9224 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9225 Remove erroneous line.
9227 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9229 * params.opt (param_evrp_mode): Change default.
9231 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9233 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9234 (sdot_prod, udot_prod): ...These.
9235 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9236 (aarch64_<sur>dot<vsi2qi>): Rename to...
9237 (<sur>dot_prod<vsi2qi>): ...This.
9238 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9241 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9243 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9245 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9247 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9249 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9251 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9253 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9255 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9256 (usdot_prod<vsi2qi>): ... This.
9257 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9258 (usdot_prod): ...This.
9259 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9260 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9262 (@<sur>dot_prod<vsi2qi>): ...This.
9263 * config/aarch64/aarch64-sve-builtins-base.cc
9264 (svusdot_impl::expand): Use it.
9266 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9268 * optabs.def (usdot_prod_optab): New.
9269 * doc/md.texi: Document it and clarify other dot prod optabs.
9270 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9271 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9272 * optabs.c (expand_widen_pattern_expr): Likewise.
9273 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9274 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9275 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9277 (vect_widened_op_tree): Optionally ignore
9279 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9281 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9284 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9285 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9286 Enable UINTR only for 64-bit codegen.
9287 * config/i386/i386-options.c
9288 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9290 * config/i386/i386.h (ARCH_ARG): New.
9291 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9292 "[arch|tune] 64" for 64-bit codegen.
9294 2021-07-14 Richard Biener <rguenther@suse.de>
9296 PR tree-optimization/101445
9297 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9298 of the IV in the correct direction for negative stride
9301 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9304 * godump.c (godump_str_hash): New type.
9305 (godump_container::pot_dummy_types): Use string_hash instead of
9306 ptr_hash in the hash_set.
9308 2021-07-14 Richard Biener <rguenther@suse.de>
9310 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9311 vector types where the old vector type has a multiple of
9312 the new vector type elements.
9313 (vect_create_partial_epilog): New function, split out from...
9314 (vect_create_epilog_for_reduction): ... here.
9315 (vect_transform_cycle_phi): Reduce the re-used accumulator
9316 to the new vector type.
9318 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9320 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9321 non-'t'-sized arg check.
9323 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9325 * tree-cfg.c (cleanup_dead_labels_eh): Update
9326 post_landing_pad label upon change of landing pad block's
9328 (cleanup_dead_labels): Check that a removed label is not that
9331 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
9333 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
9335 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
9336 Add Neon to general purpose register case for zero-extend
9338 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
9339 case to prevent some cases opting to go through memory.
9340 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
9341 * rtl.c (rtvec_series_p): Define predicate to determine
9342 whether a vector contains a linear series of integers.
9343 * rtl.h (rtvec_series_p): Define.
9344 * rtlanal.c (vec_series_lowpart_p): Define predicate to
9345 determine if a vector selection is equivalent to the low part
9347 * rtlanal.h (vec_series_lowpart_p): Define.
9348 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
9349 Add vec_select -> subreg simplification.
9351 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
9353 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
9354 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
9355 _mm_test_mix_ones_zeros): New.
9357 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
9358 Richard Biener <rguenther@suse.de>
9360 * gimple.c (gimple_could_trap_p_1): Make S argument a
9361 "const gimple*". Preserve constness in call to
9362 gimple_asm_volatile_p.
9363 (gimple_could_trap_p): Make S argument a "const gimple*".
9364 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
9365 Update function prototypes.
9367 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9369 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
9370 (_loop_vec_info::main_loop_edge): New field.
9371 (_loop_vec_info::skip_main_loop_edge): Likewise.
9372 (_loop_vec_info::skip_this_loop_edge): Likewise.
9373 (_loop_vec_info::reusable_accumulators): Likewise.
9374 (_stmt_vec_info::reduc_scalar_results): Likewise.
9375 (_stmt_vec_info::reused_accumulator): Likewise.
9376 (vect_get_main_loop_result): Declare.
9377 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
9378 reduc_scalar_inputs.
9379 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
9380 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
9381 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
9382 skip_main_loop_edge and skip_this_loop_edge fields.
9383 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
9384 (vect_emit_reduction_init_stmts): New function.
9385 (get_initial_def_for_reduction): Use it.
9386 (get_initial_defs_for_reduction): Likewise. Change the vinfo
9387 parameter to a loop_vec_info.
9388 (vect_create_epilog_for_reduction): Store the scalar results
9389 in the reduc_info. If an epilogue loop is reusing an accumulator
9390 from the main loop, and if the epilogue loop can also be skipped,
9391 try to place the reduction code in the join block. Record
9392 accumulators that could potentially be reused by epilogue loops.
9393 (vect_transform_cycle_phi): When vectorizing epilogue loops,
9394 try to reuse accumulators from the main loop. Record the initial
9395 value in reduc_info for non-SLP reductions too.
9397 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9399 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
9400 adjustment handling. Take the neutral value as an argument,
9401 in place of the code argument.
9402 (vect_transform_cycle_phi): Update accordingly. Handle the
9403 initial values of cond reductions separately from code reductions.
9404 Choose the adjustment here rather than in
9405 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
9407 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9409 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
9410 (neutral_op_for_reduction): ...this, providing a more general
9412 (vect_create_epilog_for_reduction): Update accordingly.
9413 (vectorizable_reduction): Likewise.
9414 (vect_transform_cycle_phi): Likewise.
9416 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9418 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
9419 reduc_info instead of the original stmt_vec_info.
9420 (vect_transform_cycle_phi): Update accordingly.
9422 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9424 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
9425 reduc_info as an additional parameter.
9426 (vect_transform_cycle_phi): Update accordingly.
9428 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9430 * tree-vectorizer.h: Include tree-ssa-operands.h.
9431 (vect_phi_initial_value): New function.
9432 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
9433 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
9434 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9435 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
9437 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9439 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
9440 the phi results to vectype after creating them. Remove later
9441 conversion code that thus becomes redundant.
9443 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9445 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
9446 the new_phis vector with a reduc_inputs vector. Combine handling
9447 of reduction chains and ncopies > 1.
9449 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9451 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
9452 scalar_results to group_size elements after reducing down from
9453 N*group_size elements. Construct an array_slice of the live-out
9454 stmts and assert that there is one stmt per scalar result.
9456 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9458 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
9459 nested_in_vect_loop and use double_reduc everywhere. Remove dead
9460 assignment to "loop".
9462 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9464 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
9465 vector types first. For scalar types, consider both the preferred
9466 vector mode and the alternative vector modes.
9467 * optabs-query.c (can_vec_mask_load_store_p): Use the same
9468 structure as above, in particular using related_vector_mode
9469 for modes provided by autovectorize_vector_modes.
9471 2021-07-13 Jakub Jelinek <jakub@redhat.com>
9472 Richard Biener <rguenther@suse.de>
9474 PR tree-optimization/101419
9475 * tree-pass.h (PROP_objsz): Define.
9476 (make_pass_early_object_sizes): Declare.
9477 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
9478 there to pass_early_object_sizes, drop parameter.
9479 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
9480 drop parameter, move pass_post_ipa_warn right after that.
9481 * tree-object-size.c (pass_object_sizes::execute): Rename to...
9482 (object_sizes_execute): ... this. Add insert_min_max_p argument.
9483 (pass_data_object_sizes): Move after object_sizes_execute.
9484 (pass_object_sizes): Likewise. In execute method call
9485 object_sizes_execute, drop set_pass_param method and insert_min_max_p
9486 non-static data member and its initializer in the ctor.
9487 (pass_data_early_object_sizes, pass_early_object_sizes,
9488 make_pass_early_object_sizes): New.
9489 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
9490 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
9492 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
9495 * config/riscv/constraints.md ("S"): Update description and remove
9497 * doc/md.texi (Machine Constraints): Document the 'S' constraints
9500 2021-07-13 Richard Biener <rguenther@suse.de>
9503 2021-07-12 Richard Biener <rguenther@suse.de>
9505 * tree-vect-slp.c (vect_slp_region): Show the number of
9506 SLP graph entries in the optimization message.
9508 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
9510 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
9512 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
9513 return type to long.
9514 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
9517 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9519 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
9520 Query relation between the 2 operands and use it.
9522 2021-07-12 Sergei Trofimovich <siarheit@google.com>
9524 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
9526 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9529 * config/i386/predicates.md (vec_setm_sse41_operand):
9530 Rename from vec_setm_operand.
9531 (vec_setm_avx2_operand): New predicate.
9532 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
9533 Use vec_setm_sse41_operand as operand 2 predicate.
9534 (vec_set<V_256_512:mode): New expander.
9535 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
9536 as operand 2 predicate.
9538 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9540 PR tree-optimization/101335
9541 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
9543 2021-07-12 Andrew Pinski <apinski@marvell.com>
9545 * tree-ssa-phiopt.c (match_simplify_replacement): Move
9546 insert of the sequence before the movement of the
9547 statement. Check if to see if the statement is used
9548 outside of the original phi to see if we should move it.
9550 2021-07-12 Richard Biener <rguenther@suse.de>
9552 * dump-context.h (debug_dump_context::debug_dump_context):
9553 Add FILE * parameter defaulted to stderr.
9554 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
9555 * tree-vect-slp.c (dot_slp_tree): New functions.
9557 2021-07-12 Richard Biener <rguenther@suse.de>
9559 PR tree-optimization/101373
9560 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
9561 references when the BB may not return.
9562 (compute_avail): Pass in the function we're working on and
9563 replace cfun references with it. Externally throwing
9564 const calls also possibly terminate the function.
9565 (pass_pre::execute): Pass down the function we're working on.
9566 * gcse.c (compute_hash_table_work): Externally throwing
9567 const/pure calls also need record_last_mem_set_info.
9568 * postreload-gcse.c (record_opr_changes): Looping or externally
9569 throwing const/pure calls also need record_last_mem_set_info.
9571 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9573 * recog.c (memory_address_addr_space_p): Change the type to bool.
9574 Return true/false instead of 1/0.
9575 (offsettable_memref_p): Ditto.
9576 (offsettable_nonstrict_memref_p): Ditto.
9577 (offsettable_address_addr_space_p): Ditto.
9578 Change the type of addressp indirect function to bool.
9579 * recog.h (memory_address_addr_space_p): Change the type to bool.
9580 (strict_memory_address_addr_space_p): Ditto.
9581 (offsettable_memref_p): Ditto.
9582 (offsettable_nonstrict_memref_p): Ditto.
9583 (offsettable_address_addr_space_p): Ditto.
9584 * reload.c (maybe_memory_address_addr_space_p): Ditto.
9585 (strict_memory_address_addr_space_p): Change the type to bool.
9586 Return true/false instead of 1/0.
9587 (maybe_memory_address_addr_space_p): Change the type to bool.
9589 2021-07-12 Richard Biener <rguenther@suse.de>
9591 * tree-vect-slp.c (vect_slp_region): Show the number of
9592 SLP graph entries in the optimization message.
9594 2021-07-12 Richard Biener <rguenther@suse.de>
9596 PR tree-optimization/101394
9597 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
9598 copies from abnormals for a full redundancy.
9600 2021-07-12 Richard Biener <rguenther@suse.de>
9602 PR middle-end/101423
9603 * gimple.c (gimple_could_trap_p_1): Internal function calls
9605 * tree-eh.c (tree_could_trap_p): Likewise.
9607 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9610 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
9612 (vmulq_n_u32): Likewise.
9613 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
9614 (vmulq_n_f32): Likewise.
9615 (vmul_n_f16): Likewise.
9616 (vmulq_n_f16): Likewise.
9618 2021-07-12 Martin Liska <mliska@suse.cz>
9621 * gcc.c (check_offload_target_name): Call
9622 candidates_list_and_hint only if we have a candidate.
9624 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9627 * config/arm/neon.md (vec_init): Move to ...
9628 * config/arm/vec-common.md (vec_init): ... here.
9629 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
9631 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
9633 PR tree-optimization/101403
9634 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
9635 signedness of the shift is not the same as the signedness of
9638 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
9639 Uroš Bizjak <ubizjak@gmail.com>
9641 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
9642 divmod of a constant numerator with new define_insn_and_split.
9644 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
9647 * config/i386/i386-expand.c (ix86_expand_call): If a call is
9648 to a non-local-binding, or local but to a public symbol, then
9649 assume that it might be indirected via the lazy symbol binder.
9650 Mark R10 and R10 as clobbered in that case.
9652 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
9655 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
9656 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
9657 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
9659 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
9661 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
9662 log2 range of operands[3] to [1,31].
9663 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
9665 2021-07-09 Sergei Trofimovich <siarheit@google.com>
9667 * doc/md.texi: Don't split @smallexample in multiple @groups.
9669 2021-07-09 Sergei Trofimovich <siarheit@google.com>
9671 * doc/md.texi: Add missing 'see' word.
9673 2021-07-09 Andrew Pinski <apinski@marvell.com>
9675 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
9676 to take sequence and gimple_match_op. Accept the case where
9677 op is a SSA_NAME and one statement in the sequence.
9678 Also allow constants.
9679 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
9680 Update call to phiopt_early_allow. Discard the sequence if not
9683 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
9688 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
9689 instead of ISA_HAS_MSA.
9690 (mips_expand_vec_unpack): Likewise.
9691 (mips_expand_vector_init): Likewise.
9693 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
9695 * config/rs6000/vsx.md (mods_<mode>): Rename to...
9696 (mod<mode>3): ... this.
9697 (modu_<mode>): Rename to...
9698 (umod<mode>3): ... this.
9699 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
9702 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
9704 * config/h8300/shiftrotate.md (variable shifts): Expose condition
9705 code handling for the test before the loop.
9707 2021-07-08 Martin Jambor <mjambor@suse.cz>
9710 * ipa-sra.c (class isra_call_summary): New member
9711 m_before_any_store, initialize it in the constructor.
9712 (isra_call_summary::dump): Dump the new field.
9713 (ipa_sra_call_summaries::duplicate): Copy it.
9714 (process_scan_results): Set it.
9715 (isra_write_edge_summary): Stream it.
9716 (isra_read_edge_summary): Likewise.
9717 (param_splitting_across_edge): Only override
9718 safe_to_import_accesses if m_before_any_store is set.
9720 2021-07-08 Martin Sebor <msebor@redhat.com>
9723 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
9724 Use Object Size Type 0 instead of 1.
9726 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
9728 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
9731 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
9733 * match.pd: Simplify an extend-operate-truncate sequence involving
9736 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
9737 Richard Biener <rguenther@suse.de>
9739 PR tree-optimization/40210
9740 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
9741 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
9742 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
9744 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
9747 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
9749 * config/i386/mmx.md (V_32): New mode iterator.
9750 (mov<V_32:mode>): Use V_32 mode iterator.
9751 (*mov<V_32:mode>_internal): Ditto.
9752 (*push<V_32:mode>2_rex64): Ditto.
9753 (*push<V_32:mode>2): Ditto.
9754 (movmisalign<V_32:mode>): Ditto.
9755 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
9756 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
9757 (vec_unpacks_lo_v4qi): New expander.
9758 (vec_unpacks_hi_v4qi): Ditto.
9759 (vec_unpacku_lo_v4qi): Ditto.
9760 (vec_unpacku_hi_v4qi): Ditto.
9761 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
9762 (VALID_INT_MODE_P): Ditto.
9764 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
9767 * config/rs6000/rs6000.md (udivti3): New insn.
9769 (umodti3): New insn.
9772 2021-07-07 Martin Sebor <msebor@redhat.com>
9774 PR tree-optimization/100137
9775 PR tree-optimization/99121
9776 PR tree-optimization/97027
9777 * builtins.c (access_ref::access_ref): Also set offmax.
9778 (access_ref::offset_in_range): Define new function.
9779 (access_ref::add_offset): Set offmax.
9780 (access_ref::inform_access): Handle access_none.
9781 (handle_mem_ref): Clear ostype.
9782 (compute_objsize_r): Handle ASSERT_EXPR.
9783 * builtins.h (struct access_ref): Add offmax member.
9784 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
9785 compute_objsize() and simplify.
9787 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
9789 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
9790 and VSX_BUILTIN_STXVP.
9792 2021-07-07 Martin Sebor <msebor@redhat.com>
9795 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
9796 a stray %K from error_at() missed in r12-2088.
9798 2021-07-07 Richard Biener <rguenther@suse.de>
9800 PR tree-optimization/99728
9801 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
9803 (mem_refs_may_alias_p): Add assert we handled aggregate
9805 (sm_seq_valid_bb): Give up when running into aggregate copies.
9806 (ref_indep_loop_p): Handle aggregate copies as never
9807 being invariant themselves but allow other refs to be
9808 disambiguated against them.
9809 (can_sm_ref_p): Do not try to apply store-motion to aggregate
9812 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
9815 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
9818 2021-07-06 Martin Sebor <msebor@redhat.com>
9820 * gimple-pretty-print.c (percent_G_format): Remove.
9821 * tree-diagnostic.c (default_tree_printer): Remove calls.
9822 * tree-pretty-print.c (percent_K_format): Remove.
9823 * tree-pretty-print.h (percent_K_format): Remove.
9825 2021-07-06 Martin Sebor <msebor@redhat.com>
9827 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
9828 Remove %K and use error_at.
9829 (aarch64_expand_fcmla_builtin): Same.
9830 (aarch64_expand_builtin_tme): Same.
9831 (aarch64_expand_builtin_memtag): Same.
9832 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
9833 (arm_expand_builtin): Same.
9834 * config/arm/arm.c (bounds_check): Same.
9836 2021-07-06 Martin Sebor <msebor@redhat.com>
9838 * builtins.c (warn_string_no_nul): Remove %G.
9839 (maybe_warn_for_bound): Same.
9840 (warn_for_access): Same.
9841 (check_access): Same.
9842 (check_strncat_sizes): Same.
9843 (expand_builtin_strncat): Same.
9844 (expand_builtin_strncmp): Same.
9845 (expand_builtin): Same.
9846 (expand_builtin_object_size): Same.
9847 (warn_dealloc_offset): Same.
9848 (maybe_emit_free_warning): Same.
9849 * calls.c (maybe_warn_alloc_args_overflow): Same.
9850 (maybe_warn_nonstring_arg): Same.
9851 (maybe_warn_rdwr_sizes): Same.
9852 * expr.c (expand_expr_real_1): Remove %K.
9853 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
9854 (gimple_fold_builtin_strncat): Same.
9855 * gimple-ssa-sprintf.c (format_directive): Same.
9856 (handle_printf_call): Same.
9857 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
9858 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
9859 (maybe_diag_access_bounds): Same. Call gimple_location.
9860 (check_bounds_or_overlap): Same.
9861 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
9862 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
9863 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
9864 (maybe_diag_stxncpy_trunc): Same.
9865 (handle_builtin_stxncpy_strncat): Same.
9866 (maybe_warn_pointless_strcmp): Same.
9867 * tree-ssa-uninit.c (maybe_warn_operand): Same.
9869 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
9872 * config/i386/predicates.md (vec_setm_operand): Enable
9873 register_operand for TARGET_SSE4_1.
9874 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
9875 as operand 2 predicate. Call ix86_expand_vector_set_var
9876 for non-constant index operand.
9877 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
9878 Call ix86_expand_vector_set_var for non-constant index operand.
9880 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
9882 * config/h8300/jumpcall.md (*branch): When possible, generate
9883 the comparison in CCZN mode.
9884 * config/h8300/predicates.md (simple_memory_operand): Reject all
9885 auto-increment addressing modes.
9887 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
9890 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
9893 2021-07-06 Richard Biener <rguenther@suse.de>
9895 * doc/md.texi (vec_fmaddsub<mode>4): Document.
9896 (vec_fmsubadd<mode>4): Likewise.
9897 * optabs.def (vec_fmaddsub$a4): Add.
9898 (vec_fmsubadd$a4): Likewise.
9899 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
9900 (IFN_VEC_FMSUBADD): Likewise.
9901 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
9902 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
9903 (addsub_pattern::build): Likewise.
9904 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
9905 and CFN_VEC_FMSUBADD are not transparent for permutes.
9906 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
9907 (vec_fmsubadd<mode>4): Likewise.
9909 2021-07-06 Richard Biener <rguenther@suse.de>
9911 * doc/invoke.texi (fmove-loop-stores): Document.
9912 * common.opt (fmove-loop-stores): New option.
9913 * opts.c (default_options_table): Enable -fmove-loop-stores
9915 * tree-ssa-loop-im.c (pass_lim::execute): Pass
9916 flag_move_loop_stores instead of true to
9917 loop_invariant_motion_in_fun.
9919 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
9921 * doc/install.texi: Document --with-dsymutil.
9923 2021-07-06 Andrew Pinski <apinski@marvell.com>
9925 PR tree-optimization/101256
9926 * dbgcnt.def (phiopt_edge_range): New counter.
9927 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
9928 Check to make sure the new name is defined in the same
9929 bb as the conditional before duplicating range info.
9930 Also add debug counter.
9932 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
9934 PR rtl-optimization/100328
9935 * config/i386/i386-options.c (ix86_option_override_internal):
9936 Set param_ira_consider_dup_in_all_alts to 0.
9938 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
9940 PR rtl-optimization/100328
9941 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
9943 * ira.c (ira_get_dup_out_num): Adjust as parameter
9944 param_ira_consider_dup_in_all_alts.
9945 * params.opt (ira-consider-dup-in-all-alts): New.
9946 * ira-conflicts.c (process_regs_for_copy): Add one parameter
9947 single_input_op_has_cstr_p.
9948 (get_freq_for_shuffle_copy): New function.
9949 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
9950 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
9952 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
9954 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
9955 generate condition code aware RTL directly.
9957 2021-07-05 Andrew Pinski <apinski@marvell.com>
9959 PR tree-optimization/101039
9960 * match.pd (A CMP 0 ? A : -A): New patterns.
9961 * tree-ssa-phiopt.c (abs_replacement): Delete function.
9962 (tree_ssa_phiopt_worker): Don't call abs_replacement.
9963 Update comment about abs_replacement.
9965 2021-07-05 Andrew Pinski <apinski@marvell.com>
9967 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
9968 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
9970 2021-07-05 Andrew Pinski <apinski@marvell.com>
9972 * tree-ssa-phiopt.c (match_simplify_replacement):
9973 Add early_p argument. Call gimple_simplify_phiopt
9974 instead of gimple_simplify.
9975 (tree_ssa_phiopt_worker): Update call to
9976 match_simplify_replacement and allow unconditionally.
9977 (phiopt_early_allow): New function.
9978 (gimple_simplify_phiopt): New function.
9980 2021-07-05 Andrew Pinski <apinski@marvell.com>
9982 PR middle-end/101237
9983 * fold-const.c (negate_expr_p): Remove call to element_mode
9984 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
9985 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
9986 (fold_negate_expr_1): Likewise.
9987 (const_unop): Likewise.
9988 (fold_cond_expr_with_comparison): Likewise.
9989 (fold_binary_loc): Likewise.
9990 (fold_ternary_loc): Likewise.
9991 (tree_call_nonnegative_warnv_p): Likewise.
9992 * match.pd (-(A + B) -> (-B) - A): Likewise.
9994 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
9996 * configure.ac: Handle --with-dsymutil in the same way as we
9997 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
9998 Extract the type and version for the dsymutil configured or
9999 found by the default searches.
10000 * config.in: Regenerated.
10001 * configure: Regenerated.
10002 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10003 same way as for the assembler and linker.
10004 * config/darwin.h (DSYMUTIL): Delete.
10005 * gcc.c: Report a configured dsymutil correctly.
10006 * exec-tool.in: Allow for dsymutil.
10008 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10010 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10011 Handle V4QI and V2HI modes.
10012 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10013 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10014 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10015 Handle 4-byte dual operands with TARGET_XOP and single operands
10016 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10017 mmx_pshufbv4qi3 for TARGET_SSSE3.
10018 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10019 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10020 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10021 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10022 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10023 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10024 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10025 (mmx_pshufbv4qi3): Ditto.
10026 (*mmx_pblendw32): Ditto.
10027 (*mmx_pblendw64): Rename from *mmx_pblendw.
10028 (mmx_punpckhbw_low): New insn_and_split pattern.
10029 (mmx_punpcklbw_low): Ditto.
10031 2021-07-05 Richard Biener <rguenther@suse.de>
10033 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10034 set LOOP_C_INFINITE on the vectorized loop.
10036 2021-07-05 Richard Biener <rguenther@suse.de>
10038 PR middle-end/101291
10039 * cfgloopmanip.c (loop_version): Set the loop copy of the
10040 versioned loop to the new loop.
10042 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10045 * config.gcc: Ensure that Darwin biarch definitions are
10046 added before i386.h.
10047 * config/i386/darwin.h (TARGET_64BIT): Remove.
10048 (PR80556_WORKAROUND): New.
10049 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10050 (DARWIN_SUBARCH_SPEC): New.
10051 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10052 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10053 (REAL_LIBGCC_SPEC): Remove.
10054 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10055 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10056 (REAL_LIBGCC_SPEC): Remove.
10058 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10060 PR middle-end/101294
10061 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10063 2021-07-02 Martin Sebor <msebor@redhat.com>
10065 PR middle-end/98871
10066 PR middle-end/98512
10067 * diagnostic.c (get_any_inlining_info): New.
10068 (update_effective_level_from_pragmas): Handle inlining context.
10069 (diagnostic_enabled): Same.
10070 (diagnostic_report_diagnostic): Same.
10071 * diagnostic.h (struct diagnostic_info): Add ctor.
10072 (struct diagnostic_context): Add new member.
10073 * tree-diagnostic.c (set_inlining_locations): New.
10074 (tree_diagnostics_defaults): Set new callback pointer.
10076 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10078 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10080 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10081 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10082 lxvp and stxvp built-ins.
10083 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10084 (builtin_function_type): Likewise.
10085 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10087 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10089 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10090 additional argument for the code.
10091 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10092 argument for the code. Just return if the ZN bits are useful or
10093 not rather than the old style CC_* enums.
10094 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10095 more generic shiftqi patterns.
10096 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10097 (shiftqi_noscratch_set_flags): New pattern.
10098 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10100 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10102 PR tree-optimization/101223
10103 * range-op.cc (build_lt): Add -1 for signed values.
10104 (built_gt): Subtract -1 for signed values.
10106 2021-07-02 David Faust <david.faust@oracle.com>
10108 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10109 (btf_asm_type): Likewise.
10111 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10113 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10114 an array of rtx rather than a pointer to rtx. Add code argument.
10115 (compute_a_shift_length): Similarly.
10116 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10117 shift itself isn't an operand. Create dummy operand[0] to carry
10118 a mode and pass a suitable rtx code to compute_a_shift_length.
10119 (get_shift_alg): Adjust operand number of clobber in output templates.
10120 (output_a_shift): Make first argument an array of rtx rather than
10121 a pointer to rtx. Add code argument for the type of shift.
10122 Adjust now that the shift itself is no longer an operand.
10123 (compute_a_shift_length): Similarly.
10124 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10125 iterator rather than nshift_operator.
10126 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10127 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10128 and compute_a_shift_length.
10129 (shiftqi_noscratch_clobber_flags): Likewise.
10130 (shifthi_noscratch_clobber_flags): Likewise.
10131 (shiftsi_noscratch_clobber_flags): Likewise.
10133 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10136 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10137 dsymutil for BTF/CTF.
10139 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10142 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10143 segment to include BTF.
10144 (BTF_INFO_SECTION_NAME): New.
10146 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10148 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10149 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10150 * config/m32r/m32r.c (call_operand): Adjust return type.
10151 (small_data_operand, memreg_operand): Likewise.
10153 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10155 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10157 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10158 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10159 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10160 (move_source_operand, move_destination_operand): Likewise.
10161 (condexec_source_operand, condexec_dest_operand): Likewise.
10162 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10163 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10164 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10165 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10166 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10167 (even_fpr_operand, odd_fpr_operand): Likewise.
10168 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10169 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10170 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10171 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10172 (relational_operator, float_relational_operator): Likewise.
10173 (ccr_eqne_operator, minmax_operator): Likewise.
10174 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10175 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10176 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10177 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10178 (quad_acc_operand, accg_operand): Likewise.
10180 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10182 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10183 return type to a bool.
10184 (nonimmediate_nonstack_operand): Likewise.
10185 (xstormy16_splittable_below100_operand): Likewise.
10186 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10188 (xstormy16_splittable_below100_operand): Likewise.
10190 2021-07-02 Richard Biener <rguenther@suse.de>
10192 PR tree-optimization/101293
10193 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10194 with combined offsets.
10195 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10196 combined with the rest of the offset.
10198 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10200 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10201 PIC encodings for PE-COFF targets.
10203 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10206 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10207 Return nullptr for TImode inner mode.
10209 2021-07-02 Richard Biener <rguenther@suse.de>
10211 PR tree-optimization/101280
10212 PR tree-optimization/101173
10213 * gimple-loop-interchange.cc
10214 (tree_loop_interchange::valid_data_dependences): Properly
10215 guard all dependence checks with DDR_REVERSED_P or its
10218 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10220 * config/i386/i386-expand.c (ix86_expand_builtin):
10221 Add branch to clear odata when ZF is set for asedecenc_expand
10222 and wideaesdecenc_expand.
10224 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10226 * config/i386/gcc-auto-profile: regenerate
10228 2021-07-02 liuhongt <hongtao.liu@intel.com>
10230 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10231 (trunc<mode><pmov_dst_4_lower>2): this.
10233 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10235 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10236 * input.c (class fcache): Rename to...
10237 (class file_cache_slot): ...this, making most members private and
10238 prefixing fields with "m_".
10239 (file_cache_slot::get_file_path): New accessor.
10240 (file_cache_slot::get_use_count): New accessor.
10241 (file_cache_slot::missing_trailing_newline_p): New accessor.
10242 (file_cache_slot::inc_use_count): New.
10243 (fcache_buffer_size): Move to...
10244 (file_cache_slot::buffer_size): ...here.
10245 (fcache_line_record_size): Move to...
10246 (file_cache_slot::line_record_size): ...here.
10247 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10248 (fcache_tab_size): Move to file_cache::num_file_slots.
10249 (diagnostic_file_cache_init): Update for move of fcache_tab
10250 to global_dc->m_file_cache.
10251 (diagnostic_file_cache_fini): Likewise.
10252 (lookup_file_in_cache_tab): Convert to...
10253 (file_cache::lookup_file): ...this.
10254 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10255 fcache_tab to global_dc->m_file_cache, moving most of
10256 implementation to...
10257 (file_cache::forcibly_evict_file): ...this new function and...
10258 (file_cache_slot::evict): ...this new function.
10259 (evicted_cache_tab_entry): Convert to...
10260 (file_cache::evicted_cache_tab_entry): ...this.
10261 (add_file_to_cache_tab): Convert to...
10262 (file_cache::add_file): ...this, moving bulk of implementation
10264 (file_cache_slot::create): ..this new function.
10265 (file_cache::file_cache): New.
10266 (file_cache::~file_cache): New.
10267 (lookup_or_add_file_to_cache_tab): Convert to...
10268 (file_cache::lookup_or_add_file): ..this new function.
10269 (fcache::fcache): Rename to...
10270 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10272 (fcache::~fcache): Rename to...
10273 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10275 (needs_read): Convert to...
10276 (file_cache_slot::needs_read_p): ...this.
10277 (needs_grow): Convert to...
10278 (file_cache_slot::needs_grow_p): ...this.
10279 (maybe_grow): Convert to...
10280 (file_cache_slot::maybe_grow): ...this.
10281 (read_data): Convert to...
10282 (file_cache_slot::read_data): ...this.
10283 (maybe_read_data): Convert to...
10284 (file_cache_slot::maybe_read_data): ...this.
10285 (get_next_line): Convert to...
10286 (file_cache_slot::get_next_line): ...this.
10287 (goto_next_line): Convert to...
10288 (file_cache_slot::goto_next_line): ...this.
10289 (read_line_num): Convert to...
10290 (file_cache_slot::read_line_num): ...this.
10291 (location_get_source_line): Update for moving of globals to
10292 global_dc->m_file_cache.
10293 (location_missing_trailing_newline): Likewise.
10294 * input.h (class file_cache_slot): New forward decl.
10295 (class file_cache): New.
10297 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10299 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10300 128-bit floating point conditional move support.
10301 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10303 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10304 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10305 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10306 (fpmask<mode>, IEEE128 iterator): New insn.
10307 (xxsel<mode>, IEEE128 iterator): New insn.
10309 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10312 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10314 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10316 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10318 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10320 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10321 (vec_duplicate<mode>): New expander.
10323 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10326 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10328 (ix86_byte_broadcast): New function.
10329 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10330 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10331 size is 16 bytes or bigger.
10332 (ix86_broadcast_from_integer_constant): New function.
10333 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
10334 to broadcast if mode size is 16 bytes or bigger.
10335 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
10337 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
10339 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10341 * config/i386/predicates.md (ix86_endbr_immediate_operand):
10342 Return true/false instead of 1/0.
10343 (movq_parallel): Ditto.
10345 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10347 * recog.c (general_operand): Return true/false instead of 1/0.
10348 (register_operand): Ditto.
10349 (immediate_operand): Ditto.
10350 (const_int_operand): Ditto.
10351 (const_scalar_int_operand): Ditto.
10352 (const_double_operand): Ditto.
10353 (push_operand): Ditto.
10354 (pop_operand): Ditto.
10355 (memory_operand): Ditto.
10356 (indirect_operand): Ditto.
10358 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10360 * genpreds.c (write_predicate_subfunction):
10361 Change the type of written subfunction to bool.
10362 (write_one_predicate_function):
10363 Change the type of written function to bool.
10364 (write_tm_preds_h): Ditto.
10365 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
10366 * recog.c (general_operand): Change the type to bool.
10367 (address_operand): Ditto.
10368 (register_operand): Ditto.
10369 (pmode_register_operand): Ditto.
10370 (scratch_operand): Ditto.
10371 (immediate_operand): Ditto.
10372 (const_int_operand): Ditto.
10373 (const_scalar_int_operand): Ditto.
10374 (const_double_operand): Ditto.
10375 (nonimmediate_operand): Ditto.
10376 (nonmemory_operand): Ditto.
10377 (push_operand): Ditto.
10378 (pop_operand): Ditto.
10379 (memory_operand): Ditto.
10380 (indirect_operand): Ditto.
10381 (ordered_comparison_operator): Ditto.
10382 (comparison_operator): Ditto.
10383 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
10384 Change the type of indirect predicate function to bool.
10385 * config/rs6000/rs6000.c (easy_vector_constant):
10386 Change the type to bool.
10387 * config/mips/mips-protos.h (m16_based_address_p):
10388 Change the type of operand 3 to bool.
10390 2021-07-01 Richard Biener <rguenther@suse.de>
10392 PR tree-optimization/101280
10393 PR tree-optimization/101173
10394 * gimple-loop-interchange.cc
10395 (tree_loop_interchange::valid_data_dependences): Revert
10396 previous change and instead correctly handle DDR_REVERSED_P
10399 2021-07-01 Richard Biener <rguenther@suse.de>
10401 PR tree-optimization/101278
10402 * tree-ssa-dse.c (dse_classify_store): First check for
10403 uses, then ignore stmt for chaining purposes.
10405 2021-07-01 Richard Biener <rguenther@suse.de>
10407 PR tree-optimization/100778
10408 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
10409 vectorized ops ahead of their scalar BB.
10411 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10414 * config/i386/i386.md (*nabs<dwi>2_doubleword):
10415 New insn_and_split pattern.
10416 (*nabs<dwi>2_1): Ditto.
10417 * config/i386/i386-features.c
10418 (general_scalar_chain::compute_convert_gain):
10419 Handle (NEG (ABS (...))) RTX. Rewrite src code
10420 scanner as switch statement.
10421 (general_scalar_chain::convert_insn):
10422 Handle (NEG (ABS (...))) RTX.
10423 (general_scalar_to_vector_candidate_p):
10424 Detect (NEG (ABS (...))) RTX. Reorder case statements
10425 for (AND (NOT (...) ...)) fallthrough.
10427 2021-07-01 Richard Biener <rguenther@suse.de>
10429 PR tree-optimization/101178
10430 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
10431 (slpg::perm_in): Add.
10432 (slpg::get_perm_in): Remove.
10433 (slpg::get_perm_materialized): Add.
10434 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
10435 during permute propagation and materialization.
10437 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10440 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
10442 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10444 PR middle-end/94366
10445 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
10446 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
10447 use boolean_type_node instead of integer_type_node as NE_EXPR type.
10448 (lower_reduction_clauses): Likewise.
10450 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10452 * config/gcn/gcn.c: Include dwarf2.h.
10453 (gcn_addr_space_debug): New function.
10454 (TARGET_ADDR_SPACE_DEBUG): New hook.
10456 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10458 * common/config/gcn/gcn-common.c
10459 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
10460 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
10462 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
10463 (gcn_frame_pointer_rqd): New function.
10464 (TARGET_FRAME_POINTER_REQUIRED): New hook.
10466 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10468 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
10469 prologue register saves.
10470 (gcn_debug_unwind_info): Use UI_DWARF2.
10471 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
10472 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
10473 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
10474 (DWARF_LINK_REGISTER): New define.
10475 (FIRST_PSEUDO_REGISTER): Increment.
10476 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
10477 (CALL_USED_REGISTERS): Likewise.
10478 (REGISTER_NAMES): Likewise.
10480 2021-06-30 Richard Biener <rguenther@suse.de>
10482 PR tree-optimization/101267
10483 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
10484 API and use SLP compatible interface of vect_is_simple_use.
10485 Reject not vectorized SLP defs for callers that do not support
10487 (vect_check_store_rhs): Handle masked stores and pass down
10488 the appropriate operator index.
10489 (vectorizable_call): Adjust.
10490 (vectorizable_store): Likewise.
10491 (vectorizable_load): Likewise. Handle SLP pecularity of
10493 (vect_is_simple_use): Remove special-casing of masked stores.
10495 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10497 * common.opt (foffload): Remove help as Driver only.
10498 * gcc.c (display_help): Add -foffload.
10500 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10502 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
10503 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
10504 a candidate list; better inform no offload target is configured
10505 and fix hint extraction when passed target is not '\0' at [len].
10506 * common.opt (foffload): Add tailing '.'.
10507 (foffload-options): Likewise; fix flag name in the help string.
10509 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10512 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
10513 (vcage_f32): Gate comparison on __FAST_MATH__.
10514 (vcageq_f32): Likewise.
10515 (vcale_f32): Likewise.
10516 (vcaleq_f32): Likewise.
10517 (vcagt_f32): Likewise.
10518 (vcagtq_f32): Likewise.
10519 (vcalt_f32): Likewise.
10520 (vcaltq_f32): Likewise.
10521 (vcage_f16): Likewise.
10522 (vcageq_f16): Likewise.
10523 (vcale_f16): Likewise.
10524 (vcaleq_f16): Likewise.
10525 (vcagt_f16): Likewise.
10526 (vcagtq_f16): Likewise.
10527 (vcalt_f16): Likewise.
10528 (vcaltq_f16): Likewise.
10530 2021-06-30 Richard Biener <rguenther@suse.de>
10532 PR tree-optimization/101264
10533 * tree-vect-slp.c (vect_optimize_slp): Propagate the
10534 computed perm_in to all "any" permute successors
10535 we cannot de-duplicate immediately.
10537 2021-06-30 liuhongt <hongtao.liu@intel.com>
10540 * config/i386/sse.md
10541 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
10543 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
10545 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
10546 * config/i386/subst.md (maskz_scalar): New define_subst.
10547 (maskz_scalar_name): New subst_attr.
10548 (maskz_scalar_op5): Ditto.
10549 (round_saeonly_maskz_scalar_op5): Ditto.
10550 (round_saeonly_maskz_scalar_operand5): Ditto.
10552 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
10554 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
10555 Increase code CSECT alignment to at least 32 bytes.
10556 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
10557 alignment designation.
10559 2021-06-29 Sergei Trofimovich <siarheit@google.com>
10561 * doc/generic.texi: Fix s/net yet/not yet/ typo.
10563 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10565 PR tree-optimization/101254
10566 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
10567 wrapping/non-wrapping when setting the result range.
10569 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10571 * value-query.cc (gimple_range_global): Allow phis.
10573 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10575 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
10576 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
10577 statement for location context.
10578 (check_for_binary_op_overflow): Ditto.
10579 (simplify_using_ranges::get_vr_for_comparison): Ditto.
10580 (simplify_using_ranges::compare_name_with_value): Ditto.
10581 (simplify_using_ranges::compare_names): Ditto.
10582 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
10583 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
10584 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
10585 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
10586 (simplify_using_ranges::two_valued_val_range_p): Ditto.
10587 (simplify_using_ranges::simplify): Ditto.
10588 * vr-values.h: Adjust prototypes.
10590 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
10593 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
10595 2021-06-29 Julian Brown <julian@codesourcery.com>
10597 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
10598 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
10599 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
10601 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
10602 (MAX_FIXED_MODE_SIZE): Change to 128.
10604 2021-06-29 Julian Brown <julian@codesourcery.com>
10606 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
10607 (s_mnemonic): Add clrsb.
10608 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
10609 (clrsb<mode>2): Add expander for SImode/DImode.
10611 2021-06-29 Julian Brown <julian@codesourcery.com>
10613 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
10614 muldi3): Add patterns.
10616 2021-06-29 Julian Brown <julian@codesourcery.com>
10618 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
10619 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
10621 2021-06-29 Julian Brown <julian@codesourcery.com>
10623 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
10625 2021-06-29 Joseph Myers <joseph@codesourcery.com>
10627 * btfout.c, ctfout.c: Include "memmodel.h".
10629 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10631 * gcc.c (check_offload_target_name): Cast len argument to
10632 %q.*s to 'int'; avoid -Wstringop-truncation warning.
10634 2021-06-29 Richard Biener <rguenther@suse.de>
10636 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
10637 to "any" permute nodes and relax "any" permute proapgation
10638 during iterative backward propagation.
10640 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10643 * common.opt (-foffload=): Update description.
10644 (-foffload-options=): New.
10645 * doc/invoke.texi (C Language Options): Document
10646 -foffload and -foffload-options.
10647 * gcc.c (check_offload_target_name): New, split off from
10648 handle_foffload_option.
10649 (check_foffload_target_names): New.
10650 (handle_foffload_option): Handle -foffload=default.
10651 (driver_handle_option): Update for -foffload-options.
10652 * lto-opts.c (lto_write_options): Use -foffload-options
10653 instead of -foffload.
10654 * lto-wrapper.c (merge_and_complain, append_offload_options):
10656 * opts.c (common_handle_option): Likewise.
10658 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10660 * doc/invoke.texi (C Language Options): Sort options
10661 alphabetically in optlist and also the description itself.
10662 Remove leftover -fallow-single-precision from and add missing
10663 -fgnu-tm to the optlist.
10665 2021-06-29 Richard Biener <rguenther@suse.de>
10667 * tree-vect-slp.c (slpg_vertex::visited): Remove.
10668 (vect_slp_perms_eq): Handle -1 permutes.
10669 (vect_optimize_slp): Rewrite permute propagation.
10671 2021-06-29 Jakub Jelinek <jakub@redhat.com>
10674 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
10675 perform the optimization in GENERIC when sanitizing and x has a
10678 2021-06-29 Richard Biener <rguenther@suse.de>
10680 PR tree-optimization/101242
10681 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
10682 PHIs with not represented initial values as leafs.
10684 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
10686 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
10687 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
10688 "register" keyword.
10689 (pdp11_initial_elimination_offset) Remove unused variable.
10690 (pdp11_cmp_length) Ditto.
10691 (pdp11_insn_cost): Ditto, and fix signedness warning.
10693 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
10695 * btfout.c: Include tm_p.h.
10698 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10700 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
10702 (bpf_expand_epilogue): Likewise.
10703 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
10704 Do not define DBX_DEBUGGING_INFO.
10706 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10708 * doc/invoke.texi: Document the CTF and BTF debug info options.
10710 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10711 David Faust <david.faust@oracle.com>
10712 Jose E. Marchesi <jose.marchesi@oracle.com>
10713 Weimin Pan <weimin.pan@oracle.com>
10715 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
10716 GTFILES. Add new object files.
10717 * common.opt: Add CTF and BTF debug info options.
10718 * btfout.c: New file.
10719 * ctfc.c: Likewise.
10720 * ctfc.h: Likewise.
10721 * ctfout.c: Likewise.
10722 * dwarf2ctf.c: Likewise.
10723 * dwarf2ctf.h: Likewise.
10724 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
10726 * dwarf2out.c (dwarf2out_source_line): Likewise.
10727 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
10729 (debug_format_do_cu): New function.
10730 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
10732 Include dwarf2ctf.c.
10733 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
10735 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
10736 (CTF_DEBUG): New bitmask.
10737 (BTF_DEBUG): Likewise.
10738 (enum ctf_debug_info_levels): New enum.
10739 * gengtype.c (open_base_files): Handle ctfc.h.
10740 (main): Handle uint32_t type.
10741 * flags.h (btf_debuginfo_p): New definition.
10742 (dwarf_based_debuginfo_p): Likewise.
10743 * opts.c (debug_type_names): Add entries for CTF and BTF.
10744 (btf_debuginfo_p): New function.
10745 (dwarf_based_debuginfo_p): Likewise.
10746 (common_handle_option): Handle -gctfN and -gbtf options.
10747 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
10748 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
10751 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
10753 * dwarf2out.c (AT_class): Function is no longer static.
10754 (AT_int): Likewise.
10755 (AT_unsigned): Likewise.
10756 (AT_loc): Likewise.
10757 (get_AT): Likewise.
10758 (get_AT_string): Likewise.
10759 (get_AT_flag): Likewise.
10760 (get_AT_unsigned): Likewise.
10761 (get_AT_ref): Likewise.
10762 (new_die_raw): Likewise.
10763 (lookup_decl_die): Likewise.
10764 (base_type_die): Likewise.
10765 (add_name_attribute): Likewise.
10766 (add_AT_int): Likewise.
10767 (add_AT_unsigned): Likewise.
10768 (add_AT_loc): Likewise.
10769 (dw_get_die_tag): New function.
10770 (dw_get_die_child): Likewise.
10771 (dw_get_die_sib): Likewise.
10772 (struct dwarf_file_data): Move from here to dwarf2out.h
10773 (struct dw_attr_struct): Likewise.
10774 * dwarf2out.h: Analogous changes.
10776 2021-06-28 Martin Jambor <mjambor@suse.cz>
10779 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
10780 members m_dead_stmts and m_dead_ssas.
10781 * ipa-param-manipulation.c
10782 (ipa_param_body_adjustments::mark_dead_statements): New function.
10783 (ipa_param_body_adjustments::common_initialization): Call it on
10784 all removed but not split parameters.
10785 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
10787 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
10789 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
10790 dead debug statements.
10791 (copy_phis_for_bb): Do not copy dead PHI nodes.
10793 2021-06-28 Martin Jambor <mjambor@suse.cz>
10796 * symtab-clones.h (clone_info): Removed member param_adjustments.
10797 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
10798 deal with pass-through splits now.
10799 (ipa_param_performed_split): Removed.
10800 (ipa_param_adjustments::modify_call): Adjusted parameters.
10801 (class ipa_param_body_adjustments): Adjusted parameters of
10802 register_replacement, modify_gimple_stmt and modify_call_stmt.
10803 (ipa_verify_edge_has_no_modifications): Declare.
10804 (ipa_edge_modifications_finalize): Declare.
10805 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
10806 performed_splits processing, pas only edge to padjs->modify_call,
10807 check that call arguments were not modified if they should not have
10809 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
10811 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
10812 (ipa_edge_modification_info): Likewise.
10813 (ipa_edge_modification_sum): Likewise.
10814 (ipa_edge_modifications): New edge summary.
10815 (ipa_verify_edge_has_no_modifications): New function.
10816 (transitive_split_p): Removed.
10817 (transitive_split_map): Likewise.
10818 (init_transitive_splits): Likewise.
10819 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
10820 summary instead of performed_splits.
10821 (ipa_param_body_adjustments::register_replacement): Drop dummy
10822 parameter, set base_index of the created ipa_param_body_replacement.
10823 (phi_arg_will_live_p): New function.
10824 (ipa_param_body_adjustments::common_initialization): Do not create
10825 IPA_SRA dummy decls.
10826 (simple_tree_swap_info): Removed.
10827 (remap_split_decl_to_dummy): Likewise.
10828 (record_argument_state_1): New function.
10829 (record_argument_state): Likewise.
10830 (ipa_param_body_adjustments::modify_call_stmt): New parameter
10831 orig_stmt. Do not work with dummy decls, save necessary info about
10832 changes to ipa_edge_modifications.
10833 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
10834 orig_stmt, pass it to modify_call_stmt.
10835 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
10836 modify_gimple_stmt.
10837 (ipa_edge_modifications_finalize): New function.
10838 * tree-inline.c (remap_gimple_stmt): Pass original statement to
10839 modify_gimple_stmt.
10840 (copy_phis_for_bb): Do not copy dead PHI nodes.
10841 (expand_call_inline): Do not remap performed_splits.
10842 (update_clone_info): Likewise.
10843 * toplev.c: Include ipa-param-manipulation.h.
10844 (toplev::finalize): Call ipa_edge_modifications_finalize.
10846 2021-06-28 Andrew Pinski <apinski@marvell.com>
10848 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
10849 info if we're the only things setting the target PHI.
10850 (value_replacement): Don't duplicate range here.
10851 (minmax_replacement): Likewise.
10853 2021-06-28 Richard Biener <rguenther@suse.de>
10855 PR tree-optimization/101229
10856 * gimple-walk.c (gimple_walk_op): Handle PHIs.
10858 2021-06-28 Martin Liska <mliska@suse.cz>
10860 * config/v850/v850.c (construct_dispose_instruction): Allocate
10862 (construct_prepare_instruction): Likewise.
10864 2021-06-28 Martin Liska <mliska@suse.cz>
10866 * config/v850/v850.c (v850_option_override): Build default
10868 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
10869 ignored for inlining.
10870 (TARGET_CAN_INLINE_P): New.
10872 2021-06-28 Richard Biener <rguenther@suse.de>
10874 PR tree-optimization/101207
10875 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
10876 permute eliding for load permutations properly.
10878 2021-06-28 Richard Biener <rguenther@suse.de>
10880 PR tree-optimization/101173
10881 * gimple-loop-interchange.cc
10882 (tree_loop_interchange::valid_data_dependences): Disallow outer
10883 loop dependence distance of zero.
10885 2021-06-28 liuhongt <hongtao.liu@intel.com>
10888 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
10889 define_insn_and_split.
10890 (*avx_cmp<mode>3_ltint): Ditto.
10891 (*avx2_pcmp<mode>3_3): Ditto.
10892 (*avx2_pcmp<mode>3_4): Ditto.
10893 (*avx2_pcmp<mode>3_5): Ditto.
10895 2021-06-28 liuhongt <hongtao.liu@intel.com>
10897 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
10898 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
10899 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
10900 IX86_BUILTIN_PBLENDVB128): Replace icode with
10902 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
10904 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
10905 New pre_reload splitter.
10907 2021-06-27 Andrew Pinski <apinski@marvell.com>
10909 PR middle-end/101230
10910 * fold-const.c (fold_ternary_loc): Check
10911 the return value of invert_tree_comparison.
10913 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
10915 * config.gcc: Add SPDX License Identifier.
10916 (powerpc-ibm-aix789): Default to aix73.h.
10917 (powerpc-ibm-aix7.2.*.*): New stanza.
10918 * config/rs6000/aix72.h: Add SPDX License Identifier.
10919 * config/rs6000/aix73.h: New file.
10921 2021-06-26 Jason Merrill <jason@redhat.com>
10923 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
10925 2021-06-26 Andrew Pinski <apinski@marvell.com>
10927 * genmatch.c (lower_cond): Copy for_subst_vec
10928 for the simplify also.
10929 (lower): Swap the order for lower_for and lower_cond.
10931 2021-06-26 Andrew Pinski <apinski@marvell.com>
10933 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
10934 flow senatitive info on the moved ssa set.
10936 2021-06-26 Andrew Pinski <apinski@marvell.com>
10938 * fold-const.c (fold_cond_expr_with_comparison):
10939 Exand arg0 into comp_code, arg00, and arg01.
10940 (fold_ternary_loc): Use invert_tree_comparison
10941 instead of fold_invert_truthvalue for the case
10942 where we have A CMP B ? C : A.
10944 2021-06-25 Martin Sebor <msebor@redhat.com>
10946 PR middle-end/101216
10947 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
10949 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
10951 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
10953 2021-06-25 Richard Biener <rguenther@suse.de>
10955 PR tree-optimization/101202
10956 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
10959 2021-06-25 Richard Biener <rguenther@suse.de>
10961 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
10962 STMT_VINFO_REDUC_DEF from the original representative.
10964 2021-06-25 Martin Sebor <msebor@redhat.com>
10966 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
10967 gimple_no_warning_p and gimple_set_no_warning with
10968 warning_suppressed_p, and suppress_warning.
10970 (maybe_warn_for_bound): Same.
10971 (warn_for_access): Same.
10972 (check_access): Same.
10973 (expand_builtin_strncmp): Same.
10974 (fold_builtin_varargs): Same.
10975 * calls.c (maybe_warn_nonstring_arg): Same.
10976 (maybe_warn_rdwr_sizes): Same.
10977 * cfgexpand.c (expand_call_stmt): Same.
10978 * cgraphunit.c (check_global_declaration): Same.
10979 * fold-const.c (fold_undefer_overflow_warnings): Same.
10980 (fold_truth_not_expr): Same.
10981 (fold_unary_loc): Same.
10982 (fold_checksum_tree): Same.
10983 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
10984 (array_bounds_checker::check_mem_ref): Same.
10985 (array_bounds_checker::check_addr_expr): Same.
10986 (array_bounds_checker::check_array_bounds): Same.
10987 * gimple-expr.c (copy_var_decl): Same.
10988 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
10989 (gimple_fold_builtin_strncat): Same.
10990 (gimple_fold_builtin_stxcpy_chk): Same.
10991 (gimple_fold_builtin_stpcpy): Same.
10992 (gimple_fold_builtin_sprintf): Same.
10993 (fold_stmt_1): Same.
10994 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
10995 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
10996 * gimple-ssa-sprintf.c (handle_printf_call): Same.
10997 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
10998 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10999 * gimple-ssa-warn-restrict.h: Adjust declarations.
11000 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11001 gimple_no_warning_p and gimple_set_no_warning with
11002 warning_suppressed_p, and suppress_warning.
11003 (check_call): Same.
11004 (check_bounds_or_overlap): Same.
11005 * gimple.c (gimple_build_call_from_tree): Same.
11006 * gimplify.c (gimplify_return_expr): Same.
11007 (gimplify_cond_expr): Same.
11008 (gimplify_modify_expr_complex_part): Same.
11009 (gimplify_modify_expr): Same.
11010 (gimple_push_cleanup): Same.
11011 (gimplify_expr): Same.
11012 * omp-expand.c (expand_omp_for_generic): Same.
11013 (expand_omp_taskloop_for_outer): Same.
11014 * omp-low.c (lower_rec_input_clauses): Same.
11015 (lower_lastprivate_clauses): Same.
11016 (lower_send_clauses): Same.
11017 (lower_omp_target): Same.
11018 * tree-cfg.c (pass_warn_function_return::execute): Same.
11019 * tree-complex.c (create_one_component_var): Same.
11020 * tree-inline.c (remap_gimple_op_r): Same.
11021 (copy_tree_body_r): Same.
11022 (declare_return_variable): Same.
11023 (expand_call_inline): Same.
11024 * tree-nested.c (lookup_field_for_decl): Same.
11025 * tree-sra.c (create_access_replacement): Same.
11026 (generate_subtree_copies): Same.
11027 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11028 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11029 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11030 * tree-ssa-loop-im.c (execute_sm): Same.
11031 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11032 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11033 (handle_builtin_strcpy): Same.
11034 (maybe_diag_stxncpy_trunc): Same.
11035 (handle_builtin_stxncpy_strncat): Same.
11036 (handle_builtin_strcat): Same.
11037 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11038 (set_no_uninit_warning): Same.
11039 (uninit_undefined_value_p): Same.
11040 (warn_uninit): Same.
11041 (maybe_warn_operand): Same.
11042 * tree-vrp.c (compare_values_warnv): Same.
11043 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11044 (test_for_singularity): Same.
11045 * gimple.h (warning_suppressed_p): New function.
11046 (suppress_warning): Same.
11047 (copy_no_warning): Same.
11048 (gimple_set_block): Call gimple_set_location.
11049 (gimple_set_location): Call copy_warning.
11051 2021-06-25 Martin Sebor <msebor@redhat.com>
11053 * tree.h (warning_suppressed_at, copy_warning,
11054 warning_suppressed_p, suppress_warning): New functions.
11056 2021-06-25 Martin Sebor <msebor@redhat.com>
11058 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11059 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11060 * diagnostic-spec.c: New file.
11061 * diagnostic-spec.h: New file.
11062 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11064 * warning-control.cc: New file.
11066 2021-06-25 liuhongt <hongtao.liu@intel.com>
11069 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11072 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11074 PR tree-optimization/101189
11075 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11076 LHS range of condition to postfold routine.
11077 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11078 FALSE edge if the LHS range supports it being taken.
11079 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11081 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11083 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11084 (relation_oracle::find_relation_block): Check correct bitmap.
11085 (relation_oracle::dump): Do not dump NULL blocks.
11087 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11089 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11090 range_on_edge instead of manually calculating.
11092 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11094 * range-op.cc: Fix comment.
11096 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11099 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11100 Handle V8QI and V4HI modes.
11101 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11103 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11104 (mmxpackmode): New mode attribute.
11105 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11106 (mmxunpackmode): New mode attribute.
11107 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11108 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11109 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11110 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11111 * config/i386/i386.md (extsuffix): Move from ...
11112 * config/i386/sse.md: ... here.
11114 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11116 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11117 (dwarf2out_finish): ...instead of here.
11119 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11121 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11122 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11123 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11124 * configure: Regenerate.
11126 2021-06-24 Richard Biener <rguenther@suse.de>
11128 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11129 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11130 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11132 2021-06-24 Richard Biener <rguenther@suse.de>
11134 * config/i386/sse.md (avx_addsubv4df3): Rename to
11136 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11137 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11138 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11139 * config/i386/i386-builtin.def: Adjust.
11140 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11141 * optabs.def (vec_addsub_optab): New optab.
11142 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11143 (slp_patterns): Add addsub_pattern.
11144 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11145 across CFN_VEC_ADDSUB.
11146 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11148 * doc/md.texi (vec_addsub<mode>3): Document.
11150 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11152 PR middle-end/101170
11153 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11154 where regno + subreg_regno_offset wraps around use 0 as starting
11157 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11159 PR middle-end/101172
11160 * stor-layout.c (finish_bitfield_representative): If nextf has
11161 error_mark_node type, set repr type to error_mark_node too.
11163 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11165 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11167 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11169 2021-06-24 Richard Biener <rguenther@suse.de>
11171 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11172 across operations that have different semantics on different
11175 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11177 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11178 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11179 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11180 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11181 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11182 initializer/combiner gimplification.
11183 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11184 on target constructs.
11185 (lower_rec_input_clauses): Likewise.
11186 (lower_omp_target): Likewise.
11187 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11188 on target if in_reduction is present.
11190 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11192 * tree-predcom.c (class pcom_worker): New class.
11193 (release_chain): Renamed to...
11194 (pcom_worker::release_chain): ...this.
11195 (release_chains): Renamed to...
11196 (pcom_worker::release_chains): ...this.
11197 (aff_combination_dr_offset): Renamed to...
11198 (pcom_worker::aff_combination_dr_offset): ...this.
11199 (determine_offset): Renamed to...
11200 (pcom_worker::determine_offset): ...this.
11201 (class comp_ptrs): New class.
11202 (split_data_refs_to_components): Renamed to...
11203 (pcom_worker::split_data_refs_to_components): ...this,
11204 and update with class comp_ptrs.
11205 (suitable_component_p): Renamed to...
11206 (pcom_worker::suitable_component_p): ...this.
11207 (filter_suitable_components): Renamed to...
11208 (pcom_worker::filter_suitable_components): ...this.
11209 (valid_initializer_p): Renamed to...
11210 (pcom_worker::valid_initializer_p): ...this.
11211 (find_looparound_phi): Renamed to...
11212 (pcom_worker::find_looparound_phi): ...this.
11213 (add_looparound_copies): Renamed to...
11214 (pcom_worker::add_looparound_copies): ...this.
11215 (determine_roots_comp): Renamed to...
11216 (pcom_worker::determine_roots_comp): ...this.
11217 (determine_roots): Renamed to...
11218 (pcom_worker::determine_roots): ...this.
11219 (single_nonlooparound_use): Renamed to...
11220 (pcom_worker::single_nonlooparound_use): ...this.
11221 (remove_stmt): Renamed to...
11222 (pcom_worker::remove_stmt): ...this.
11223 (execute_pred_commoning_chain): Renamed to...
11224 (pcom_worker::execute_pred_commoning_chain): ...this.
11225 (execute_pred_commoning): Renamed to...
11226 (pcom_worker::execute_pred_commoning): ...this.
11227 (struct epcc_data): New member worker.
11228 (execute_pred_commoning_cbck): Call execute_pred_commoning
11229 with pcom_worker pointer.
11230 (find_use_stmt): Renamed to...
11231 (pcom_worker::find_use_stmt): ...this.
11232 (find_associative_operation_root): Renamed to...
11233 (pcom_worker::find_associative_operation_root): ...this.
11234 (find_common_use_stmt): Renamed to...
11235 (pcom_worker::find_common_use_stmt): ...this.
11236 (combinable_refs_p): Renamed to...
11237 (pcom_worker::combinable_refs_p): ...this.
11238 (reassociate_to_the_same_stmt): Renamed to...
11239 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11240 (stmt_combining_refs): Renamed to...
11241 (pcom_worker::stmt_combining_refs): ...this.
11242 (combine_chains): Renamed to...
11243 (pcom_worker::combine_chains): ...this.
11244 (try_combine_chains): Renamed to...
11245 (pcom_worker::try_combine_chains): ...this.
11246 (prepare_initializers_chain): Renamed to...
11247 (pcom_worker::prepare_initializers_chain): ...this.
11248 (prepare_initializers): Renamed to...
11249 (pcom_worker::prepare_initializers): ...this.
11250 (prepare_finalizers_chain): Renamed to...
11251 (pcom_worker::prepare_finalizers_chain): ...this.
11252 (prepare_finalizers): Renamed to...
11253 (pcom_worker::prepare_finalizers): ...this.
11254 (tree_predictive_commoning_loop): Renamed to...
11255 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11256 some calls and remove some cleanup code.
11257 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11258 (static variable looparound_phis): Remove.
11259 (static variable name_expansions): Remove.
11261 2021-06-24 Richard Biener <rguenther@suse.de>
11263 * tree-vect-slp.c (slpg_vertex): New struct.
11264 (vect_slp_build_vertices): Adjust.
11265 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11266 and a materialized one.
11268 2021-06-24 Richard Biener <rguenther@suse.de>
11270 PR tree-optimization/101105
11271 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11272 Only ignore steps when they are equal or scalar order is preserved.
11274 2021-06-24 liuhongt <hongtao.liu@intel.com>
11277 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11278 Adjust comments for ix86_expand_vecop_qihi2.
11279 (ix86_expand_vecmul_qihi): Renamed to ..
11280 (ix86_expand_vecop_qihi2): Adjust function prototype to
11281 support shift operation, add static to definition.
11282 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11283 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11284 ix86_expand_vec_shift_qihi_constant.
11285 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11286 (ix86_expand_vec_shift_qihi_constant): Deleted.
11287 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11289 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11290 condition TARGET_64BIT.
11291 (mul<mode>3): Ditto.
11292 (<insn><mode>3): Ditto.
11293 (vlshr<mode>3): Extend to support avx512 vlshr.
11294 (v<insn><mode>3): New expander for
11296 (v<insn>v8qi3): Ditto.
11297 (vashrv8hi3<mask_name>): Renamed to ..
11298 (vashr<mode>3): And extend to support V16QImode for avx512.
11299 (vashrv16qi3): Deleted.
11300 (vashrv2di3<mask_name>): Extend expander to support avx512
11303 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11305 * doc/lto.texi (Design Overview): Update that slim objects are
11308 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11310 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11311 of OTHER_POWER10_MASKS so it will not be enabled by default.
11313 2021-06-23 Richard Biener <rguenther@suse.de>
11314 Martin Jambor <mjambor@suse.cz>
11316 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11317 param replacement unconditionally. Adjust comment.
11319 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11321 * Makefile.in (OBJS): Add gimple-range-fold.o
11322 * gimple-range-fold.cc: New.
11323 * gimple-range-fold.h: New.
11324 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11325 (gimple_range_calc_op2): Ditto.
11326 * gimple-range-gori.h: Move prototypes to here.
11327 * gimple-range.cc: Adjust include files.
11328 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11329 (fur_source::get_operand): Ditto.
11330 (fur_source::get_phi_operand): Ditto.
11331 (fur_source::query_relation): Ditto.
11332 (fur_source::register_relation): Ditto.
11333 (class fur_edge): Ditto.
11334 (fur_edge::fur_edge): Ditto.
11335 (fur_edge::get_operand): Ditto.
11336 (fur_edge::get_phi_operand): Ditto.
11337 (fur_stmt::fur_stmt): Ditto.
11338 (fur_stmt::get_operand): Ditto.
11339 (fur_stmt::get_phi_operand): Ditto.
11340 (fur_stmt::query_relation): Ditto.
11341 (class fur_depend): Relocate to gimple-range-fold.h.
11342 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
11343 (fur_depend::register_relation): Ditto.
11344 (fur_depend::register_relation): Ditto.
11345 (class fur_list): Ditto.
11346 (fur_list::fur_list): Ditto.
11347 (fur_list::get_operand): Ditto.
11348 (fur_list::get_phi_operand): Ditto.
11349 (fold_range): Ditto.
11350 (adjust_pointer_diff_expr): Ditto.
11351 (gimple_range_adjustment): Ditto.
11352 (gimple_range_base_of_assignment): Ditto.
11353 (gimple_range_operand1): Ditto.
11354 (gimple_range_operand2): Ditto.
11355 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
11356 (gimple_range_calc_op2): Ditto.
11357 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
11358 (fold_using_range::range_of_range_op): Ditto.
11359 (fold_using_range::range_of_address): Ditto.
11360 (fold_using_range::range_of_phi): Ditto.
11361 (fold_using_range::range_of_call): Ditto.
11362 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11363 (fold_using_range::range_of_builtin_call): Ditto.
11364 (fold_using_range::range_of_cond_expr): Ditto.
11365 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
11366 (fold_using_range::relation_fold_and_or): Ditto.
11367 (fold_using_range::postfold_gcond_edges): Ditto.
11368 * gimple-range.h: Add gimple-range-fold.h to include files. Change
11369 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
11370 (gimple_range_handler): Relocate to gimple-range-fold.h.
11371 (gimple_range_ssa_p): Ditto.
11372 (range_compatible_p): Ditto.
11373 (class fur_source): Ditto.
11374 (class fur_stmt): Ditto.
11375 (class fold_using_range): Ditto.
11376 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
11377 (gimple_range_calc_op2): Ditto.
11379 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11381 PR tree-optimization/101148
11382 PR tree-optimization/101014
11383 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
11384 (ranger_cache::~ranger_cache): Adjust.
11385 (ranger_cache::block_range): Check if propagation disallowed.
11386 (ranger_cache::propagate_cache): Disallow propagation if new value
11387 can't be stored properly.
11388 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
11390 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11392 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
11393 (sbr_vector::set_bb_range): Return true.
11394 (class sbr_sparse_bitmap): Adjust.
11395 (sbr_sparse_bitmap::set_bb_range): Return value.
11396 (block_range_cache::set_bb_range): Return value.
11397 (ranger_cache::propagate_cache): Use return value to print msg.
11398 * gimple-range-cache.h (class block_range_cache): Adjust.
11400 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11402 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
11404 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
11406 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
11407 so this pattern can be used for test/compare removal. Pass
11408 current insn to compute_logical_op_length and output_logical_op.
11409 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
11410 (h8300_and_costs): Add argument to compute_logical_op_length.
11411 (output_logical_op): Add new argument. Use it to determine if the
11412 condition codes are used and adjust the output accordingly.
11413 (compute_logical_op_length): Add new argument and update length
11414 computations when condition codes are used.
11415 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
11417 (output_logical_op): Likewise.
11419 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11422 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
11423 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
11424 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
11425 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
11426 * config/i386/sse.md (unspec): ... here.
11428 2021-06-23 Martin Liska <mliska@suse.cz>
11431 * optc-save-gen.awk: Put back arm_fp16_format to
11434 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11437 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
11440 (clz<mode>2): Update RTX pattern for additions.
11442 2021-06-23 Jakub Jelinek <jakub@redhat.com>
11444 PR middle-end/101167
11445 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
11446 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
11448 2021-06-22 Sergei Trofimovich <siarheit@google.com>
11450 * doc/rtl.texi: drop unbalanced parenthesis.
11452 2021-06-22 Richard Biener <rguenther@suse.de>
11454 PR middle-end/101156
11455 * gimplify.c (gimplify_expr): Remove premature incorrect
11458 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11460 PR tree-optimization/101159
11461 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
11464 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11466 PR middle-end/101160
11467 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
11468 clear crtl->return_rtx instead of keeping it referencing a pseudo.
11470 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11471 Andrew Pinski <apinski@marvell.com>
11473 PR tree-optimization/101162
11474 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
11477 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11479 * range-op.cc (range_relational_tests): New.
11480 (range_op_tests): Call range_relational_tests.
11482 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11484 * range-op.cc (operator_cast::lhs_op1_relation): New.
11485 (operator_identity::lhs_op1_relation): Mew.
11487 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11489 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
11491 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11493 * range-op.cc (operator_plus::lhs_op1_relation): New.
11494 (operator_plus::lhs_op2_relation): New.
11496 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11498 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
11499 relation_oracle if dominators exist.
11500 (ranger_cache::~ranger_cache): Dispose of oracle.
11501 (ranger_cache::dump_bb): Dump oracle.
11502 * gimple-range.cc (fur_source::fur_source): New.
11503 (fur_source::get_operand): Use mmeber query.
11504 (fur_source::get_phi_operand): Use member_query.
11505 (fur_source::query_relation): New.
11506 (fur_source::register_dependency): Delete.
11507 (fur_source::register_relation): New.
11508 (fur_edge::fur_edge): Adjust.
11509 (fur_edge::get_phi_operand): Fix comment.
11510 (fur_edge::query): Delete.
11511 (fur_stmt::fur_stmt): Adjust.
11512 (fur_stmt::query): Delete.
11513 (fur_depend::fur_depend): Adjust.
11514 (fur_depend::register_relation): New.
11515 (fur_depend::register_relation): New.
11516 (fur_list::fur_list): Adjust.
11517 (fur_list::get_operand): Use member query.
11518 (fold_using_range::range_of_range_op): Process and query relations.
11519 (fold_using_range::range_of_address): Adjust dependency call.
11520 (fold_using_range::range_of_phi): Ditto.
11521 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
11522 (fold_using_range::relation_fold_and_or): New.
11523 (fold_using_range::postfold_gcond_edges): New.
11524 * gimple-range.h (class gimple_ranger): Adjust.
11525 (class fur_source): Adjust members.
11526 (class fur_stmt): Ditto.
11527 (class fold_using_range): Ditto.
11529 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11531 * range-op.cc (range_operator::wi_fold): Apply relation effect.
11532 (range_operator::fold_range): Adjust and apply relation effect.
11533 (*::fold_range): Add relation parameters.
11534 (*::op1_range): Ditto.
11535 (*::op2_range): Ditto.
11536 (range_operator::lhs_op1_relation): New.
11537 (range_operator::lhs_op2_relation): New.
11538 (range_operator::op1_op2_relation): New.
11539 (range_operator::op1_op2_relation_effect): New.
11540 (relop_early_resolve): New.
11541 (operator_equal::op1_op2_relation): New.
11542 (operator_equal::fold_range): Call relop_early_resolve.
11543 (operator_not_equal::op1_op2_relation): New.
11544 (operator_not_equal::fold_range): Call relop_early_resolve.
11545 (operator_lt::op1_op2_relation): New.
11546 (operator_lt::fold_range): Call relop_early_resolve.
11547 (operator_le::op1_op2_relation): New.
11548 (operator_le::fold_range): Call relop_early_resolve.
11549 (operator_gt::op1_op2_relation): New.
11550 (operator_gt::fold_range): Call relop_early_resolve.
11551 (operator_ge::op1_op2_relation): New.
11552 (operator_ge::fold_range): Call relop_early_resolve.
11553 * range-op.h (class range_operator): Adjust parameters and methods.
11555 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11557 * Makefile.in (OBJS): Add value-relation.o.
11558 * gimple-range.h: Adjust include files.
11559 * tree-data-ref.c: Adjust include file order.
11560 * value-query.cc (range_query::get_value_range): Default to no oracle.
11561 (range_query::query_relation): New.
11562 (range_query::query_relation): New.
11563 * value-query.h (class range_query): Adjust.
11564 * value-relation.cc: New.
11565 * value-relation.h: New.
11567 2021-06-22 Richard Biener <rguenther@suse.de>
11569 PR tree-optimization/101151
11570 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
11573 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
11575 * config/riscv/riscv.c (thead_c906_tune_info): New.
11576 (riscv_tune_info_table): Use new tune.
11578 2021-06-22 Richard Biener <rguenther@suse.de>
11580 PR tree-optimization/101158
11581 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
11582 checking after checking for matching operation.
11584 2021-06-22 Richard Biener <rguenther@suse.de>
11586 PR tree-optimization/101159
11587 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
11588 missing NULL vectype check.
11590 2021-06-22 Richard Biener <rguenther@suse.de>
11592 PR tree-optimization/101154
11593 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
11595 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11598 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
11599 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
11600 * config/i386/i386.c (ix86_expand_prologue): Clear it.
11601 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
11602 Remove "" from match_operand. Emit new insns using emit_move_insn and
11603 set ix86_last_zero_store_uid to INSN_UID of the last store.
11604 Add peephole2s for 1/2/4 stores of const0_rtx following previous
11607 2021-06-22 Martin Liska <mliska@suse.cz>
11609 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
11612 2021-06-22 Martin Liska <mliska@suse.cz>
11614 * gcov-io.h: Remove padding entries.
11616 2021-06-22 liuhongt <hongtao.liu@intel.com>
11618 PR tree-optimization/97770
11619 * tree-vect-patterns.c (vect_recog_popcount_pattern):
11621 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
11623 2021-06-22 liuhongt <hongtao.liu@intel.com>
11626 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
11627 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
11628 (expand<mode>_mask): this ..
11629 (*expand<mode>_mask): New pre_reload splitter to transform
11630 v{,p}expand* to vmov* when mask is zero, all ones, or has all
11631 ones in it's lower part, otherwise still generate
11634 2021-06-22 liuhongt <hongtao.liu@intel.com>
11637 * config/i386/i386-expand.c
11638 (ix86_expand_special_args_builtin): Keep constm1_operand only
11639 if it satisfies insn's operand predicate.
11641 2021-06-21 Jason Merrill <jason@redhat.com>
11644 * df-scan.c (df_ref_record): Check that regno < endregno.
11645 * function.c (assign_parms, expand_function_end): Do nothing with a
11646 TYPE_EMPTY_P result.
11648 2021-06-21 Richard Biener <rguenther@suse.de>
11650 PR tree-optimization/101120
11651 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
11653 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
11654 DR chain DCE capability.
11655 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
11656 * tree-vect-stmts.c (vectorizable_load): Remove unused
11657 loads in the DR chain for SLP.
11659 2021-06-21 Jakub Jelinek <jakub@redhat.com>
11661 PR inline-asm/100785
11662 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
11663 output or input operands were already error_mark_node.
11664 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
11665 remove all inputs, outputs and clobbers from the asm and
11666 set template to "".
11668 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11670 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
11671 (vceq_s16): Likewise.
11672 (vceq_s32): Likewise.
11673 (vceq_u8): Likewise.
11674 (vceq_u16): Likewise.
11675 (vceq_u32): Likewise.
11676 (vceq_p8): Likewise.
11677 (vceqq_s8): Likewise.
11678 (vceqq_s16): Likewise.
11679 (vceqq_s32): Likewise.
11680 (vceqq_u8): Likewise.
11681 (vceqq_u16): Likewise.
11682 (vceqq_u32): Likewise.
11683 (vceqq_p8): Likewise.
11684 (vceq_f32): Gate __a == __b on __FAST_MATH__.
11685 (vceqq_f32): Likewise.
11686 (vceq_f16): Likewise.
11687 (vceqq_f16): Likewise.
11689 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11692 * config/arm/iterators.md (NEON_VACMP): Remove.
11693 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
11695 (neon_vca<cmp_op><mode>_insn): Likewise.
11696 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
11699 2021-06-21 Richard Biener <rguenther@suse.de>
11701 PR tree-optimization/101121
11702 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
11703 when we just lack a stmt with the desired op when doing permutation.
11704 (vect_build_slp_tree): When caching a failed SLP build attempt
11705 assert that at least one lane is marked as not matching.
11707 2021-06-21 liuhongt <hongtao.liu@intel.com>
11710 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
11711 register alternative.
11712 (*and<mode>_1): Ditto.
11714 (*andn<mode>_1): Ditto.
11715 (*<code><mode>_1): Ditto.
11716 (*<code>qi_1): Ditto.
11717 (*one_cmpl<mode>2_1): Ditto.
11718 (*one_cmplsi2_1_zext): Ditto.
11719 (*one_cmplqi2_1): Ditto.
11720 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
11721 the order of mask registers to be before general registers.
11723 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
11726 * config/i386/i386.md: New define_peephole2s to shrink writing
11727 1, 2 or 4 consecutive zeros to memory when optimizing for size.
11729 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
11731 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
11732 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
11733 more efficient code when the source can be trivially simplified.
11735 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
11737 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
11738 a range if global is not available.
11739 (ranger_cache::entry_range): Fallback to range_of_def.
11740 * gimple-range-cache.h (range_of_def): Adjust prototype.
11742 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
11744 PR tree-optimization/101014
11745 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
11747 (ranger_cache::~ranger_cache): Ditto.
11748 (ranger_cache::enable_new_values): Delete.
11749 (ranger_cache::push_poor_value): Delete.
11750 (ranger_cache::range_of_def): Remove poor value processing.
11751 (ranger_cache::entry_range): Ditto.
11752 (ranger_cache::fill_block_cache): Ditto.
11753 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
11754 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
11755 * gimple-range.h (class gimple_ranger): Adjust.
11757 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
11760 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
11761 derived from arm_canon_arch.
11762 (arm_canon_arch_option): Call it.
11763 (arm_canon_arch_multilib_option): New function.
11764 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
11765 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
11766 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
11767 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
11768 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
11769 * config/arm/arm.opt (mlibarch): New option.
11770 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
11771 of march on RHS with mlibarch.
11773 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
11775 * config.in: Regenerate.
11776 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
11778 * configure: Regenerate.
11779 * configure.ac: Fix for global_load assembler functions.
11781 2021-06-18 Richard Biener <rguenther@suse.de>
11783 PR tree-optimization/101112
11784 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
11785 to lookup a pattern stmt def.
11787 2021-06-18 Jakub Jelinek <jakub@redhat.com>
11789 PR middle-end/101062
11790 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
11791 representatives in QUAL_UNION_TYPE.
11793 2021-06-18 Andrew Pinski <apinski@marvell.com>
11795 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
11796 Add counting of how many times it is done.
11797 (factor_out_conditional_conversion): Likewise.
11798 (match_simplify_replacement): Likewise.
11799 (value_replacement): Likewise.
11800 (spaceship_replacement): Likewise.
11801 (cond_store_replacement): Likewise.
11802 (cond_if_else_store_replacement_1): Likewise.
11803 (hoist_adjacent_loads): Likewise.
11805 2021-06-18 Andrew Pinski <apinski@marvell.com>
11807 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
11808 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
11809 (verify_gimple_assign_binary): Reject point and offset types on
11810 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
11811 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
11812 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
11814 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
11816 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
11817 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
11818 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
11821 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
11823 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
11824 earlyclobber to alts 0/1.
11825 (gen_addadd): Add earlyclobber to alts 0/1.
11826 * config/rs6000/fusion.md: Regenerate file.
11828 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11830 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
11832 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
11834 * gimple-range-cache.cc: Comment cleanups.
11835 * gimple-range-gori.cc: Comment cleanups.
11836 * gimple-range.cc: Comment/spacing cleanups
11837 * value-range.h: Comment cleanups.
11839 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
11842 * calls.c (expand_call): Replace PUSH_ARGS with
11843 targetm.calls.push_argument (0).
11844 (emit_library_call_value_1): Likewise.
11845 * defaults.h (PUSH_ARGS): Removed.
11846 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
11847 targetm.calls.push_argument (0).
11848 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
11849 (emit_push_insn): Pass the number bytes to push to
11850 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
11851 * hooks.c (hook_bool_uint_true): New.
11852 * hooks.h (hook_bool_uint_true): Likewise.
11853 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
11854 targetm.calls.push_argument (0).
11855 * target.def (push_argument): Add a targetm.calls hook.
11856 * targhooks.c (default_push_argument): New.
11857 * targhooks.h (default_push_argument): Likewise.
11858 * config/bpf/bpf.h (PUSH_ARGS): Removed.
11859 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
11860 * config/cr16/cr16.h (PUSH_ARGS): Removed.
11861 * config/i386/i386.c (ix86_push_argument): New.
11862 (TARGET_PUSH_ARGUMENT): Likewise.
11863 * config/i386/i386.h (PUSH_ARGS): Removed.
11864 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
11865 * config/m32c/m32c.h (PUSH_ARGS): Removed.
11866 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
11867 * config/pru/pru.h (PUSH_ARGS): Likewise.
11868 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
11869 TARGET_PUSH_ARGUMENT hook.
11870 * doc/tm.texi: Regenerated.
11872 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
11875 * config/i386/i386-expand.c (expand_vector_set_var):
11876 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
11877 ix86_expand_vector_init_duplicate.
11878 (ix86_expand_vector_init_duplicate): Emit insv_1 for
11879 QImode for !TARGET_PARTIAL_REG_STALL.
11880 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
11881 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
11882 as operand 2 predicate. Call ix86_expand_vector_set_var
11883 for non-constant index operand.
11884 (vec_setv2si): Ditto.
11885 (vec_setv4hi): Ditto.
11886 (vec_setv8qi): ditto.
11888 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
11890 PR tree-optimization/100790
11891 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
11894 2021-06-17 Martin Liska <mliska@suse.cz>
11896 * doc/invoke.texi: Use consistently -O1 instead of -O.
11898 2021-06-17 Martin Liska <mliska@suse.cz>
11900 * gcov-io.h: Update documentation entry about string format.
11902 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
11905 * config/s390/vecintrin.h (vec_doublee): Fix to use
11906 __builtin_s390_vflls.
11907 (vec_floate): Fix to use __builtin_s390_vflrd.
11909 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11911 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
11912 * dominance.h (get_dominated_to_depth): Likewise.
11913 (get_all_dominated_blocks): Likewise.
11914 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
11915 * gcse.c (hoist_code): Likewise.
11916 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
11917 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
11918 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
11919 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
11921 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11923 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
11924 * dominance.h (get_dominated_by_region): Likewise.
11925 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
11926 (gimple_duplicate_sese_tail): Likewise.
11927 (move_sese_region_to_fn): Likewise.
11929 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11931 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
11932 * dominance.h (get_dominated_by): Likewise.
11933 * auto-profile.c (afdo_find_equiv_class): Adjust.
11934 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
11935 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
11936 * tree-cfg.c (test_linear_chain): Likewise.
11937 (test_diamond): Likewise.
11939 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11941 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
11942 * cfgloopanal.c (get_loop_hot_path): Likewise.
11943 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
11945 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11947 * cgraph.c (cgraph_node::collect_callers): Return
11948 auto_vec<cgraph_edge *>.
11949 * cgraph.h (cgraph_node::collect_callers): Likewise.
11950 * ipa-cp.c (create_specialized_node): Adjust.
11951 (decide_about_value): Likewise.
11952 (decide_whether_version_node): Likewise.
11953 * ipa-sra.c (process_isra_node_results): Likewise.
11955 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11957 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
11958 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
11960 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
11963 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
11965 * gimple-range.cc (debug_seed_ranger): New.
11966 (dump_ranger): New.
11967 (debug_ranger): New.
11969 2021-06-17 Richard Biener <rguenther@suse.de>
11971 PR tree-optimization/54400
11972 * tree-vectorizer.h (enum slp_instance_kind): Add
11973 slp_inst_kind_bb_reduc.
11974 (reduction_fn_for_scalar_code): Declare.
11975 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
11976 Check SLP_INSTANCE_KIND instead of looking at the
11978 (vect_slp_analyze_instance_alignment): Likewise.
11979 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
11980 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
11981 chain linearization from vect_build_slp_tree_2 and generalize
11982 for the use of BB reduction vectorization.
11983 (vect_build_slp_tree_2): Adjust accordingly.
11984 (vect_optimize_slp): Elide permutes at the root of BB reduction
11986 (vectorizable_bb_reduc_epilogue): New function.
11987 (vect_slp_prune_covered_roots): Likewise.
11988 (vect_slp_analyze_operations): Use them.
11989 (vect_slp_check_for_constructors): Recognize associatable
11990 chains for BB reduction vectorization.
11991 (vectorize_slp_instance_root_stmt): Generate code for the
11992 BB reduction epilogue.
11994 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
11996 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
11998 (gori_compute::may_recompute_p): New.
11999 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12000 * gimple-range-gori.h (class gori_compute): Add prototype.
12002 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12004 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12005 true when a range can be calculated.
12006 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12008 2021-06-16 Martin Sebor <msebor@redhat.com>
12010 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12011 Correct documented defaults.
12013 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12015 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12016 m_new_value_p directly.
12018 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12021 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12022 Handle 64bit modes for TARGET_SSE4_1.
12023 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12024 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12025 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12026 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12027 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12028 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12030 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12032 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12033 Change to an expander that emits the correct instruction
12034 depending on endianness.
12035 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12036 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12038 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12040 * config/aarch64/aarch64-simd-builtins.def: Split generator
12041 for aarch64_<su>qmovn builtins into scalar and vector
12043 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12045 (aarch64_<su>qmovn<mode>_insn_be): Define.
12046 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12047 variants. Change vector variant to an expander that emits the
12048 correct instruction depending on endianness.
12050 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12052 * config/aarch64/aarch64-simd-builtins.def: Split generator
12053 for aarch64_sqmovun builtins into scalar and vector variants.
12054 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12055 Split into scalar and vector variants. Change vector variant
12056 to an expander that emits the correct instruction depending
12058 (aarch64_sqmovun<mode>_insn_le): Define.
12059 (aarch64_sqmovun<mode>_insn_be): Define.
12061 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12063 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12064 Define - modeling zero-high-half semantics.
12065 (aarch64_xtn<mode>): Change to an expander that emits the
12066 appropriate instruction depending on endianness.
12067 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12069 (aarch64_xtn2<mode>_le): Rename to...
12070 (aarch64_xtn2<mode>_insn_le): This.
12071 (aarch64_xtn2<mode>_be): Rename to...
12072 (aarch64_xtn2<mode>_insn_be): This.
12073 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12075 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12076 attribute iterator.
12078 2021-06-16 Martin Jambor <mjambor@suse.cz>
12080 PR tree-optimization/100453
12081 * tree-sra.c (create_access): Disqualify any const candidates
12082 which are written to.
12083 (sra_modify_expr): Do not store sub-replacements back to a const base.
12084 (handle_unscalarized_data_in_subtree): Likewise.
12085 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12086 instead of constant_decl_p.
12088 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12090 PR middle-end/101062
12091 * stor-layout.c (finish_bitfield_representative): For fields in unions
12092 assume nextf is always NULL.
12093 (finish_bitfield_layout): Compute bit field representatives also in
12094 unions, but handle it as if each bitfield was the only field in the
12097 2021-06-16 Richard Biener <rguenther@suse.de>
12099 PR tree-optimization/101088
12100 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12101 supported refs on edges. Do not assert same ref but
12102 different kind stores are unsuported but mark them so.
12103 (hoist_memory_references): Only look for supported refs
12106 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12108 PR rtl-optimization/46235
12109 * config/i386/i386.md: New define_split for bt followed by cmov.
12110 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12111 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12112 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12113 by setnc with zero extension.
12115 2021-06-16 Richard Biener <rguenther@suse.de>
12117 PR tree-optimization/101083
12118 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12119 vectype as argument.
12120 (vect_build_slp_tree_2): Adjust.
12122 2021-06-15 Martin Sebor <msebor@redhat.com>
12124 PR middle-end/100876
12125 * builtins.c: (gimple_call_return_array): Account for size_t
12126 mangling as either unsigned int or unsigned long
12128 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12130 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12131 up before eliminating comparisons.
12133 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12135 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12136 nonzero|X is nonzero.
12137 (range_op_bitwise_and_tests): Add tests for above.
12139 2021-06-15 Carl Love <cel@us.ibm.com>
12142 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12144 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12147 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12150 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12151 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12152 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12153 (omp_notice_variable): Update type classification for Fortran.
12154 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12155 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12156 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12157 * langhooks.c (lhd_omp_scalar_p): Likewise.
12158 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12159 (LANG_HOOKS_DECLS): Add them.
12160 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12161 omp_scalar_p pointer type to include the new bool argument.
12163 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12165 * doc/analyzer.texi
12166 (Special Functions for Debugging the Analyzer): Add
12167 __analyzer_dump_capacity.
12169 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12172 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12173 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12175 2021-06-15 Richard Biener <rguenther@suse.de>
12177 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12178 check to identify loop latches.
12179 * cfgloop.c (verify_loop_structure): Likewise.
12180 * loop-init.c (apply_loop_flags): Allow marked irreducible
12181 regions even with multiple latches.
12182 * predict.c (rebuild_frequencies): Simplify.
12184 2021-06-15 Richard Biener <rguenther@suse.de>
12186 * tree-ssa-threadupdate.c
12187 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12188 have marked irreducible regions.
12190 2021-06-14 Martin Sebor <msebor@redhat.com>
12193 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12194 Handle calls to placement new.
12195 (ndecl_dealloc_argno): Avoid placement delete.
12197 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12200 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12201 create_tmp_reg_or_ssa_name().
12203 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12205 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12206 (ranger_cache::enable_new_values): Set to specified value and
12207 return the old value.
12208 (ranger_cache::disable_new_values): Delete.
12209 (ranger_cache::fill_block_cache): Disable non 1st order derived
12211 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12212 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12214 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12217 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12218 Return true early when testing with V2HImode.
12219 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12221 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12223 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12224 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12225 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12226 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12227 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12229 (vec_unpack<US>_lo_<mode>): Likewise.
12230 (vec_pack_trunc_<mode>): Rename to
12231 neon_quad_vec_pack_trunc_<mode>.
12232 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12234 (vec_unpack<US>_lo_<mode>): New.
12235 (vec_pack_trunc_<mode>): New.
12237 2021-06-14 Richard Biener <rguenther@suse.de>
12239 PR tree-optimization/100934
12240 * tree-ssa-dom.c (pass_dominator::execute): Properly
12241 mark irreducible regions.
12243 2021-06-14 Martin Liska <mliska@suse.cz>
12245 * doc/invoke.texi: Put r{...} on the same line as @item.
12247 2021-06-14 Martin Liska <mliska@suse.cz>
12249 * doc/invoke.texi: Add missing newline.
12251 2021-06-14 Martin Liska <mliska@suse.cz>
12253 * doc/invoke.texi: Remove '+' charasters.
12255 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12257 * config.gcc (arc): Add support for with_cpu option.
12258 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12260 2021-06-14 Richard Biener <rguenther@suse.de>
12262 PR tree-optimization/101031
12263 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12264 instead of size when accounting for a possibly string
12267 2021-06-14 Martin Liska <mliska@suse.cz>
12269 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12271 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12273 * value-query.cc (gimple_range_global): Call get_range_global
12274 if called after inlining.
12276 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12279 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12280 Emit constant permutation insn directly from here.
12282 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12284 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12286 * auto-profile.c (afdo_find_equiv_class): Likewise.
12287 * gcc.c (do_specs_vec): Likewise.
12288 (do_spec_1): Likewise.
12289 (driver::set_up_specs): Likewise.
12290 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12291 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12292 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12293 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12294 (get_location_for_stmts): Likewise.
12295 * graphite-poly.c (print_iteration_domains): Likewise.
12296 (free_poly_bb): Likewise.
12297 (remove_gbbs_in_scop): Likewise.
12298 (free_scop): Likewise.
12299 (dump_gbb_cases): Likewise.
12300 (dump_gbb_conditions): Likewise.
12301 (print_pdrs): Likewise.
12302 (print_scop): Likewise.
12303 * ifcvt.c (cond_move_process_if_block): Likewise.
12304 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12305 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12306 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12307 * sel-sched-dump.c (dump_insn_vector): Likewise.
12308 * store-motion.c (store_ops_ok): Likewise.
12309 (store_killed_in_insn): Likewise.
12310 * timevar.c (timer::named_items::print): Likewise.
12311 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12312 (cleanup_tree_cfg_noloop): Likewise.
12313 * tree-data-ref.c (dump_data_references): Likewise.
12314 (print_dir_vectors): Likewise.
12315 (print_dist_vectors): Likewise.
12316 (dump_data_dependence_relations): Likewise.
12317 (dump_dist_dir_vectors): Likewise.
12318 (dump_ddrs): Likewise.
12319 (create_runtime_alias_checks): Likewise.
12320 (free_subscripts): Likewise.
12321 (save_dist_v): Likewise.
12322 (save_dir_v): Likewise.
12323 (invariant_access_functions): Likewise.
12324 (same_access_functions): Likewise.
12325 (access_functions_are_affine_or_constant_p): Likewise.
12326 (find_data_references_in_stmt): Likewise.
12327 (graphite_find_data_references_in_stmt): Likewise.
12328 (free_dependence_relations): Likewise.
12329 (free_data_refs): Likewise.
12330 * tree-inline.c (copy_debug_stmts): Likewise.
12331 * tree-into-ssa.c (dump_currdefs): Likewise.
12332 (rewrite_update_phi_arguments): Likewise.
12333 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12334 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12336 (vect_slp_analyze_node_dependences): Likewise.
12337 (vect_slp_analyze_instance_dependence): Likewise.
12338 (vect_record_base_alignments): Likewise.
12339 (vect_get_peeling_costs_all_drs): Likewise.
12340 (vect_peeling_supportable): Likewise.
12341 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
12342 (vec_info::free_stmt_vec_infos): Likewise.
12344 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
12346 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
12347 (andqi3_1<cczn>): Removed.
12348 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
12349 (H8/SX bit logicals): Split out from other patterns.
12350 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
12351 mulqihi3_const_clobber_flags.
12352 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
12354 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
12357 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
12358 to true if red zone is used.
12359 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
12360 ix86_red_zone_used.
12361 * config/i386/i386.h (machine_function): Add red_zone_used.
12362 (ix86_red_zone_size): Removed.
12363 (ix86_red_zone_used): New.
12364 * config/i386/i386.md (peephole2 patterns): Replace
12365 ix86_red_zone_size with ix86_red_zone_used.
12367 2021-06-12 Jason Merrill <jason@redhat.com>
12369 * doc/extend.texi (unused variable attribute): Applies to
12370 structure fields as well.
12372 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
12374 * auto-profile.c (read_profile): fix a typo in an error string
12376 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
12378 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
12380 * tree-pretty-print.c (dump_omp_clauses): Update.
12381 (dump_generic_node) <OMP_CLAUSE>: Use it.
12383 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12386 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
12387 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
12388 the polymorphic variants matching code.
12389 (__arm_vld1q_z): Likewise.
12390 (__arm_vld2q): Likewise.
12391 (__arm_vld4q): Likewise.
12392 (__arm_vldrbq_gather_offset): Likewise.
12393 (__arm_vldrbq_gather_offset_z): Likewise.
12395 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12397 PR tree-optimization/96392
12398 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
12400 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12402 PR tree-optimization/96392
12403 * fold-const.c (fold_real_zero_addition_p): Take both arguments
12404 of the addition or subtraction, not just the zero. Use this
12405 other argument in tests for signaling NaNs and signed zeros.
12406 (tree_expr_maybe_real_minus_zero_p): New predicate.
12407 * fold-const.h (fold_real_zero_addition_p): Update prototype.
12408 (tree_expr_maybe_real_minus_zero_p): New function prototype.
12409 * match.pd: Update calls to fold_real_zero_addition_p.
12410 Replace HONOR_NANS with tree_expr_maybe_nan_p.
12411 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
12412 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
12413 * tree-ssa-reassoc.c (eliminate_using_constants): Update
12414 call to fold_real_zero_addition_p.
12416 2021-06-11 Richard Biener <rguenther@suse.de>
12418 PR tree-optimization/101025
12419 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
12420 all refs that require dependence checking.
12422 2021-06-11 Richard Biener <rguenther@suse.de>
12424 PR tree-optimization/101028
12425 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
12426 reassoc discovery fails fatally, mark appropriate lanes
12429 2021-06-11 Richard Biener <rguenther@suse.de>
12431 PR tree-optimization/101026
12432 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
12433 have a representative for the associated chain nodes.
12435 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12437 PR rtl-optimization/101008
12438 * simplify-rtx.c (relational_result): New function.
12439 (simplify_logical_relational_operation,
12440 simplify_relational_operation): Use it.
12442 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12445 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
12447 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
12450 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
12451 false if the permutation can be implemented with constant
12452 permutation instruction in wider mode.
12453 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
12454 Handle V8QImode and V4HImode.
12456 2021-06-11 Martin Liska <mliska@suse.cz>
12458 PR gcov-profile/100788
12459 * common.opt: Add new option.
12460 * coverage.c (coverage_begin_function): Emit warning instead on
12461 the internal compiler error.
12462 * doc/invoke.texi: Document the option.
12463 * toplev.c (process_options): Enable it by default.
12465 2021-06-11 Richard Biener <rguenther@suse.de>
12467 PR middle-end/101009
12468 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
12469 to set *init_b to true when we encounter a constant equal
12471 (compute_affine_dependence): Also dump the actual DR_REF.
12473 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12475 PR tree-optimization/100984
12476 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
12477 replacements table.
12478 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
12480 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
12482 * config/rs6000/rs6000.md
12483 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
12484 define_insn_and_split.
12486 2021-06-11 Richard Biener <rguenther@suse.de>
12488 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
12489 to sort operands of the associative chain.
12491 2021-06-11 Richard Biener <rguenther@suse.de>
12493 * system.h (gcc_stablesort_r): Declare.
12494 * sort.cc (gcc_sort_r): Support stable sort.
12495 (gcc_stablesort_r): Define.
12496 * vec.h (vec<>::stablesort): Add.
12498 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
12501 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
12502 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
12503 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
12504 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
12505 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
12506 (expand_vec_perm_interleave2): Handle 64bit modes.
12507 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
12508 (expand_vec_perm_even_odd_1): Ditto.
12509 (ix86_vectorize_vec_perm_const): Ditto.
12510 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
12511 * config/i386/sse.md: ... here.
12512 * config/i386/mmx.md (*vec_interleave_lowv2sf):
12513 New insn_and_split pattern.
12514 (*vec_interleave_highv2sf): Ditto.
12515 (mmx_pshufbv8qi3): New insn pattern.
12516 (*mmx_pblendw): Ditto.
12518 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
12520 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
12521 (build_acc): Likewise.
12522 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
12523 source operands in little-endian mode.
12524 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
12525 (mma_init_builtins): Likewise.
12526 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
12527 ordering for the MMA assemble and build source operands.
12528 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
12530 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
12533 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
12535 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
12537 * config/h8300/extensions.md: Replace _clobber_flags patterns
12540 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
12542 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
12543 (vcond_mask_<mode><tointvec>): this.
12545 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
12546 Thomas Schwinge <thomas@codesourcery.com>
12548 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
12549 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
12550 * gimple.h (enum gf_mask): Split
12551 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
12552 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
12553 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
12554 (is_gimple_omp_oacc): Update.
12555 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
12556 * gimplify.c (gimplify_omp_target_update): Likewise.
12557 * omp-expand.c (expand_omp_target, build_omp_regions_1)
12558 (omp_make_gimple_edges): Likewise.
12559 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
12562 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12564 * value-query.cc (value_query::value_on_edge): Rename name to
12566 (range_query::range_on_edge): Same.
12567 (range_query::value_of_expr): Same.
12568 (range_query::value_on_edge): Same.
12569 * value-query.h (class value_query): Same.
12570 (class range_query): Same.
12572 2021-06-10 Richard Biener <rguenther@suse.de>
12574 PR tree-optimization/101003
12575 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
12576 use the pattern stmt defs when linearizing a chain.
12578 2021-06-10 Jakub Jelinek <jakub@redhat.com>
12581 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
12582 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
12584 2021-06-10 Clement Chigot <clement.chigot@atos.net>
12586 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
12587 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
12589 2021-06-09 Andrew Pinski <apinski@marvell.com>
12591 PR tree-optimization/100925
12592 * match.pd (a ? CST1 : CST2): Limit transformations
12593 that would produce a negative to integeral types only.
12594 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
12596 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12599 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12601 * doc/tm.texi: Correctly update.
12603 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12605 * doc/tm.texi: Correctly update.
12607 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
12610 * doc/tm.texi.in (Trampolines): Add a missing blank line.
12612 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
12615 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
12616 and -ftrampolines work only with Ada.
12617 * doc/tm.texi.in (Trampolines): Likewise.
12618 * doc/tm.texi: Regenerated.
12620 2021-06-09 Carl Love <cel@us.ibm.com>
12622 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
12623 Add define for new builtins.
12624 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
12625 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
12626 overloaded builtin definitions.
12627 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
12628 VSIGNEXTSD2Q): Add builtin expansions.
12629 (SIGNEXT): Add P10 overload definition.
12630 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
12631 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
12632 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
12633 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
12634 vsignextend_si_v2di)[VIlong]: Add define_expand.
12635 Make define_insn vsx_sign_extend_si_v2di visible.
12636 * doc/extend.texi: Add documentation for the vec_signexti,
12637 vec_signextll builtins and vec_signextq.
12639 2021-06-09 Carl Love <cel@us.ibm.com>
12641 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
12642 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
12643 __floattikf_sw, __floatuntikf_sw respectively.
12644 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
12645 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
12646 define_insn for mode IEEE 128.
12648 2021-06-09 Carl Love <cel@us.ibm.com>
12650 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
12651 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
12652 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
12653 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
12654 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
12655 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
12656 uses of VSX_TI to VEC_TI.
12658 2021-06-09 Carl Love <cel@us.ibm.com>
12660 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
12662 2021-06-09 Carl Love <cel@us.ibm.com>
12664 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
12666 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
12667 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
12668 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
12669 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
12670 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
12671 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
12672 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
12674 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
12675 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
12676 altivec_vrlqnm): New define_expands.
12677 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
12678 VCMPGTUT_P): Add macro expansions.
12679 (BU_P10V_AV_P): Add builtin predicate definition.
12680 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
12681 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
12682 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
12683 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
12684 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
12685 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
12686 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
12687 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
12688 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
12689 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
12690 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
12691 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
12692 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
12693 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
12694 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
12695 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
12696 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
12697 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
12698 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
12699 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
12700 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
12701 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
12702 P10V_BUILTIN_MODU_V1TI):
12703 New overloaded definitions.
12704 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
12705 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
12706 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
12707 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
12708 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
12709 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
12711 (altivec_init_builtins): New E_V1TImode case statement.
12712 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
12713 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
12714 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
12715 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
12716 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
12717 E_V1TImode]: New case statements.
12718 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
12719 value RS6000_BTI_bool_V1TI.
12720 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
12721 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
12722 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
12723 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
12724 vlshrv1ti3, vashrv1ti3): New define_expands.
12725 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
12726 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
12727 UNSPEC_VSX_MODUQ): New unspecs.
12728 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
12729 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
12731 (vcmpnet): New define_expand.
12732 * doc/extend.texi: Add documentation for the new builtins vec_rl,
12733 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
12734 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
12735 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
12736 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
12737 vec_any_ge, vec_any_le.
12739 2021-06-09 Carl Love <cel@us.ibm.com>
12741 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
12742 bug in argument generation.
12744 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
12746 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
12748 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
12749 remove <supf> iterator.
12750 (mve_vclzq_u<mode>): New.
12751 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
12752 (neon_vclz<mode): Move to ...
12753 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
12754 * config/arm/vec-common.md: ... here. Add support for MVE.
12756 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
12758 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
12759 (@mve_vrhaddq_<supf><mode): Likewise.
12760 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
12761 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
12762 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
12764 2021-06-09 imba-tjd <109224573@qq.com>
12766 * doc/invoke.texi: Fix typo.
12768 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
12770 PR middle-end/53267
12771 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
12772 Support evaluation of fmod/fmodf/fmodl at compile-time.
12774 2021-06-09 Richard Biener <rguenther@suse.de>
12776 PR tree-optimization/100981
12777 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
12778 gimple_get_lhs to also handle calls.
12779 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
12782 2021-06-09 Richard Biener <rguenther@suse.de>
12784 PR tree-optimization/97832
12785 * tree-vectorizer.h (_slp_tree::failed): New.
12786 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
12788 (_slp_tree::~_slp_tree): Free failed.
12789 (vect_build_slp_tree): Retain failed nodes and record
12790 matches in them, copying that back out when running
12791 into a cached fail. Dump start and end of discovery.
12792 (dt_sort_cmp): New.
12793 (vect_build_slp_tree_2): Handle associatable chains
12794 together doing more aggressive operand swapping.
12796 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
12799 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
12801 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
12802 Linux and GNU targets.
12804 2021-06-09 Richard Biener <rguenther@suse.de>
12806 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
12809 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
12811 * config/arc/arc.md (loop_end): Change it to
12812 define_insn_and_split.
12814 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
12816 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
12817 (machi): New pattern.
12818 (umaddhisi4): Use VMAC2HU instruction.
12819 (umachi): New pattern.
12821 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
12823 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
12824 * config/arc/arc.c (arc_split_move_p): New function.
12825 (arc_split_move): Clean up.
12826 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
12827 (movdf_insn): Likewise.
12828 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
12830 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
12833 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
12834 argument to "raw". Do not emit segment overrides when "raw" is true.
12836 2021-06-09 Martin Liska <mliska@suse.cz>
12838 * doc/gcov.texi: Create a proper JSON files.
12839 * doc/invoke.texi: Remove dots in order to make it a valid
12842 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
12844 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
12845 (insn_is_load_p): Use pattern_is_rotate64.
12846 (insn_is_swap_p): Likewise.
12847 (quad_aligned_load_p): Likewise.
12848 (const_load_sequence_p): Likewise.
12849 (replace_swapped_aligned_load): Likewise.
12850 (recombine_lvx_pattern): Likewise.
12851 (recombine_stvx_pattern): Likewise.
12853 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
12855 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
12856 fur_stmt source record.
12857 * gimple-range.cc (fur_source::get_operand): Generic range query.
12858 (fur_source::get_phi_operand): New.
12859 (fur_source::register_dependency): New.
12860 (fur_source::query): New.
12861 (class fur_edge): New. Edge source for operands.
12862 (fur_edge::fur_edge): New.
12863 (fur_edge::get_operand): New.
12864 (fur_edge::get_phi_operand): New.
12865 (fur_edge::query): New.
12866 (fur_stmt::fur_stmt): New.
12867 (fur_stmt::get_operand): New.
12868 (fur_stmt::get_phi_operand): New.
12869 (fur_stmt::query): New.
12870 (class fur_depend): New. Statement source and process dependencies.
12871 (fur_depend::fur_depend): New.
12872 (fur_depend::register_dependency): New.
12873 (class fur_list): New. List source for operands.
12874 (fur_list::fur_list): New.
12875 (fur_list::get_operand): New.
12876 (fur_list::get_phi_operand): New.
12877 (fold_range): New. Instantiate appropriate fur_source class and fold.
12878 (fold_using_range::range_of_range_op): Use new API.
12879 (fold_using_range::range_of_address): Ditto.
12880 (fold_using_range::range_of_phi): Ditto.
12881 (imple_ranger::fold_range_internal): Use fur_depend class.
12882 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
12883 * gimple-range.h (class fur_source): Now a base class.
12884 (class fur_stmt): New.
12885 (fold_range): New prototypes.
12886 (fur_source::fur_source): Delete.
12888 2021-06-08 Andrew Pinski <apinski@marvell.com>
12890 PR tree-optimization/25290
12891 * tree-ssa-phiopt.c (xor_replacement): Delete.
12892 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
12893 (match_simplify_replacement): Allow one cheap preparation
12894 statement that can be moved to before the if.
12896 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
12898 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
12899 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
12901 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
12903 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
12904 Create length attribute on define_insn_and_split. Only split for cases which we
12906 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
12907 fix length computation.
12908 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
12910 2021-06-08 Richard Biener <rguenther@suse.de>
12912 PR tree-optimization/100923
12913 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
12914 the operand vector to be valueized.
12915 (valueize_refs): Likewise.
12916 (valueize_shared_reference_ops_from_ref): Adjust.
12917 (valueize_shared_reference_ops_from_call): Likewise.
12918 (vn_reference_lookup_3): Likewise.
12919 (vn_reference_lookup_pieces): Likewise. Re-valueize
12920 with honoring availability when we are about to create
12921 the ao_ref and valueized before.
12922 (vn_reference_lookup): Likewise.
12923 (vn_reference_insert_pieces): Adjust.
12925 2021-06-08 Richard Biener <rguenther@suse.de>
12927 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
12928 (_slp_instance::root_stmts): ... a vector.
12929 (SLP_INSTANCE_ROOT_STMT): Rename to ...
12930 (SLP_INSTANCE_ROOT_STMTS): ... this.
12931 (slp_root::root): Change to...
12932 (slp_root::roots): ... a vector.
12933 (slp_root::slp_root): Adjust.
12934 * tree-vect-slp.c (_slp_instance::location): Adjust.
12935 (vect_free_slp_instance): Release the root stmt vector.
12936 (vect_build_slp_instance): Adjust.
12937 (vect_analyze_slp): Likewise.
12938 (_bb_vec_info::~_bb_vec_info): Likewise.
12939 (vect_slp_analyze_operations): Likewise.
12940 (vect_bb_vectorization_profitable_p): Likewise. Adjust
12941 costs for the root stmt.
12942 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
12944 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
12946 (vectorize_slp_instance_root_stmt): Adjust.
12947 (vect_schedule_slp): Likewise.
12949 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
12951 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
12952 (ssa_equiv_stack::ssa_equiv_stack): New.
12953 (ssa_equiv_stack::~ssa_equiv_stack): New.
12954 (ssa_equiv_stack::enter): New.
12955 (ssa_equiv_stack::leave): New.
12956 (ssa_equiv_stack::push_replacement): New.
12957 (ssa_equiv_stack::get_replacement): New.
12958 (is_pointer_ssa): New.
12959 (class pointer_equiv_analyzer): New.
12960 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
12961 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
12962 (pointer_equiv_analyzer::set_global_equiv): New.
12963 (pointer_equiv_analyzer::set_cond_equiv): New.
12964 (pointer_equiv_analyzer::get_equiv): New.
12965 (pointer_equiv_analyzer::enter): New.
12966 (pointer_equiv_analyzer::leave): New.
12967 (pointer_equiv_analyzer::get_equiv_expr): New.
12968 (pta_valueize): New.
12969 (pointer_equiv_analyzer::visit_stmt): New.
12970 (pointer_equiv_analyzer::visit_edge): New.
12971 (hybrid_folder::value_of_expr): Call PTA.
12972 (hybrid_folder::value_on_edge): Same.
12973 (hybrid_folder::pre_fold_bb): New.
12974 (hybrid_folder::post_fold_bb): New.
12975 (hybrid_folder::pre_fold_stmt): New.
12976 (rvrp_folder::pre_fold_bb): New.
12977 (rvrp_folder::post_fold_bb): New.
12978 (rvrp_folder::pre_fold_stmt): New.
12979 (rvrp_folder::value_of_expr): Call PTA.
12980 (rvrp_folder::value_on_edge): Same.
12982 2021-06-08 Jakub Jelinek <jakub@redhat.com>
12985 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
12986 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
12988 2021-06-08 Richard Biener <rguenther@suse.de>
12990 PR middle-end/100951
12991 * tree-vect-generic.c (expand_vector_piecewise): Build a
12992 VECTOR_CST if all elements are constant.
12993 (expand_vector_condition): Likewise.
12994 (lower_vec_perm): Likewise.
12995 (expand_vector_conversion): Likewise.
12997 2021-06-08 Martin Liska <mliska@suse.cz>
12999 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13001 2021-06-08 Martin Liska <mliska@suse.cz>
13003 * genautomata.c (create_automata): Fix typo.
13005 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13007 PR tree-optimization/100794
13008 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13009 allow_unroll_p and only allow unrolling when it's true.
13010 (tree_predictive_commoning): Add parameter allow_unroll_p and
13012 (run_tree_predictive_commoning): Likewise.
13013 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13014 global_options_set.x_flag_predictive_commoning.
13015 (pass_predcom::execute): Adjust for allow_unroll_p.
13017 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13019 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13020 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13021 lambda function cleanup, remove scev_reset call, and adjust return
13023 (tree_predictive_commoning): Adjust for different changed values,
13024 only set flag TODO_update_ssa_only_virtuals if changed.
13025 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13026 from todo_flags_finish.
13028 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13030 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13031 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13032 (sbr_sparse_bitmap::bitmap_set_quad): New.
13033 (sbr_sparse_bitmap::bitmap_get_quad): New.
13034 (sbr_sparse_bitmap::set_bb_range): New.
13035 (sbr_sparse_bitmap::get_bb_range): New.
13036 (sbr_sparse_bitmap::bb_range_p): New.
13037 (block_range_cache::block_range_cache): initialize bitmap obstack.
13038 (block_range_cache::~block_range_cache): Destruct obstack.
13039 (block_range_cache::set_bb_range): Decide when to utilze the
13040 sparse on entry cache.
13041 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13042 * params.opt (-param=evrp-sparse-threshold): New.
13044 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13046 * bitmap.c (bitmap_set_aligned_chunk): New.
13047 (bitmap_get_aligned_chunk): New.
13048 (test_aligned_chunk): New.
13049 (bitmap_c_tests): Call test_aligned_chunk.
13050 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13052 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13055 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13057 (ix86_expand_vector_init_one_nonzero): Ditto.
13058 (ix86_expand_vector_init_one_var): Ditto.
13059 (ix86_expand_vector_init_general): Ditto.
13060 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13062 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13064 * config/h8300/movepush.md: Change most _clobber_flags
13065 patterns to instead use <cczn> subst.
13066 (movsi_cczn): New pattern with usable CC cases split out.
13067 (movsi_h8sx_cczn): Likewise.
13069 2021-06-07 Martin Liska <mliska@suse.cz>
13071 * common/common-target.def: Split long lines and replace them
13073 * target.def: Likewise.
13074 * doc/tm.texi: Re-generated.
13076 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13079 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13080 read from a CONSTRUCTOR with vector type elements.
13082 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13084 PR middle-end/100898
13085 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13086 should copy any arguments. Don't call gimple_call_num_args
13087 on id->call_stmt or call_stmt more than once.
13089 2021-06-07 liuhongt <hongtao.liu@intel.com>
13092 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13094 (<insn>v4siv4di2): Delete constraints for define_expand.
13096 2021-06-07 liuhongt <hongtao.liu@intel.com>
13099 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13100 assignment of cfun->machine->has_explicit_vzeroupper.
13101 * config/i386/i386-features.c
13102 (ix86_add_reg_usage_to_vzerouppers): Delete.
13103 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13104 (rest_of_handle_insert_vzeroupper): Remove
13105 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13107 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13108 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13110 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13111 (ix86_initialize_callee_abi): Ditto.
13112 (ix86_expand_avx_vzeroupper): Ditto.
13113 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13115 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13116 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13118 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13119 has_explicit_vzeroupper.
13120 * config/i386/i386.md (enum unspec): New member
13122 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13123 define_constants for insn callee abi index.
13124 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13125 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13126 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13127 (*avx_vzeroupper): Rename to ..
13128 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13129 call_insn which has a special vzeroupper ABI.
13130 (*avx_vzeroupper_1): Deleted.
13132 2021-06-07 liuhongt <hongtao.liu@intel.com>
13135 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13136 it won't use stack pointer reg.
13137 * final.c (leaf_function_p): When call_insn is a fake call, it
13138 won't affect caller as a leaf function.
13139 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13140 (subst_stack_regs): When call_insn doesn't clobber any stack
13141 reg, don't clear the arguments.
13142 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13144 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13145 frame for a fake call.
13146 * rtl.h (FAKE_CALL_P): New macro.
13148 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13150 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13152 (sparc_order_regs_for_local_alloc): ...this.
13153 (sparc_leaf_reg_remap): Declare.
13154 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13155 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13156 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13157 (order_regs_for_local_alloc): Rename to...
13158 (sparc_order_regs_for_local_alloc): ...this.
13159 (sparc_leaf_reg_remap): New function.
13160 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13162 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13164 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13165 Use assemble_name to output BSS section name.
13167 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13169 * config/i386/constraints.md (Bs):
13170 Remove boolean operators from match_test RTX.
13173 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13176 2021-06-06 Martin Liska <mliska@suse.cz>
13178 * doc/extend.texi: Add missing @headitem.
13179 * doc/invoke.texi: Likewise.
13180 * doc/objc.texi: Likewise.
13182 2021-06-06 Martin Liska <mliska@suse.cz>
13184 * genhooks.c (emit_findices): Remove unused function.
13185 (emit_documentation): Do not call emit_findices
13186 and do not search for @Fcode directives.
13188 2021-06-06 Martin Liska <mliska@suse.cz>
13190 * doc/invoke.texi: Remove extra character.
13192 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13194 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13196 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13198 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13199 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13200 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13201 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13203 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13205 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13207 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13209 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13212 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13214 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13215 *zero_extendsidi2): Fix empty split condition.
13217 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13219 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13221 * config/h8300/bitfield.md: Likewise.
13222 * config/h8300/combiner.md: Likewise.
13223 * config/h8300/divmod.md: Likewise.
13224 * config/h8300/extensions.md: Likewise.
13225 * config/h8300/jumpcall.md: Likewise.
13226 * config/h8300/movepush.md: Likewise.
13227 * config/h8300/multiply.md: Likewise.
13228 * config/h8300/other.md: Likewise.
13229 * config/h8300/shiftrotate.md: Likewise.
13230 * config/h8300/logical.md: Likewise. Fix split pattern to use
13231 code iterator that somehow slipped through.
13233 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13235 PR middle-end/100905
13236 * tree-nested.c (convert_nonlocal_omp_clauses,
13237 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13239 2021-06-04 Martin Sebor <msebor@redhat.com>
13241 PR middle-end/100732
13242 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13243 with either source or destination argument of invalid type.
13244 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13245 calls with arguments of invalid type.
13247 2021-06-04 Martin Sebor <msebor@redhat.com>
13249 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13251 (attr_access::vla_bounds): Also handle VLA bounds.
13253 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13255 * config/i386/predicates.md (GOT_memory_operand):
13256 Implement using match_code RTXes.
13257 (GOT32_symbol_operand): Ditto.
13259 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13262 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13264 (ix86_expand_vector_init_general): Ditto.
13265 Use SImode instead of word_mode for logic operations
13266 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13267 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13268 implemented by expand_vec_perm_1.
13269 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13270 should be implemented using standard shuffle patterns.
13271 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13272 V2HI modes to modes, implementable with shuffle for one operand.
13273 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13274 (*pshufw_1): New insn pattern.
13275 (*vec_dupv2hi): Ditto.
13276 (vec_initv2hihi): New expander.
13278 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13280 * config/arm/vfp.md (no_literal_pool_df_immediate,
13281 no_literal_pool_sf_immediate): Fix empty split condition.
13283 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13285 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13286 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13287 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13288 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13289 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13291 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13294 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13295 concatenation from half-sized modes with TImode elements.
13297 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13299 * config/arc/arc.c (arc_override_options): Disable millicode
13300 thunks when RF16 is on.
13302 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13304 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13306 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13308 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13309 Replace PROMOTE_MODE marco with its content.
13311 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13313 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13315 2021-06-03 Jim Wilson <jimw@sifive.com>
13317 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13318 turn it on for all riscv targets.
13320 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13323 * config/i386/i386-expand.c (ix86_expand_vector_set):
13324 Handle V2HI and V4QI modes.
13325 (ix86_expand_vector_extract): Ditto.
13326 * config/i386/mmx.md (*pinsrw): New insn pattern.
13329 (*pextrw_zext): Ditto.
13331 (*pextrb_zext): Ditto.
13332 (vec_setv2hi): New expander.
13333 (vec_extractv2hihi): Ditto.
13334 (vec_setv4qi): Ditto.
13335 (vec_extractv4qiqi): Ditto.
13336 (vec_setv8qi): Enable only for TARGET_SSE4_1.
13337 (vec_extractv8qiqi): Ditto.
13339 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
13341 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
13342 order to subf instruction.
13343 * config/rs6000/fusion.md: Regenerate.
13345 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13347 * calls.c (get_size_range): Use range_of_expr instead of
13348 determine_value_range.
13349 * tree-affine.c (expr_to_aff_combination): Same.
13350 * tree-data-ref.c (split_constant_offset): Same.
13351 * tree-vrp.c (determine_value_range_1): Remove.
13352 (determine_value_range): Remove.
13353 * tree-vrp.h (determine_value_range): Remove.
13355 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13357 * function-tests.c (test_ranges): Call gimple_range_tests.
13358 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
13360 * gimple-range.cc (fur_source::get_operand): Do not call
13361 get_tree_range or gimple_range_global.
13363 (get_tree_range): Move to value-query.cc.
13364 Call get_arith_expr_range.
13365 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
13366 Include gimple-range-tests.cc.
13367 * gimple-range.h (fold_range): Add argument.
13368 (get_tree_range): Remove.
13369 * selftest.h (gimple_range_tests): New.
13370 * value-query.cc (global_range_query::range_of_expr): Add
13372 (range_query::get_tree_range): Move from gimple-range.cc.
13373 * value-query.h (class range_query): Add get_tree_range and
13374 get_arith_expr_range. Make fur_source a friend.
13375 * vr-values.c (vr_values::range_of_expr): Pass stmt to
13377 * gimple-range-tests.cc: New file.
13379 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13381 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
13382 update_global_range.
13383 * value-query.cc (update_global_range): New.
13384 * value-query.h (update_global_range): New.
13386 2021-06-03 David Malcolm <dmalcolm@redhat.com>
13388 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
13389 printing the same location twice if there are fix-it hints,
13390 multiple locations, or a label.
13392 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
13394 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
13395 thresholds to narrow the upper bound on epilogue iterations.
13397 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
13399 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
13400 (mve_vabsq_s<mode>): Likewise.
13401 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
13402 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
13403 * config/arm/vec-common.md (neg<mode>2): Rename to
13404 <absneg_str><mode>2.
13406 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
13408 * common/config/arc/arc-common.c (arc_option_optimization_table):
13409 Remove malign-call.
13410 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
13411 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
13412 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
13413 * config/arc/arc.md (abssi2_mixed): Remove pattern.
13414 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
13415 (malign-call): Likewise.
13416 (mmixed-code): Likewise.
13417 * doc/invoke.texi (ARC): Update doc.
13419 2021-06-03 Martin Liska <mliska@suse.cz>
13421 * common.opt: Use proper Enum values.
13422 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
13423 (parse_sanitizer_options): Handle only sanitizer_opts.
13424 (common_handle_option): Just assign value.
13426 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13429 * tree-inline.c (inline_forbidden_p): Remove test on return type.
13431 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13433 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
13434 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
13435 (type_byte_size): Inline into...
13436 (add_byte_size_attribute): ...this and call add_scalar_info.
13438 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13440 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
13441 (typed_binop_from_tree): New function.
13442 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
13443 turn a divide by a power of 2 into a shift.
13444 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
13445 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
13446 typed divide by calling typed_binop_from_tree.
13448 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13450 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
13451 (is_handled_procedure_type): Likewise.
13452 (struct loc_descr_context): Add strict_signedness field.
13453 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
13454 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
13455 (resolve_args_picking): Minor tweak.
13456 (function_to_dwarf_procedure): Initialize strict_signedness field.
13457 (type_byte_size): Likewise.
13458 (field_byte_offset): Likewise.
13459 (gen_descr_array_type_die): Likewise.
13460 (gen_variant_part): Likewise.
13461 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
13462 to true when a context is present before evaluating the arguments.
13463 <COND_EXPR>: Do not generate a useless comparison with zero.
13464 When dereferencing an address, if strict_signedness is true and the
13465 type is small and signed, use DW_OP_deref_type to do the dereference
13466 and then DW_OP_convert to convert back to the generic type.
13468 2021-06-03 Jakub Jelinek <jakub@redhat.com>
13471 * tree-inline.c (copy_tree_body_r): Handle iterators on
13472 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
13474 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13476 * config/arc/arc.md (*bbit_di): Remove.
13478 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
13480 PR rtl-optimization/100264
13481 * ree.c (get_sub_rtx): Ignore SET expressions without register
13482 destinations and remove assertion, as it is not valid anymore
13483 with this new behaviour.
13484 (merge_def_and_ext): Eliminate destination check for register
13485 as such SET expressions can't occur anymore.
13486 (combine_reaching_defs): Likewise.
13488 2021-06-02 Jakub Jelinek <jakub@redhat.com>
13491 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
13492 -Wtype-limits warnings.
13493 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
13494 in operands to avoid -Wsign-compare warnings.
13496 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
13498 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
13501 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
13503 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
13505 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
13507 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
13509 * config/s390/subst.md(cconly_subst): Use a single constraint
13510 in (match_scratch).
13512 2021-06-02 Martin Liska <mliska@suse.cz>
13514 * ipa-icf.h: Use auto_vec for memory_access_types.
13516 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
13518 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
13519 argument from prototype.
13520 (output_logical_op): Add rtx_code argument.
13521 (compute_logical_op_length): Likewise.
13522 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
13523 to compute_a_shift_length.
13524 (output_logical_op); New argument with the rtx code rather than
13525 extracting it from an operand. Handle QImode too.
13526 (compute_logical_op_length): Similary.
13527 (compute_a_shift_length): Drop unused argument.
13528 * config/h8300/h8300.md (logicals): New code iterator.
13529 * config/h8300/logical.md (<code><mode>3 expander): Combine
13530 the "and" expander with the "ior"/"xor" expander.
13531 (bclr<mode>msx): Combine the QI/HI mode patterns.
13532 (<logical><mode>3 insns): Use code iterator rather than match_operator.
13533 Handle QImode as well. Update call to output_logical_op and
13534 compute_logical_op_length to pass in rtx_code
13535 Fix split condition on all define_insn_and_split patterns.
13536 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
13537 the flags and setting ZN via existing define_subst.
13538 * config/h8300/shiftrotate.md: Drop unused argument from
13539 calls to compute_a_shift_length.
13540 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
13542 2021-06-01 Andrew Pinski <apinski@marvell.com>
13544 PR tree-optimization/25290
13545 * tree-ssa-phiopt.c (match_simplify_replacement):
13547 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
13548 (two_value_replacement): Change the comment about
13549 conditional_replacement.
13550 (conditional_replacement): Delete.
13552 2021-06-01 Andrew Pinski <apinski@marvell.com>
13554 PR tree-optimization/95481
13555 * tree-tailcall.c (find_tail_calls): Handle empty typed
13558 2021-06-01 Andrew Pinski <apinski@marvell.com>
13560 * gimplify.c (zero_sized_field_decl): Delete
13561 (zero_sized_type): Delete
13562 (gimplify_init_ctor_eval): Use is_empty_type instead
13563 of zero_sized_field_decl.
13564 (gimplify_modify_expr): Use is_empty_type instead of
13567 2021-06-01 Jason Merrill <jason@redhat.com>
13570 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
13572 2021-06-01 Jason Merrill <jason@redhat.com>
13575 * diagnostic.h (warning_enabled_at): Declare.
13576 * diagnostic.c (diagnostic_enabled): Factor out from...
13577 (diagnostic_report_diagnostic): ...here.
13578 (warning_enabled_at): New.
13580 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
13582 * gimple-ssa-evrp.c: Enable exporting of global ranges.
13584 2021-06-01 Martin Liska <mliska@suse.cz>
13587 * doc/invoke.texi: Mention that -fgcse-after-reload
13588 is enabled with -O3.
13590 2021-06-01 liuhongt <hongtao.liu@intel.com>
13592 PR tree-optimization/98365
13593 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
13594 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
13595 (convert_scalar_cond_reduction): Ditto.
13596 (predicate_scalar_phi): Ditto.
13598 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13600 PR tree-optimization/100781
13601 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
13602 value calculation by default.
13603 (ranger_cache::enable_new_values): New.
13604 (ranger_cache::disable_new_values): New.
13605 (ranger_cache::push_poor_value): Check if new values are allowed.
13606 * gimple-range-cache.h (class ranger_cache): New member/methods.
13607 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
13608 statement, and disable/renable new value calculation.
13610 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13612 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
13613 (ranger_cache::range_of_def): New.
13614 (ranger_cache::entry_range): New.
13615 (ranger_cache::exit_range): New.
13616 (ranger_cache::range_of_expr): Adjust.
13617 (ranger_cache::range_on_edge): Adjust.
13618 (ranger_cache::propagate_cache): Call exit_range directly.
13619 * gimple-range-cache.h (class ranger_cache): Adjust.
13621 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13623 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
13624 gori_compute being a member rather than base class.
13625 dervied call to member call.
13626 (ranger_cache::dump): No longer dump gori_map.
13627 (ranger_cache::dump_bb): New.
13628 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
13629 being a member rather than base class.
13630 (ranger_cache::set_global_range): Ditto.
13631 (ranger_cache::ssa_range_in_bb): Ditto.
13632 (ranger_cache::range_of_expr): New.
13633 (ranger_cache::range_on_edge): New.
13634 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
13635 (ranger_cache::propagate_cache): Adjust debugging output.
13636 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
13638 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
13639 member, and inherit from range_query instead.
13640 (ranger_cache::dump_bb): New. split from dump.
13641 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
13642 (gori_compute::expr_range_at_stmt): Delete.
13643 (gori_compute::compute_name_range_op): Delete.
13644 (gori_compute::compute_operand_range_switch): Add fur_source.
13645 (gori_compute::compute_operand_range): Add fur_source param, inline
13646 old compute_name_range_op and optimize_logical_operands.
13647 (struct tf_range): Delete.
13648 (gori_compute::logical_combine): Adjust
13649 (gori_compute::optimize_logical_operands): Delete.
13650 (gori_compute::compute_logical_operands_in_chain): Delete.
13651 (gori_compute::compute_logical_operands): Adjust.
13652 (gori_compute::compute_operand1_range): Adjust to fur_source.
13653 (gori_compute::compute_operand2_range): Ditto.
13654 (gori_compute::compute_operand1_and_operand2_range): Ditto.
13655 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
13656 and adjust to fur_source.
13657 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
13658 range_query and fur_source.
13659 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
13660 from the ranger_cache..
13661 (gimple_ranger::fold_range_internal): Adjust to base class change of
13663 (gimple_ranger::dump_bb): Adjust dump.
13664 * gimple-range.h (gimple_ranger):export gori computes object.
13666 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13668 PR tree-optimization/100774
13669 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
13670 Constant values are also not stale.
13671 (ranger_cache::set_global_range): Range invariant values should also
13672 have the correct timestamp.
13674 2021-05-31 Martin Liska <mliska@suse.cz>
13676 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
13677 Unpack FUNCTION_DECL_DECL_TYPE.
13678 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
13679 Stream FUNCTION_DECL_DECL_TYPE instead of
13680 DECL_IS_OPERATOR_NEW_P.
13681 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
13683 (DECL_IS_OPERATOR_NEW_P): Likewise.
13684 (DECL_IS_OPERATOR_DELETE_P): Likewise.
13685 (DECL_LAMBDA_FUNCTION_P): Likewise.
13687 2021-05-31 Richard Biener <rguenther@suse.de>
13690 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
13691 * internal-fn.def (SHUFFLEVECTOR): New.
13692 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
13693 * doc/extend.texi: Document __builtin_shufflevector.
13695 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
13698 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
13699 indexed form addresses.
13701 2021-05-29 Jeff Law <jlaw@tachyum.com>
13703 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
13704 parameter. Call callers fixed.
13706 (output_plussi): Add FALLTHRU markers.
13707 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
13709 2021-05-29 Jakub Jelinek <jakub@redhat.com>
13711 PR middle-end/99928
13712 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
13713 combined with parallel, make sure to add shared clause to
13714 parallel for explicit linear clause.
13716 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
13718 PR tree-optimization/100787
13719 * gimple-ssa-evrp.c: Disable exporting of global ranges.
13721 2021-05-28 Jason Merrill <jason@redhat.com>
13723 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
13724 operator--, operator*, operator==, and operator!=.
13725 (class tsi_range): New.
13727 2021-05-28 Richard Biener <rguenther@suse.de>
13729 PR tree-optimization/100778
13730 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
13731 trapping ops in different BBs.
13733 2021-05-28 Richard Biener <rguenther@suse.de>
13736 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
13737 copy fntype from original call.
13739 2021-05-28 Martin Liska <mliska@suse.cz>
13741 PR gcov-profile/100751
13742 * doc/gcov.texi: Revert partially a hunk that was wrong.
13744 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
13746 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
13748 (HAVE_sync_compare_and_swaphi): Likewise.
13749 (HAVE_sync_compare_and_swapsi): Likewise.
13751 2021-05-28 Jakub Jelinek <jakub@redhat.com>
13753 PR middle-end/99928
13754 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
13756 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
13758 * gimplify.c (gimplify_omp_affinity): New.
13759 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
13760 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
13761 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
13762 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
13763 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
13765 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
13766 Richard Biener <rguenther@suse.de>
13768 * match.pd <popcount & / + pattern matching>:
13769 When generating popcount directly fails, try doing it in two halves.
13771 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
13773 * Makefile.in (generated_files): Add gimple-match.c and
13776 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
13778 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
13780 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
13782 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
13784 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
13786 PR tree-optimization/99398
13787 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
13788 where the fed operands are CTOR/CST and propagated through
13789 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
13790 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
13792 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
13795 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
13797 * config/i386/mmx.md (addv2sf3): Do not call
13798 ix86_fixup_binary_operands_no_copy.
13801 (<smaxmin:code>v2sf3): Ditto.
13802 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
13803 (<plusminus:insn><VI_32:mode>3): Remove expander.
13804 (<plusminus:insn><VI_32:mode>3): Rename from
13805 "*<plusminus:insn><VI_32:mode>3".
13806 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
13807 (mulv2hi3): Remove expander.
13808 (mulv2hi3): Rename from *mulv2hi3.
13809 (<s>mulv2hi3_highpart): Remove expander.
13810 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
13811 (<smaxmin:code><MMXMODE14:mode>3): Rename from
13812 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
13813 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
13814 (SMAXMIN_MMXMODEI): Remove mode iterator.
13815 (<smaxmin:code>v4hi3): New expander.
13816 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
13817 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
13818 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
13819 (SMAXMIN_VI_32): Remove mode iterator.
13820 (<umaxmin:code><MMXMODE24:mode>3): Rename from
13821 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
13822 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
13823 (UMAXMIN_MMXMODEI): Remove mode iterator.
13824 (<umaxmin:code>v8qi3): New expander.
13825 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
13826 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
13827 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
13828 (UMAXMIN_VI_32): Remove mode iterator.
13829 (<any_shift:insn>v2hi3): Remove expander.
13830 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
13831 (<any_logic:code><MMXMODEI:mode>3): Do not call
13832 ix86_fixup_binary_operands_no_copy.
13833 (<any_logic:code><VI_32:mode>3): Remove expander.
13834 (<any_logic:code><VI_32:mode>3): Rename from
13835 "*<any_logic:code><VI_32:mode>3".
13836 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
13837 * config/i386/sse.md (div<VF2:mode>3): Do not call
13838 ix86_fixup_binary_operands_no_copy.
13839 (div<VF1:mode>3): Ditto.
13840 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
13841 (smulhrsv4hi3): Ditto.
13842 (smulhrsv2hi3): Ditto.
13844 2021-05-27 Martin Sebor <msebor@redhat.com>
13846 * ggc.h (gt_ggc_mx): Add overloads for all integers.
13848 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
13850 (hash_map::operator==): New function.
13852 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
13855 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
13856 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
13857 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
13858 (*xop_maskcmp<VI_32:mode>3): Ditto.
13859 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
13860 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
13862 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
13865 * config/arm/arm.c (arm_configure_build_target): Remove parameter
13866 opts_set, directly check opts parameters for being non-null.
13867 (arm_option_restore): Update call to arm_configure_build_target.
13868 (arm_option_override): Likewise.
13869 (arm_can_inline_p): Likewise.
13870 (arm_valid_target_attribute_tree): Likewise.
13871 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
13872 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
13875 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
13877 * vr-values.c (simplify_conversion_using_ranges): Use
13878 get_range_query instead of get_global_range_query.
13880 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
13882 * gimple-range.cc (get_range_global): Move to value-query.cc.
13883 (gimple_range_global): Same.
13884 (get_global_range_query): Same.
13885 (global_range_query::range_of_expr): Same.
13886 * gimple-range.h (class global_range_query): Move to
13888 (gimple_range_global): Same.
13889 * tree-ssanames.c (get_range_info): Move to value-query.cc.
13890 (get_ptr_nonnull): Same.
13891 * tree-ssanames.h (get_range_info): Remove.
13892 (get_ptr_nonnull): Remove.
13893 * value-query.cc (get_ssa_name_range_info): Move from
13895 (get_ssa_name_ptr_info_nonnull): Same.
13896 (get_range_global): Move from gimple-range.cc.
13897 (gimple_range_global): Same.
13898 (get_global_range_query): Same.
13899 (global_range_query::range_of_expr): Same.
13900 * value-query.h (class global_range_query): Move from
13902 (gimple_range_global): Same.
13904 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
13907 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
13908 (uavgv2hi3_ceil): Ditto.
13910 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
13913 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
13915 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
13917 * tree-ssanames.c (get_range_info): Merge both copies of
13918 get_range_info into one that works with irange.
13919 * tree-ssanames.h (get_range_info): Remove version that works on
13922 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
13924 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
13925 (expand_builtin_strnlen): Same.
13926 (determine_block_size): Same.
13927 * fold-const.c (expr_not_equal_to): Same.
13928 * gimple-fold.c (size_must_be_zero_p): Same.
13929 * gimple-match-head.c: Include gimple-range.h.
13930 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
13931 * gimple-ssa-warn-restrict.c
13932 (builtin_memref::extend_offset_range): Same.
13933 * graphite-sese-to-poly.c (add_param_constraints): Same.
13934 * internal-fn.c (get_min_precision): Same.
13935 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
13936 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
13938 * tree-data-ref.c (split_constant_offset): Same.
13939 (dr_step_indicator): Same.
13940 * tree-dfa.c (get_ref_base_and_extent): Same.
13941 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13942 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
13943 (determine_value_range): Same.
13944 (record_nonwrapping_iv): Same.
13945 (infer_loop_bounds_from_signedness): Same.
13946 (scev_var_range_cant_overflow): Same.
13947 * tree-ssa-phiopt.c (two_value_replacement): Same.
13948 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
13949 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
13950 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
13952 (dump_strlen_info): Same.
13953 (set_strlen_range): Same.
13954 (maybe_diag_stxncpy_trunc): Same.
13955 (get_len_or_size): Same.
13956 (handle_integral_assign): Same.
13957 * tree-ssa-structalias.c (find_what_p_points_to): Same.
13958 * tree-ssa-uninit.c (find_var_cmp_const): Same.
13959 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
13960 * tree-vect-patterns.c (vect_get_range_info): Same.
13961 (vect_recog_divmod_pattern): Same.
13962 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
13963 (register_edge_assert_for_2): Same.
13964 (determine_value_range_1): Same.
13965 * tree.c (get_range_pos_neg): Same.
13966 * vr-values.c (vr_values::get_lattice_entry): Same.
13967 (vr_values::update_value_range): Same.
13968 (simplify_conversion_using_ranges): Same.
13970 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
13972 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
13973 get_range_query instead of query argument.
13974 (pass_walloca::execute): Enable and disable global ranger.
13976 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
13978 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
13980 (rvrp_folder::~rvrp_folder): Call disable_ranger.
13981 (hybrid_folder::hybrid_folder): Call enable_ranger.
13982 (hybrid_folder::~hybrid_folder): Call disable_ranger.
13984 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
13986 * function.c (allocate_struct_function): Set cfun->x_range_query.
13987 * function.h (struct function): Declare x_range_query.
13988 (get_range_query): New.
13989 (get_global_range_query): New.
13990 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
13991 Remove call to safe_grow_cleared.
13992 * gimple-range.cc (get_range_global): New.
13993 (gimple_range_global): Move from gimple-range.h.
13994 (get_global_range_query): New.
13995 (global_range_query::range_of_expr): New.
13996 (enable_ranger): New.
13997 (disable_ranger): New.
13998 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
13999 (class global_range_query): New.
14000 (enable_ranger): New.
14001 (disable_ranger): New.
14002 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14003 dump_all_value_ranges to dump.
14004 * tree-vrp.c (vrp_prop::finalize): Same.
14005 * value-query.cc (range_query::dump): New.
14006 * value-query.h (range_query::dump): New.
14007 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14008 (vr_values::dump): ...this.
14009 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14010 dump and make virtual.
14012 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14014 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14015 Add V4QImode and V16QImode for TARGET_SSE2.
14016 * doc/sourcebuild.texi (Vector-specific attributes):
14017 Add vect64 and vect32 description.
14019 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14021 * gimple-range-gori.cc (range_def_chain::register_dependency):
14022 Resize m_def_chain when needed.
14024 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14026 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14027 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14028 * config/arm/vec-common.md: .. here. Add support for MVE.
14030 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14032 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14034 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14036 microblaze_option_override, print_operand): Likewise.
14037 * config/microblaze/microblaze.md (call_internal_plt,
14038 call_value_intern_plt, call_value_intern): Likewise.
14039 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14040 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14041 call_value_multiple_internal1): Likewise.
14042 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14044 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14046 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14047 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14048 arc_raw_symbolic_reference_mentioned_p): Remove register
14051 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14054 * omp-low.c: Include omp-offload.h.
14055 (create_omp_child_function): If current_function_decl has
14056 "omp declare target" attribute and is_gimple_omp_offloaded,
14057 remove that attribute from the copy of attribute list and
14058 add "omp target entrypoint" attribute instead.
14059 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14060 variables for offloading if in omp_maybe_offloaded_ctx.
14061 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14062 argument to GOMP_target_data_ext in offloaded code.
14064 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14066 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14067 For csky, HF/SF mode use the low bits of VREGS.
14069 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14071 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14072 DECL which is a reference for OMP.
14074 2021-05-26 Martin Liska <mliska@suse.cz>
14076 PR gcov-profile/100751
14077 * doc/gcov.texi: Document that __gcov_dump can be called just
14078 once and that __gcov_reset resets run-time counters.
14080 2021-05-26 Martin Liska <mliska@suse.cz>
14082 * doc/install.texi: Port relevant part from install-old.texi
14083 and re-generate list of CPUs and systems.
14085 2021-05-26 Martin Liska <mliska@suse.cz>
14087 * Makefile.in: Remove it.
14088 * doc/include/fdl.texi: Update next/previous chapters.
14089 * doc/install.texi: Likewise.
14090 * doc/install-old.texi: Removed.
14092 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14094 * config/csky/csky.c (ck810_legitimate_index_p): Support
14095 "base + index" with DF mode.
14096 * config/csky/constraints.md ("Y"): New constraint for memory operands
14097 without index register.
14098 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14099 when mov between memory and general registers, and lower their priority.
14100 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14102 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14104 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14106 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14108 * config/csky/csky.md (untyped_call): Emit clobber for return
14109 registers to mark them used.
14111 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14113 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14115 2021-05-26 Andrew Pinski <apinski@marvell.com>
14117 * match.pd (x < 0 ? ~y : y): New patterns.
14119 2021-05-26 Andrew Pinski <apinski@marvell.com>
14121 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14122 A?POW2:0 and A?0:POW2.
14124 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14126 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14127 (logical_stmt_cache::logical_stmt_cache ): Delete.
14128 (logical_stmt_cache::~logical_stmt_cache): Delete.
14129 (logical_stmt_cache::cache_entry::dump): Delete.
14130 (logical_stmt_cache::get_range): Delete.
14131 (logical_stmt_cache::cached_name ): Delete.
14132 (logical_stmt_cache::same_cached_name): Delete.
14133 (logical_stmt_cache::cacheable_p): Delete.
14134 (logical_stmt_cache::slot_diagnostics ): Delete.
14135 (logical_stmt_cache::dump): Delete.
14136 (gori_compute_cache::gori_compute_cache): Delete.
14137 (gori_compute_cache::~gori_compute_cache): Delete.
14138 (gori_compute_cache::compute_operand_range): Delete.
14139 (gori_compute_cache::cache_stmt): Delete.
14140 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14142 (class gori_compute_cache): Delete.
14144 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14146 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14148 (fold_using_range::range_of_address): Adjust.
14149 (fold_using_range::range_of_phi): Adjust.
14150 * gimple-range.h (class fur_source): Adjust.
14151 (fur_source::fur_source): Adjust.
14153 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14155 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14156 from expr_range_in_bb and adjust.
14157 (gori_compute::compute_name_range_op): Adjust.
14158 (gori_compute::optimize_logical_operands): Adjust.
14159 (gori_compute::compute_logical_operands_in_chain): Adjust.
14160 (gori_compute::compute_operand1_range): Adjust.
14161 (gori_compute::compute_operand2_range): Adjust.
14162 (ori_compute_cache::cache_stmt): Adjust.
14163 * gimple-range-gori.h (gori_compute): Rename prototype.
14165 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14167 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14168 checked only after range_of_stmt, not range_on_entry.
14169 (gimple_ranger::range_on_entry): Check for non-null in any
14170 predecessor block, if it is not already non-null.
14171 (gimple_ranger::range_on_exit): DOnt check for non-null after
14172 range on entry call.
14173 (gimple_ranger::dump_bb): New. Split from dump.
14174 (gimple_ranger::dump): Adjust.
14175 * gimple-range.h (class gimple_ranger): Adjust.
14177 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14179 * gimple-range-cache.cc (struct range_timestamp): Delete.
14180 (class temporal_cache): Adjust.
14181 (temporal_cache::get_timestamp): Delete.
14182 (temporal_cache::set_dependency): Delete.
14183 (temporal_cache::temporal_value): Adjust.
14184 (temporal_cache::current_p): Take dependencies as params.
14185 (temporal_cache::set_timestamp): Adjust.
14186 (temporal_cache::set_always_current): Adjust.
14187 (ranger_cache::get_non_stale_global_range): Adjust.
14188 (ranger_cache::register_dependency): Delete.
14189 * gimple-range-cache.h (class range_cache): Adjust.
14191 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14193 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14195 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14196 each individual bitmap.
14197 (range_def_chain::set_import): New.
14198 (range_def_chain::get_imports): New.
14199 (range_def_chain::chain_import_p): New.
14200 (range_def_chain::register_dependency): Rename from build_def_chain
14202 (range_def_chain::def_chain_in_bitmap_p): New.
14203 (range_def_chain::add_def_chain_to_bitmap): New.
14204 (range_def_chain::has_def_chain): Just check first depenedence.
14205 (range_def_chain::get_def_chain): Process imports, use generic
14206 register_dependency routine.
14207 (range_def_chain::dump): New.
14208 (gori_map::gori_map): Allocate import list.
14209 (gori_map::~gori_map): Release imports.
14210 (gori_map::exports): Check for past allocated block size.
14211 (gori_map::imports): New.
14212 (gori_map::def_chain_in_export_p): Delete.
14213 (gori_map::is_import_p): New.
14214 (gori_map::maybe_add_gori): Handle imports.
14215 (gori_map::dump): Adjust output, add imports.
14216 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14217 (gori_export_iterator::gori_export_iterator): New.
14218 (gori_export_iterator::next): New.
14219 (gori_export_iterator::get_name): New.
14220 * gimple-range-gori.h (range_def_chain): Add imports and direct
14221 dependecies via struct rdc.
14222 (range_def_chain::depend1): New.
14223 (range_def_chain::depend2): New.
14224 (class gori_map): Adjust.
14225 (FOR_EACH_GORI_IMPORT_NAME): New.
14226 (FOR_EACH_GORI_EXPORT_NAME): New.
14227 (class gori_export_iterator): New.
14229 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14231 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14232 export cache filling to here.
14233 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14235 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14237 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14238 (gori_map): Move to gimple-range-gori.h.
14239 (gori_compute::gori_compute): Adjust.
14240 (gori_compute::~gori_compute): Delete.
14241 (gori_compute::compute_operand_range_switch): Adjust.
14242 (gori_compute::compute_operand_range): Adjust.
14243 (gori_compute::compute_logical_operands): Adjust.
14244 (gori_compute::has_edge_range_p ): Adjust.
14245 (gori_compute::set_range_invariant): Delete.
14246 (gori_compute::dump): Adjust.
14247 (gori_compute::outgoing_edge_range_p): Adjust.
14248 * gimple-range-gori.h (class range_def_chain): Relocate here.
14249 (class gori_map): Relocate here.
14250 (class gori_compute): Inherit from gori_map, and adjust.
14252 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14254 * value-range.cc (range_tests_legacy): Use
14255 build_nonstandard_integer_type instead of int and short.
14257 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14259 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14260 when really creating an initialization statement for it.
14262 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14264 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14266 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14268 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14270 2021-05-25 Martin Liska <mliska@suse.cz>
14272 PR tree-optimization/92860
14274 * optc-save-gen.awk: Remove exceptions.
14276 2021-05-25 Martin Liska <mliska@suse.cz>
14278 * asan.h (sanitize_coverage_p): New function.
14279 * doc/extend.texi: Document it.
14280 * fold-const.c (fold_range_test): Use sanitize_flags_p
14281 instead of flag_sanitize_coverage.
14282 (fold_truth_andor): Likewise.
14283 * sancov.c: Likewise.
14284 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14285 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14286 -fsanitize-coverage when inlining.
14288 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14290 * config/csky/csky-modes.def : Fix copyright.
14292 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14294 * config/csky/csky-modes.def : Amend copyright.
14295 * config/csky/csky_insn_fpuv2.md : Likewise.
14296 * config/csky/csky_insn_fpuv3.md : Likewise.
14298 2021-05-25 Richard Biener <rguenther@suse.de>
14300 PR middle-end/100727
14301 * calls.c (initialize_argument_information): Explicitely test
14302 for WITH_SIZE_EXPR.
14303 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14305 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14307 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14308 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14309 the signle definition. The signle definition may not work well
14310 at simplify_subreg_regno().
14311 (HARD_FRAME_POINTER_REGNUM): New.
14312 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14313 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14314 csky_initial_elimination_offset, csky_expand_prologue,
14315 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14317 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14319 * config/csky/csky.c (csky_option_override):
14320 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14321 and TARGET_DIV can be set well.
14323 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14325 * config/csky/constraints.md ("l", "h"): Delete.
14326 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14327 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14328 * config/csky/csky.c (regno_reg_classm,
14329 csky_secondary_reload, csky_register_move_cost):
14330 Use HILO_REGS instead of LO_REGS and HI_REGS.
14332 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14334 * config/csky/constraints.md ("W"): New constriant for mem operand
14335 with base reg, index register.
14336 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
14337 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
14339 ("Dv"): New constraint for const double value that can be used at
14341 * config/csky/csky-modes.def (HFmode): New mode.
14342 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
14343 to "csky_valid_mem_constraint_operand" and support new constraint
14345 (csky_get_movedouble_length): New.
14346 (fpuv3_output_move): New.
14347 (fpuv3_const_double): New.
14348 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
14349 (decompose_csky_address): Refine.
14350 (csky_print_operand): New "CONST_DOUBLE" operand.
14351 (csky_output_move): Support fpv3 instructions.
14352 (csky_get_movedouble_length): New.
14353 (fpuv3_output_move): New.
14354 (fpuv3_const_double): New.
14355 (csky_emit_compare): Cover float comparsion.
14356 (csky_emit_compare_float): Refine.
14357 (csky_vaild_fpuv2_mem_operand): Rename to
14358 "csky_valid_mem_constraint_operand" and support new constraint "W".
14359 (ck860_rtx_costs): New.
14360 (csky_rtx_costs): Add the cost calculation of CK860.
14361 (regno_reg_class): New vregs for fpuv3.
14362 (csky_dbx_regno): Likewise.
14363 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
14364 (csky_conditional_register_usage): Suporrot fpuv3.
14365 (csky_dwarf_register_span): Suporrot fpuv3.
14366 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
14367 (ck810_legitimate_index_p): Support fp16.
14368 * config/csky/csky.h (TARGET_TLS): ADD CK860.
14369 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
14370 (TARGET_SINGLE_FPU): Support fpuv3.
14371 (TARGET_SUPPORT_FPV3): New.
14372 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
14373 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
14374 REG_CLASS_CONTENTS): Support fpuv3.
14375 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
14376 (csky_movsf_fpv2): Likewise.
14377 (ck801_movsf): Likewise.
14378 (csky_movsf): Likewise.
14380 (csky_movdf_fpv2): Likewise.
14381 (ck801_movdf): Likewise.
14382 (csky_movdf): Likewise.
14383 (movsicc): Refine. Use "comparison_operatior" instead of
14384 "ordered_comparison_operatior".
14385 (addsicc): Likewise.
14386 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
14387 (call_value_internal_vh): New.
14388 * config/csky/csky_cores.def (CK860): New arch and cpu.
14393 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
14394 into emit-patterns and match-patterns, remain the emit-patterns here,
14395 and move the match-patterns to csky_insn_fpuv2.md or
14396 csky_insn_fpuv3.md.
14397 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
14398 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
14400 * config/csky/csky_isa.def (fcr): New.
14405 (CK860): New definition for ck860.
14406 * config/csky/csky_tables.opt (ck860): New processors ck860,
14407 ck860f. And new arch ck860.
14412 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
14413 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
14414 * config/csky/t-csky-elf: Support 860.
14415 * config/csky/t-csky-linux: Likewise.
14416 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
14418 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
14420 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
14421 add generation of logical-add and add-logical fusion pairs.
14422 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
14424 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
14425 logical-add and add-logical fusion by default.
14426 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
14427 -mpower10-fusion-add-logical options.
14428 * config/rs6000/fusion.md: Regenerate file.
14430 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
14432 * value-range.cc (irange::legacy_equal_p): Check type when
14433 comparing VR_VARYING types.
14434 (range_tests_legacy): Test comparing VARYING ranges of different
14437 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14439 * config/aarch64/aarch64.c (neoversen1_tunings):
14440 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
14442 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14444 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
14445 extern weak symbols. Limit symbol offsets for non-GOT symbols with
14448 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14450 * config/arm/neon.md (vec_load_lanesxi<mode>)
14451 (vec_store_lanexoi<mode>): Move ...
14452 * config/arm/vec-common.md: here.
14454 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14456 * config/arm/neon.md (vec_load_lanesoi<mode>)
14457 (vec_store_lanesoi<mode>): Move ...
14458 * config/arm/vec-common.md: here.
14460 2021-05-24 liuhongt <hongtao.liu@intel.com>
14463 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
14464 stmt with GIMPLE_NOP when lhs doesn't exist.
14466 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
14469 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
14470 New instruction pattern.
14471 (*push<VI_32:mode>2): Ditto.
14472 (push splitter for SSE registers): New splitter.
14474 2021-05-23 Andrew Pinski <apinski@marvell.com>
14476 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
14478 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
14480 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
14481 * config/rs6000/fusion.md: Regenerate file.
14483 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14485 * config/rs6000/genfusion.pl (gen_addadd): New function.
14486 * config/rs6000/fusion.md: Regenerate file.
14487 * config/rs6000/rs6000-cpus.def: Add
14488 OPTION_MASK_P10_FUSION_2ADD to masks.
14489 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14490 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
14491 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
14493 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14495 PR middle-end/99928
14496 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
14497 * gimplify.c (enum gimplify_omp_var_data): Fix up
14498 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
14499 (omp_lastprivate_for_combined_outer_constructs): If combined target
14500 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
14501 GOVD_MAP | GOVD_SEEN.
14502 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
14503 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
14504 (gimplify_adjust_omp_clauses): For firstprivate clauses with
14505 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
14506 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
14507 let it be replaced by implicit map clause.
14509 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14511 PR middle-end/99928
14512 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
14514 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
14515 (gimplify_omp_for): Likewise.
14517 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14519 PR middle-end/90115
14520 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
14521 'external' in blocks.
14523 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14525 PR middle-end/90115
14526 * flag-types.h (enum openacc_privatization): New.
14527 * params.opt (-param=openacc-privatization): New.
14528 * doc/invoke.texi (openacc-privatization): Document it.
14529 * omp-general.h (get_openacc_privatization_dump_flags): New
14531 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
14532 * omp-offload.c (execute_oacc_device_lower)
14533 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
14534 * target.def (goacc.adjust_private_decl): Add 'location_t'
14536 * doc/tm.texi: Regenerate.
14537 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
14538 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
14539 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
14540 Likewise. Preserve it for...
14541 (nvptx_goacc_expand_var_decl): ... use here.
14543 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14545 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
14548 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14550 PR middle-end/90115
14551 * omp-low.c (oacc_privatization_candidate_p): New function.
14552 (oacc_privatization_scan_clause_chain)
14553 (oacc_privatization_scan_decl_chain): Use it. Also
14554 'gcc_checking_assert' that we're not seeing duplicates.
14556 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14558 PR middle-end/90115
14559 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
14562 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14564 PR middle-end/90115
14565 * omp-offload.c (execute_oacc_device_lower): Explain.
14567 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14569 PR middle-end/90115
14570 * omp-offload.c (execute_oacc_device_lower)
14571 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
14573 * internal-fn.c (expand_UNIQUE): Don't expect
14574 'IFN_UNIQUE_OACC_PRIVATE'.
14576 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14578 PR middle-end/90115
14579 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
14581 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14583 PR middle-end/90115
14584 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
14585 (nvptx_goacc_expand_var_decl): Tighten.
14587 2021-05-21 Julian Brown <julian@codesourcery.com>
14588 Chung-Lin Tang <cltang@codesourcery.com>
14589 Thomas Schwinge <thomas@codesourcery.com>
14591 PR middle-end/90115
14592 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
14593 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
14594 * doc/tm.texi: Regenerate.
14595 * expr.c (expand_expr_real_1): Expand decls using the
14596 expand_var_decl OpenACC hook if defined.
14597 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
14598 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
14599 * omp-low.c (omp_context): Add oacc_privatization_candidates
14601 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
14603 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
14604 private marker's gimple call arguments, and pass it to
14605 lower_oacc_reductions.
14606 (oacc_privatization_scan_clause_chain)
14607 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
14609 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
14610 * omp-offload.c (convert.h): Include.
14611 (oacc_loop_xform_head_tail): Treat private-variable markers like
14612 fork/join when transforming head/tail sequences.
14613 (struct var_decl_rewrite_info): Add struct.
14614 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
14615 (execute_oacc_device_lower): Support rewriting gang-private
14616 variables using target hook, and fix up addr_expr and var_decl
14618 * target.def (adjust_private_decl, expand_var_decl): New hooks.
14619 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
14621 (gcn_goacc_adjust_private_decl): ...this.
14622 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
14624 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
14625 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
14626 definition using gcn_goacc_adjust_gangprivate_decl...
14627 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
14628 gcn_goacc_adjust_private_decl.
14629 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
14630 (gang_private_shared_size): New global variable.
14631 (gang_private_shared_align): Likewise.
14632 (gang_private_shared_sym): Likewise.
14633 (gang_private_shared_hmap): Likewise.
14634 (nvptx_option_override): Initialize these.
14635 (nvptx_file_end): Output gang_private_shared_sym.
14636 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
14638 (nvptx_set_current_function): Clear gang_private_shared_hmap.
14639 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
14640 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
14642 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
14644 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
14646 2021-05-21 Richard Biener <rguenther@suse.de>
14647 H.J. Lu <hjl.tools@gmail.com>
14649 PR middle-end/90773
14650 * expr.c (expand_constructor): Elide expand_constructor if
14651 move by pieces is preferred.
14653 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14655 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
14656 Take a flag and mode value as arguments.
14657 (aarch64_modifies_global_state_p): Likewise.
14658 (aarch64_reads_global_state_p): Likewise.
14659 (aarch64_could_trap_p): Likewise.
14660 (aarch64_get_attributes): Likewise.
14661 (aarch64_init_simd_builtins): Adjust callsite of above.
14662 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
14663 function attributes to apply to builtins.
14664 (aarch64_init_crc32_builtins): Likewise.
14665 (aarch64_init_builtin_rsqrt): Likewise.
14667 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14669 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
14670 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
14671 (gen_2logical): Use new fusion types.
14672 * config/rs6000/fusion.md: Regenerate.
14674 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
14677 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
14678 Handle V4QI and V2HI modes.
14679 (ix86_expand_sse_movcc): Ditto.
14680 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
14681 New instruction pattern.
14682 (*eq<VI_32:mode>3): Ditto.
14683 (*gt<VI_32:mode>3): Ditto.
14684 (*xop_pcmov_<VI_32:mode>): Ditto.
14685 (mmx_pblendvb32): Ditto.
14686 (mmx_pblendvb64): Rename from mmx_pblendvb.
14687 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
14688 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
14689 (vcond<VI_32:mode><VI_32:mode>): Ditto.
14690 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
14691 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
14693 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14695 PR tree-optimization/94589
14696 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
14697 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
14698 not the same as cmp3) treat <= the same as < and >= the same as >.
14699 Don't require that cond2_phi_edge is true edge, instead take
14700 false/true edges into account based on cmp1/cmp2 comparison kinds.
14702 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
14705 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
14706 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
14707 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
14708 using SMAXMIN_MMXMODEI mode iterator.
14709 (*<smaxmin:code>v4qi3): New insn pattern.
14710 (*<smaxmin:code>v2hi3): Ditto.
14711 (SMAXMIN_VI_32): New mode iterator.
14712 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
14713 (UMAXMIN_MMXMODEI): New mode iterator.
14714 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
14715 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
14716 using UMAXMIN_MMXMODEI mode iterator.
14717 (*<umaxmin:code>v4qi3): New insn pattern.
14718 (*<umaxmin:code>v2hi3): Ditto.
14719 (UMAXMIN_VI_32): New mode iterator.
14720 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
14721 (abs<VI_32:mode>2): New insn pattern.
14722 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
14723 * config/i386/sse.md: ... here.
14725 2021-05-20 Clement Chigot <clement.chigot@atos.net>
14726 David Edelsohn <dje.gcc@gmail.com>
14728 * collect2.c (scan_prog_file): Issue non-fatal warning for
14731 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
14733 * doc/invoke.texi (-Wno-c++11-extensions)
14734 (-Wno-c++14-extensions, -Wno-c++17-extensions)
14735 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
14738 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
14740 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
14741 * config/darwin.c (darwin_override_options): Likewise.
14742 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
14743 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
14744 (DWARF2_FRAME_REG_OUT): Likewise.
14745 * config/mips/mips.c (mips_output_filename): Likewise.
14746 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
14748 (rs6000_dbx_register_number): Likewise.
14749 * dbxout.c: Include flags.h.
14750 * dwarf2cfi.c (cfi_label_required_p): Likewise.
14751 (dwarf2out_do_frame): Likewise.
14752 * except.c: Include flags.h.
14753 * final.c (dwarf2_debug_info_emitted_p): Likewise.
14754 (final_scan_insn_1): Likewise.
14755 * flags.h (dwarf_debuginfo_p): New function declaration.
14756 * opts.c (dwarf_debuginfo_p): New function definition.
14757 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
14758 * toplev.c (process_options): Likewise.
14760 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
14762 * common.opt: Change type to support bitmasks.
14763 * flag-types.h (enum debug_info_type): Rename enumerator constants.
14764 (NO_DEBUG): New bitmask.
14765 (DBX_DEBUG): Likewise.
14766 (DWARF2_DEBUG): Likewise.
14767 (XCOFF_DEBUG): Likewise.
14768 (VMS_DEBUG): Likewise.
14769 (VMS_AND_DWARF2_DEBUG): Likewise.
14770 * flags.h (debug_set_to_format): New function declaration.
14771 (debug_set_count): Likewise.
14772 (debug_set_names): Likewise.
14773 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
14774 (debug_set_to_format): New function definition.
14775 (debug_set_count): Likewise.
14776 (debug_set_names): Likewise.
14777 (set_debug_level): Update access to debug_type_names.
14778 * toplev.c: Likewise.
14780 2021-05-20 Martin Sebor <msebor@redhat.com>
14782 PR middle-end/100684
14783 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
14785 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
14788 * config/i386/i386.md (isa): Remove x64_bmi.
14789 (enabled): Remove x64_bmi.
14790 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
14791 Remove general register alternative.
14792 (*andnot<VI_32:mode>3): Ditto.
14793 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
14794 (*<any_logic:code><VI_32:mode>3): Ditto.
14796 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
14798 * config/arm/arm.c: Include head files tree-vectorizer.h and
14801 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
14804 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
14805 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
14806 (<s>mulv4hi3_highpart): New expander.
14807 (*<s>mulv2hi3_highpart): New insn pattern.
14808 (<s>mulv2hi3_higpart): New expander.
14809 (*<any_shift:insn>v2hi3): New insn pattern.
14810 (<any_shift:insn>v2hi3): New expander.
14811 * config/i386/sse.md (smulhrsv2hi3): New expander.
14812 (*smulhrsv2hi3): New insn pattern.
14814 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
14816 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
14818 * params.opt (vect-inner-loop-cost-factor): New.
14819 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
14820 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
14821 tree-vectorizer.h and its required ones.
14822 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
14823 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
14824 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
14825 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
14826 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
14827 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
14829 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
14830 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
14831 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
14833 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
14834 Torbjörn Svensson <torbjorn.svensson@st.com>
14837 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
14839 2021-05-20 Jakub Jelinek <jakub@redhat.com>
14841 PR middle-end/99928
14842 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
14843 explicit linear clause when combined with target, make it map(tofrom:)
14844 instead of no clause or firstprivate.
14846 2021-05-20 Jakub Jelinek <jakub@redhat.com>
14848 PR tree-optimization/94589
14849 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
14850 of integral conversions.
14852 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
14854 * gimple-range.cc (fur_source::get_operand): New.
14855 (gimple_range_fold): Delete.
14856 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
14857 (fold_using_range::range_of_range_op): Move from gimple_ranger.
14858 (fold_using_range::range_of_address): Ditto.
14859 (fold_using_range::range_of_phi): Ditto.
14860 (fold_using_range::range_of_call): Ditto.
14861 (fold_using_range::range_of_builtin_ubsan_call): Move from
14862 range_of_builtin_ubsan_call.
14863 (fold_using_range::range_of_builtin_call): Move from
14864 range_of_builtin_call.
14865 (gimple_ranger::range_of_builtin_call): Delete.
14866 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
14867 (gimple_ranger::fold_range_internal): New.
14868 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
14869 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
14870 gimple_ranger. Improve ranges of SSA_NAMES when possible.
14871 * gimple-range.h (gimple_ranger): Remove various range_of routines.
14872 (class fur_source): New.
14873 (class fold_using_range): New.
14874 (fur_source::fur_source): New.
14876 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
14877 instead of range_of_builtin_call.
14879 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
14881 * doc/cpp.texi (Common Predefined Macros): Update documentation
14882 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
14884 2021-05-19 Alex Coplan <alex.coplan@arm.com>
14887 * config/arm/arm.md (nonsecure_call_internal): Always ensure
14888 callee's address is in a register.
14890 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
14892 * common/config/riscv/riscv-common.c
14893 (riscv_subset_list::parsing_subset_version): Properly parse the letter
14895 (riscv_subset_list::parse_std_ext,
14896 riscv_subset_list::parse_multiletter_ext): To handle errors generated
14897 in riscv_subset_list::parsing_subset_version.
14899 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
14901 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
14902 type attribute in patterns generating XTN(2).
14904 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
14906 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
14907 Remove as duplicate of...
14908 (aarch64_xtn<mode>): This.
14909 (aarch64_xtn2<mode>_le): Move position in file.
14910 (aarch64_xtn2<mode>_be): Move position in file.
14911 (aarch64_xtn2<mode>): Move position in file.
14912 (vec_pack_trunc_<mode>): Define as an expander.
14914 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
14916 * config/aarch64/aarch64-simd-builtins.def: Split builtin
14917 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
14918 separate scalar and vector generators.
14919 * config/aarch64/aarch64-simd.md
14920 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
14922 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
14923 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
14924 * config/aarch64/iterators.md: Define SD_HSDI iterator.
14926 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
14928 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
14930 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
14932 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
14934 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
14935 Implement as an expand emitting a big/little endian
14936 instruction pattern.
14937 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
14938 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
14940 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
14942 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
14943 Implement as an expand emitting a big/little endian
14944 instruction pattern.
14945 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
14946 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
14947 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
14948 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
14950 2021-05-19 Richard Biener <rguenther@suse.de>
14952 PR middle-end/100672
14953 * fold-const.c (fold_negate_expr_1): Use element_precision.
14954 (negate_expr_p): Likewise.
14956 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
14958 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
14959 (pred_load): New int attribute.
14960 * config/aarch64/aarch64-sve.md
14961 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
14962 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
14963 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
14964 code_for_aarch64_load.
14966 2021-05-19 Richard Biener <rguenther@suse.de>
14968 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
14969 sure TARGET_MEM_REF bases are expanded as memory.
14970 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
14971 Do not mark TARGET_MEM_REF bases addressable.
14972 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
14973 TARGET_MEM_REF bases as never rewritable.
14974 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
14975 walk TARGET_MEM_REF bases as address-takens.
14976 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
14978 2021-05-19 Richard Biener <rguenther@suse.de>
14980 * builtins.c (get_object_alignment_1): Strip outer
14982 * tree-dfa.c (get_ref_base_and_extent): Handle outer
14983 WITH_SIZE_EXPR for size processing and process the
14985 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
14986 outer WITH_SIZE_EXPR.
14987 (ao_ref_base_alias_ptr_type): Likewise.
14988 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
14989 and handle that accordingly, stripping it for the
14990 core alias workers.
14991 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
14992 looking through it instead of returning NULL.
14994 2021-05-19 Jakub Jelinek <jakub@redhat.com>
14996 PR middle-end/100576
14997 * builtins.c (check_read_access): Convert bound to size_type_node if
15000 2021-05-19 Richard Biener <rguenther@suse.de>
15002 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15003 (verify_types_in_gimple_reference): ... here. Sanitize.
15004 (verify_gimple_call): Verify references in LHS and arguments.
15005 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15007 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15009 * config/i386/i386.h (VALID_INT_MODE_P):
15010 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15011 * config/i386/i386.md (isa): Add x64_bmi.
15012 (enabled): Handle x64_bmi.
15013 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15014 Add alternative using 64bit general registers.
15015 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15017 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15019 PR middle-end/99928
15020 * tree.h (OMP_MASTER_COMBINED): Define.
15021 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15022 handling for outer combined/composite constructs to a loop.
15023 Handle lastprivate on combined target.
15024 (gimplify_expr): Formatting fix.
15026 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15028 * passes.def: Add sink_code pass before store_merging.
15029 * tree-ssa-sink.c (pass_sink_code:clone): New.
15031 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15033 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15034 rs6000_special_adjust_field_align_p.
15035 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15036 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15037 Remove ABI warning.
15038 (rs6000_function_arg): Likewise.
15039 * config/rs6000/rs6000-protos.h
15040 (rs6000_special_adjust_field_align_p): Remove prototype.
15041 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15043 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15044 rs6000_special_adjust_field_align_p.
15046 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15049 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15050 Add V4QI and V2HI modes.
15051 (VALID_INT_MODE_P): Ditto.
15052 * config/i386/mmx.md (VI_32): New mode iterator.
15053 (mmxvecsize): Handle V4QI and V2HI.
15055 (mov<VI_32:mode>): New expander.
15056 (*mov<mode>_internal): New insn pattern.
15057 (movmisalign<VI_32:mode>): New expander.
15058 (neg<VI_32:mode>): New expander.
15059 (<plusminus:insn><VI_32:mode>3): New expander.
15060 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15061 (mulv2hi3): New expander.
15062 (*mulv2hi3): New insn pattern.
15063 (one_cmpl<VI_32:mode>2): New expander.
15064 (*andnot<VI_32:mode>3): New insn pattern.
15065 (<any_logic:code><VI_32:mode>3): New expander.
15066 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15068 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15070 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15071 Fix a mode mismatch with operand 1.
15073 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15076 * config/i386/i386-expand.c (split_double_mode): Return
15077 temporary register when simplify_gen_subreg fails with
15078 the high half od the paradoxical subreg.
15080 2021-05-18 Richard Biener <rguenther@suse.de>
15082 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15083 and honor it when expanding.
15084 (expand_used_vars_for_block): Pass through forced_stack_var.
15085 (expand_used_vars): Likewise.
15086 (discover_nonconstant_array_refs_r): Set bits in
15087 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15088 (avoid_type_punning_on_regs): Likewise.
15089 (discover_nonconstant_array_refs): Likewise.
15090 (pass_expand::execute): Create and pass down forced_stack_var
15091 bitmap. For parameters and returns temporarily set
15092 TREE_ADDRESSABLE when expand_function_start.
15094 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15096 * doc/sourcebuild.texi: Document 'dg-note'.
15098 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15101 * configure: Regenerate.
15102 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15104 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15106 * gimple.h (is_gimple_omp_oacc): Tighten.
15107 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15109 2021-05-18 Richard Biener <rguenther@suse.de>
15111 * tree-ssa-operands.c (mark_address_taken): Simplify.
15113 2021-05-18 Martin Liska <mliska@suse.cz>
15115 * config/gcn/mkoffload.c (STR): Redefine.
15116 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15117 * config/nvptx/mkoffload.c (STR): Likewise.
15119 2021-05-18 Martin Liska <mliska@suse.cz>
15121 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15122 Use startswith function instead of strncmp.
15123 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15124 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15125 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15126 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15127 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15128 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15129 * config/arm/arm.c (arm_file_start): Likewise.
15130 (arm_valid_target_attribute_rec): Likewise.
15131 (thumb1_md_asm_adjust): Likewise.
15132 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15133 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15134 (avr_set_current_function): Likewise.
15135 (avr_handle_addr_attribute): Likewise.
15136 (avr_asm_output_aligned_decl_common): Likewise.
15137 (avr_asm_named_section): Likewise.
15138 (avr_section_type_flags): Likewise.
15139 (avr_asm_select_section): Likewise.
15140 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15141 (c6x_section_type_flags): Likewise.
15142 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15143 (darwin_objc_declare_unresolved_class_reference): Likewise.
15144 (darwin_objc_declare_class_definition): Likewise.
15145 * config/darwin.c (indirect_data): Likewise.
15146 (darwin_encode_section_info): Likewise.
15147 (darwin_objc2_section): Likewise.
15148 (darwin_objc1_section): Likewise.
15149 (machopic_select_section): Likewise.
15150 (darwin_globalize_label): Likewise.
15151 (darwin_label_is_anonymous_local_objc_name): Likewise.
15152 (darwin_asm_named_section): Likewise.
15153 (darwin_asm_output_dwarf_offset): Likewise.
15154 * config/frv/frv.c (frv_string_begins_with): Likewise.
15155 (frv_in_small_data_p): Likewise.
15156 * config/gcn/mkoffload.c (STR): Likewise.
15158 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15159 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15160 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15161 (ix86_md_asm_adjust): Likewise.
15162 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15163 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15164 (i386_pe_file_end): Likewise.
15165 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15166 (ia64_section_type_flags): Likewise.
15167 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15168 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15169 (mips16_stub_function_p): Likewise.
15170 (mips_function_rodata_section): Likewise.
15171 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15172 (msp430_function_section): Likewise.
15173 (msp430_section_type_flags): Likewise.
15174 (msp430_expand_helper): Likewise.
15175 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15176 (nios2_valid_target_attribute_rec): Likewise.
15177 * config/nvptx/mkoffload.c (process): Likewise.
15179 * config/pa/som.h: Likewise.
15180 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15181 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15182 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15183 (rs6000_inner_target_options): Likewise.
15184 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15185 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15186 * config/vax/vax.c (vax_output_int_move): Likewise.
15187 * config/vms/vms-ld.c (startswith): Likewise.
15188 (process_args): Likewise.
15190 * config/vms/vms.c: Likewise.
15192 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15194 PR rtl-optimization/100590
15195 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15196 they are NONJUMP_INSN_P.
15198 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15201 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15202 DECL_ASSEMBLER_NAME on the fn_decl.
15204 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15206 PR tree-optimization/94589
15207 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15208 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15209 res == 2 would be UB with -ffinite-math-only.
15211 2021-05-18 Martin Liska <mliska@suse.cz>
15213 * Makefile.in: genversion.o should depend on DATESTAMP.
15215 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15217 * config/arc/simdext.md (negv2si2): Remove round bracket.
15219 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15221 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15222 _Bool as macro expanding to _Bool.
15224 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15227 * tree.c (build_reference_type_for_mode)
15228 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15230 (build_reference_type, build_pointer_type): Invoke
15231 build_*_type_for_mode with VOIDmode.
15233 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15235 PR tree-optimization/100512
15236 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15237 and non-zero pointer ranges as invariant.
15238 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15239 processing from here.
15241 2021-05-17 Tom de Vries <tdevries@suse.de>
15244 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15245 * config/nvptx/nvptx.c (nvptx_output_barrier)
15246 (nvptx_output_atomic_insn): New function.
15247 (nvptx_print_operand): Add support for 'B'.
15248 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15251 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15253 PR tree-optimization/100349
15254 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15257 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15259 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15260 (host_detect_local_cpu): Use it.
15262 2021-05-17 Martin Liska <mliska@suse.cz>
15264 * doc/invoke.texi: Add 2 missing dots.
15266 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15268 PR bootstrap/100552
15269 * configure.ac: Replace pattern substitution with call to sed.
15270 * configure: Regenerate.
15272 2021-05-17 Richard Biener <rguenther@suse.de>
15274 PR middle-end/100582
15275 * tree.c (array_at_struct_end_p): Get to the base of the
15276 reference before looking for the underlying decl.
15278 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15280 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15281 alternatives" message more specific, and remove assumption on where
15284 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15286 * config/arm/iterators.md (V16): New iterator.
15287 (VH_cvtto): New iterator.
15288 (v_cmp_result): Added V4HF and V8HF support.
15289 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15290 (vcond<mode><mode>): Likewise.
15291 (vcond_mask_<mode><v_cmp_result>): Likewise.
15292 (vcond<VH_cvtto><mode>): New expander.
15294 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15296 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15298 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15300 (arm_expand_vcond): Likewise.
15301 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15302 VCMPEQQ_N_S, VCMPNEQ_N_S.
15303 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15304 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15305 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15306 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15307 (@mve_vpselq_<supf><mode>): Likewise.
15308 (@mve_vpselq_f<mode>"): Likewise.
15309 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15310 and move to vec-common.md.
15311 (vec_cmpu<mode><mode>): Likewise.
15312 (vcond<mode><mode>): Likewise.
15313 (vcond<V_cvtto><mode>): Likewise.
15314 (vcondu<mode><v_cmp_result>): Likewise.
15315 (vcond_mask_<mode><v_cmp_result>): Likewise.
15316 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15317 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15318 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15319 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15320 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15321 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15323 (vec_cmpu<mode><mode>): Likewise.
15324 (vcond<mode><mode>): Likewise.
15325 (vcond<V_cvtto><mode>): Likewise.
15326 (vcondu<mode><v_cmp_result>): Likewise.
15327 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15330 2021-05-17 liuhongt <hongtao.liu@intel.com>
15333 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
15334 gsi_insert_seq_before instead.
15336 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15338 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
15339 (arm_sat_ok): ...this.
15341 2021-05-17 Martin Liska <mliska@suse.cz>
15343 * lto-wrapper.c (merge_flto_options): Factor out a new function.
15344 (merge_and_complain): Use it.
15345 (run_gcc): Merge also linker command line -flto=foo argument
15348 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
15350 * config/arm/arm.h (CPP_SPEC): Remove error message about
15351 -mlittle-endian/-mbig-endian conflict.
15353 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15355 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
15356 __ROP_PROTECT__ if -mrop-protect is selected.
15358 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15360 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
15361 rop_hash_save_offset and rop_hash_size.
15362 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
15363 rop_hash_size and rop_hash_save_offset.
15364 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
15365 (rs6000_emit_prologue): Emit hashst[p] in prologue.
15366 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
15367 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
15369 (hashst): New define_insn.
15370 (hashchk): Likewise.
15372 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15374 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15375 Disable shrink wrap when inserting ROP-protect instructions.
15376 * config/rs6000/rs6000.opt (mrop-protect): New option.
15377 (mprivileged): Likewise.
15378 * doc/invoke.texi: Document mrop-protect and mprivileged.
15380 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
15382 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
15385 2021-05-15 Martin Jambor <mjambor@suse.cz>
15388 2021-05-13 Martin Jambor <mjambor@suse.cz>
15390 PR tree-optimization/100453
15391 * tree-sra.c (sra_modify_assign): All const base accesses do not
15392 need refreshing, not just those from decl_pool.
15393 (sra_modify_assign): Do not refresh into a const base decl.
15395 2021-05-15 Jakub Jelinek <jakub@redhat.com>
15397 PR rtl-optimization/100342
15398 * regcprop.c (copy_value): When copying a source reg in a wider
15399 mode than it has recorded for the value, adjust recorded destination
15400 mode too or punt if !REG_CAN_CHANGE_MODE_P.
15402 2021-05-14 Jason Merrill <jason@redhat.com>
15404 * intl.h: Add comments.
15406 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15408 * config/aarch64/aarch64-simd.md
15409 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
15410 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
15411 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
15412 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
15413 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
15414 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
15415 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
15416 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
15417 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
15419 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15422 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
15423 boolean logic equivalent.
15424 (vtst_s16): Likewise.
15425 (vtst_s32): Likewise.
15426 (vtst_u8): Likewise.
15427 (vtst_u16): Likewise.
15428 (vtst_u32): Likewise.
15429 (vtst_p8): Likewise.
15430 (vtst_p16): Likewise.
15431 (vtstq_s8): Likewise.
15432 (vtstq_s16): Likewise.
15433 (vtstq_s32): Likewise.
15434 (vtstq_u8): Likewise.
15435 (vtstq_u16): Likewise.
15436 (vtstq_u32): Likewise.
15437 (vtstq_p8): Likewise.
15438 (vtstq_p16): Likewise.
15439 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
15440 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
15442 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15444 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
15445 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
15446 (aarch64_sqdmlsl2<mode>): Delete.
15447 (aarch64_sqdmlal2_lane<mode>): Merge this...
15448 (aarch64_sqdmlsl2_lane<mode>): ... And this...
15449 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
15450 (aarch64_sqdmlal2_laneq<mode>): Merge this...
15451 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
15452 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
15453 (aarch64_sqdmlal2_n<mode>): Merge this...
15454 (aarch64_sqdmlsl2_n<mode>): ... And this...
15455 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
15457 2021-05-13 Martin Sebor <msebor@redhat.com>
15459 PR middle-end/100574
15460 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
15461 all null arguments.
15463 2021-05-13 Martin Sebor <msebor@redhat.com>
15465 PR tree-optimization/93100
15466 PR middle-end/98583
15467 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
15468 don't modify referenced objects.
15470 2021-05-13 Martin Jambor <mjambor@suse.cz>
15472 PR tree-optimization/100453
15473 * tree-sra.c (sra_modify_assign): All const base accesses do not
15474 need refreshing, not just those from decl_pool.
15475 (sra_modify_assign): Do not refresh into a const base decl.
15477 2021-05-13 Martin Liska <mliska@suse.cz>
15479 * tree-ssa-dom.c: Remove m_simplifier.
15481 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
15484 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
15485 canonicalize DImode inequality comparisons against the
15486 maximum integral value.
15488 2021-05-13 Jakub Jelinek <jakub@redhat.com>
15490 PR tree-optimization/98856
15491 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
15492 Expect V2DI and V4DI arithmetic right shifts to be emulated.
15493 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
15495 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
15496 expand_vec_perm_2perm_pblendv): New functions.
15497 (ix86_expand_vec_perm_const_1): Use them.
15498 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
15499 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
15500 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
15501 (ashrv4di3): New define_expand.
15502 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
15503 and !TARGET_AVX512VL expansion.
15505 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
15508 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
15509 sizes < 16 to a register when constructing vpcmov pattern.
15510 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
15512 2021-05-13 Martin Liska <mliska@suse.cz>
15514 * gcov-io.c (gcov_write_block): Remove.
15515 (gcov_write_words): Likewise.
15516 (gcov_read_words): Re-implement using gcov_read_bytes.
15517 (gcov_allocate): Remove.
15518 (GCOV_BLOCK_SIZE): Likewise.
15519 (struct gcov_var): Remove most of the fields.
15520 (gcov_position): Implement with ftell.
15521 (gcov_rewrite): Remove setting of start and offset fields.
15522 (from_file): Re-format.
15523 (gcov_open): Remove setbuf call. It should not be needed.
15524 (gcov_close): Remove internal buffer handling.
15525 (gcov_magic): Use __builtin_bswap32.
15526 (gcov_write_counter): Use directly gcov_write_unsigned.
15527 (gcov_write_string): Use direct fwrite and do not round
15529 (gcov_seek): Use directly fseek.
15530 (gcov_write_tag): Use gcov_write_unsigned directly.
15531 (gcov_write_length): Likewise.
15532 (gcov_write_tag_length): Likewise.
15533 (gcov_read_bytes): Use directly fread.
15534 (gcov_read_unsigned): Use gcov_read_words.
15535 (gcov_read_counter): Likewise.
15536 (gcov_read_string): Use gcov_read_bytes.
15537 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
15538 that size is not in bytes, but words (4B).
15539 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
15540 (GCOV_TAG_ARCS_LENGTH): Likewise.
15541 (GCOV_TAG_ARCS_NUM): Likewise.
15542 (GCOV_TAG_COUNTER_LENGTH): Likewise.
15543 (GCOV_TAG_COUNTER_NUM): Likewise.
15544 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
15546 2021-05-13 liuhongt <hongtao.liu@intel.com>
15549 * config/i386/sse.md (ssedoublevecmode): Add attribute for
15550 V64QI/V32HI/V16SI/V4DI.
15551 (ssehalfvecmode): Add attribute for V2DI/V2DF.
15552 (*vec_concatv4si_0): Extend to VI124_128.
15553 (*vec_concat<mode>_0): New pre-reload splitter.
15554 * config/i386/predicates.md (movq_parallel): New predicate.
15556 2021-05-13 Alexandre Oliva <oliva@adacore.com>
15558 * targhooks.c (default_zero_call_used_regs): Retry using
15559 successfully-zeroed registers as sources.
15561 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
15563 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
15565 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
15568 * gimple-range.cc (range_of_builtin_call): Skip out on
15569 processing __builtin_clz when varying.
15571 2021-05-12 Tom de Vries <tdevries@suse.de>
15574 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
15575 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
15576 to ptx_version_option.
15577 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
15578 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
15579 (define_insn "nvptx_vote_ballot"): Use sync variant for
15581 * config/nvptx/nvptx.opt (ptx_version): Add enum.
15582 (mptx): Add option.
15583 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
15585 2021-05-12 Richard Biener <rguenther@suse.de>
15587 PR tree-optimization/100566
15588 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
15589 allow_back for all edge queries.
15591 2021-05-12 liuhongt <hongtao.liu@intel.com>
15594 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
15595 splitters for pblendvb of NOT mask register.
15597 2021-05-12 Richard Biener <rguenther@suse.de>
15599 PR tree-optimization/100519
15600 * tree-ssa-reassoc.c (can_associate_p): Split into...
15601 (can_associate_op_p): ... this
15602 (can_associate_type_p): ... and this.
15603 (is_reassociable_op): Call can_associate_op_p.
15604 (break_up_subtract_bb): Call the appropriate predicates.
15605 (reassociate_bb): Likewise.
15607 2021-05-12 Martin Liska <mliska@suse.cz>
15609 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
15610 (run_gcc): Use -flto argument detection for merged
15613 2021-05-12 Martin Liska <mliska@suse.cz>
15615 * lto-wrapper.c (print_lto_docs_link): New function.
15616 (run_gcc): Print warning about missing job server detection
15617 after we know NR of partitions. Do the same for -flto{,=1}.
15618 * opts.c (get_option_html_page): Support -flto option.
15620 2021-05-12 Martin Liska <mliska@suse.cz>
15622 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
15624 (append_option): Remove.
15625 (find_option): Rework to use the vector type.
15626 (remove_option): Remove.
15627 (merge_and_complain): Use vectors for cl_decoded_option data
15629 (append_compiler_options): Likewise.
15630 (append_diag_options): Likewise.
15631 (append_linker_options): Likewise.
15632 (append_offload_options): Likewise.
15633 (compile_offload_image): Likewise.
15634 (compile_images_for_offload_targets): Likewise.
15635 (find_and_merge_options): Likewise.
15636 (run_gcc): Likewise.
15638 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
15641 * dwarf2out.c (dwarf2out_finish): Set
15642 have_multiple_function_sections with multi-range text_section.
15644 2021-05-12 Martin Liska <mliska@suse.cz>
15646 PR bootstrap/100560
15647 * Makefile.in: Remove version.h from linker command line.
15649 2021-05-12 Richard Biener <rguenther@suse.de>
15651 PR middle-end/100547
15652 * rtl.h (rtvec_alloc): Make argument size_t.
15653 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
15655 2021-05-12 Jakub Jelinek <jakub@redhat.com>
15657 PR middle-end/100508
15658 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
15659 type, don't reuse DECL_RTL if it has different mode, instead force
15660 creation of a new DEBUG_EXPR.
15662 2021-05-12 Jakub Jelinek <jakub@redhat.com>
15663 Marc Glisse <marc.glisse@inria.fr>
15665 PR tree-optimization/94589
15666 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
15667 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
15669 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
15672 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
15673 * config/i386/mmx.md (MMXMODE124): New mode iterator.
15675 (mmxintvecmode): New mode attribute.
15676 (mmxintvecmodelower): Ditto.
15677 (*mmx_maskcmpv2sf3_comm): New insn pattern.
15678 (*mmx_maskcmpv2sf3): Ditto.
15679 (vec_cmpv2sfv2si): New expander.
15680 (vcond<V2FI:mode>v2si): Ditto.
15681 (mmx_vlendvps): New insn pattern.
15682 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
15683 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
15684 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
15686 2021-05-11 Martin Sebor <msebor@redhat.com>
15688 PR middle-end/21433
15689 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
15691 2021-05-11 Richard Biener <rguenther@suse.de>
15693 * gimple-fold.c (gimple_fold_call): Do not call
15694 maybe_fold_reference on call arguments or the static chain.
15695 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
15698 2021-05-11 Martin Liska <mliska@suse.cz>
15700 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
15701 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
15702 (DEF_HSAIL_SAT_BUILTIN): Likewise.
15703 (DEF_HSAIL_INTR_BUILTIN): Likewise.
15704 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
15705 * doc/frontends.texi: Remove BRIG.
15706 * doc/install.texi: Likewise.
15707 * doc/invoke.texi: Likewise.
15708 * doc/standards.texi: Likewise.
15709 * brig-builtins.def: Removed.
15710 * brig/ChangeLog: Removed.
15711 * brig/Make-lang.in: Removed.
15712 * brig/brig-builtins.h: Removed.
15713 * brig/brig-c.h: Removed.
15714 * brig/brig-lang.c: Removed.
15715 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
15716 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
15717 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
15718 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
15719 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
15720 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
15721 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
15722 * brig/brigfrontend/brig-comment-handler.cc: Removed.
15723 * brig/brigfrontend/brig-control-handler.cc: Removed.
15724 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
15725 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
15726 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
15727 * brig/brigfrontend/brig-function-handler.cc: Removed.
15728 * brig/brigfrontend/brig-function.cc: Removed.
15729 * brig/brigfrontend/brig-function.h: Removed.
15730 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
15731 * brig/brigfrontend/brig-label-handler.cc: Removed.
15732 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
15733 * brig/brigfrontend/brig-machine.c: Removed.
15734 * brig/brigfrontend/brig-machine.h: Removed.
15735 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
15736 * brig/brigfrontend/brig-module-handler.cc: Removed.
15737 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
15738 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
15739 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
15740 * brig/brigfrontend/brig-to-generic.cc: Removed.
15741 * brig/brigfrontend/brig-to-generic.h: Removed.
15742 * brig/brigfrontend/brig-util.cc: Removed.
15743 * brig/brigfrontend/brig-util.h: Removed.
15744 * brig/brigfrontend/brig-variable-handler.cc: Removed.
15745 * brig/brigfrontend/hsa-brig-format.h: Removed.
15746 * brig/brigfrontend/phsa.h: Removed.
15747 * brig/brigspec.c: Removed.
15748 * brig/config-lang.in: Removed.
15749 * brig/gccbrig.texi: Removed.
15750 * brig/lang-specs.h: Removed.
15751 * brig/lang.opt: Removed.
15753 2021-05-11 Richard Biener <rguenther@suse.de>
15756 * ipa-param-manipulation.c
15757 (ipa_param_body_adjustments::modify_call_stmt): Avoid
15758 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
15759 via gimple_call_lhs_ptr.
15761 2021-05-11 Alex Coplan <alex.coplan@arm.com>
15764 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
15765 Avoid emitting CFA adjusts on the sp if we have the fp.
15767 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
15769 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
15770 (VMULD): New iterator.
15771 (VCOND): Handle V4HF and V8HF.
15772 (VCONQ): Fix entry for V2SF.
15773 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
15774 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
15775 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
15776 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
15777 Use a 128-bit vector mode for the indexed operand. Use stype for
15778 the scheduling type.
15780 2021-05-11 Richard Biener <rguenther@suse.de>
15782 * gimple-fold.c (maybe_fold_reference): Only return
15783 is_gimple_min_invariant values.
15785 2021-05-11 Richard Biener <rguenther@suse.de>
15787 PR middle-end/100509
15788 * gimple-fold.c (fold_gimple_assign): Only call
15789 get_symbol_constant_value on register type symbols.
15791 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
15792 Joe Ramsay <joe.ramsay@arm.com>
15795 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
15796 (__arm_vcmpneq): Remove duplicate definition.
15797 (__arm_vstrwq_scatter_offset_p): Likewise.
15798 (__arm_vmaxq_x): Likewise.
15799 (__arm_vmlsdavaq): Likewise.
15800 (__arm_vmlsdavaxq): Likewise.
15801 (__arm_vmlsdavq_p): Likewise.
15802 (__arm_vmlsdavxq_p): Likewise.
15803 (__arm_vrmlaldavhaq): Likewise.
15804 (__arm_vstrbq_p): Likewise.
15805 (__arm_vstrbq_scatter_offset): Likewise.
15806 (__arm_vstrbq_scatter_offset_p): Likewise.
15807 (__arm_vstrdq_scatter_offset): Likewise.
15808 (__arm_vstrdq_scatter_offset_p): Likewise.
15809 (__arm_vstrdq_scatter_shifted_offset): Likewise.
15810 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
15812 2021-05-11 Jakub Jelinek <jakub@redhat.com>
15814 PR middle-end/100471
15815 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
15816 is 0, bypass the reduction loop including
15817 GOMP_taskgroup_reduction_unregister call.
15819 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
15821 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
15822 costing_for_scalar.
15823 (rs6000_density_test): Early return if costing_for_scalar is true.
15824 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
15826 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
15828 * doc/tm.texi: Regenerated.
15829 * target.def (init_cost): Add new parameter costing_for_scalar.
15830 * targhooks.c (default_init_cost): Adjust for new parameter.
15831 * targhooks.h (default_init_cost): Likewise.
15832 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
15833 (vect_compute_single_scalar_iteration_cost): Likewise.
15834 (vect_analyze_loop_2): Likewise.
15835 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
15836 (vect_bb_vectorization_profitable_p): Likewise.
15837 * tree-vectorizer.h (init_cost): Likewise.
15838 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
15839 * config/i386/i386.c (ix86_init_cost): Likewise.
15840 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
15842 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
15844 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
15845 vect_nonmem and moved into...
15846 (struct rs6000_cost_data): ...here.
15847 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
15848 (rs6000_add_stmt_cost): Likewise.
15849 (rs6000_finish_cost): Likewise.
15851 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
15853 * range-op.cc (get_bool_state): Adjust head comment.
15854 (operator_not_equal::op1_range): Fix comment.
15855 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
15857 2021-05-10 Martin Sebor <msebor@redhat.com>
15859 PR middle-end/100425
15860 PR middle-end/100510
15861 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
15862 (pass_walloca::xlimit_certain_p): ...to this.
15863 (pass_walloca::gate): Execute for any kind of handled warning.
15864 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
15865 warnings when xlimit_certain_p is set.
15867 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
15869 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
15870 Return ALTIVEC_REGS if that is best_class.
15871 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
15873 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
15875 * config/arm/arm.h (CPP_SPEC): Remove error message about
15878 2021-05-10 Martin Jambor <mjambor@suse.cz>
15880 * ipa-prop.h (IPA_NODE_REF): Removed.
15881 (IPA_NODE_REF_GET_CREATE): Likewise.
15882 (IPA_EDGE_REF): Likewise.
15883 (IPA_EDGE_REF_GET_CREATE): Likewise.
15884 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
15885 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
15886 use of ipa_node_params_sum.
15887 (ipcp_versionable_function_p): Likewise.
15888 (push_node_to_stack): Likewise.
15889 (pop_node_from_stack): Likewise.
15890 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
15891 direct use of ipa_node_params_sum.
15892 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
15893 ipa_node_params_sum.
15894 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
15896 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
15897 use of ipa_node_params_sum.
15898 (self_recursively_generated_p): Likewise.
15899 (propagate_scalar_across_jump_function): Likewise.
15900 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
15901 direct use of ipa_edge_args_sum, moved the lookup after the early
15902 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
15903 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
15904 direct uses of ipa_node_params_sum.
15905 (propagate_vr_across_jump_function): Likewise.
15906 (propagate_aggregate_lattice): Likewise.
15907 (propagate_aggs_across_jump_function): Likewise.
15908 (propagate_constants_across_call): Likewise, also replaced
15909 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
15910 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
15911 of ipa_node_params_sum.
15912 (estimate_local_effects): Likewise.
15913 (add_all_node_vals_to_toposort): Likewise.
15914 (propagate_constants_topo): Likewise.
15915 (ipcp_propagate_stage): Likewise.
15916 (ipcp_discover_new_direct_edges): Likewise.
15917 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
15918 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
15919 (get_info_about_necessary_edges): Likewise.
15920 (want_remove_some_param_p): Likewise.
15921 (create_specialized_node): Likewise.
15922 (self_recursive_pass_through_p): Likewise.
15923 (self_recursive_agg_pass_through_p): Likewise.
15924 (find_more_scalar_values_for_callers_subset): Likewise and also
15925 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
15926 case replacing two of those with a single query.
15927 (find_more_contexts_for_caller_subset): Likewise for the
15928 ipa_polymorphic_call_context overload.
15929 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
15930 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
15931 ipa_node_params_sum.
15932 (find_aggregate_values_for_callers_subset): Likewise, also reusing
15933 results of ipa_edge_args_sum->get.
15934 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
15935 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
15936 direct use of ipa_edge_args_sum.
15937 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
15938 summary query after the early exit and reused the result later.
15939 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
15940 ipa_node_params_sum.
15941 (decide_whether_version_node): Likewise. Removed re-querying for
15942 summaries after cloning.
15943 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
15944 ipa_node_params_sum.
15945 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
15947 (identify_dead_nodes): Likewise.
15948 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
15949 ipa_node_params_sum.
15950 (ipcp_store_vr_results): Likewise.
15951 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
15952 (ipa_fn_summary_t::duplicate): Likewise.
15953 (analyze_function_body): Likewise.
15954 (estimate_calls_size_and_time): Likewise.
15955 (ipa_cached_call_context::duplicate_from): Likewise.
15956 (ipa_call_context::equal_to): Likewise.
15957 (remap_edge_params): Likewise.
15958 (ipa_merge_fn_summary_after_inlining): Likewise.
15959 (inline_read_section): Likewise.
15960 * ipa-icf.c (sem_function::param_used_p): Likewise.
15961 * ipa-modref.c (compute_parm_map): Likewise.
15962 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
15964 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
15965 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
15967 * ipa-profile.c (check_argument_count): Likewise.
15968 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
15969 with a direct use of ipa_node_params_sum.
15970 (ipa_initialize_node_params): Likewise.
15971 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
15972 direct use of ipa_edge_args_sum and reused the query result.
15973 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
15974 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
15975 direct use of ipa_edge_args_sum.
15976 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
15977 ipa_node_params_sum and reused the result of the query.
15978 (ipa_analyze_node): Likewise.
15979 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
15980 of ipa_node_params_sum.
15981 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
15982 direct uses of ipa_edge_args_sum.
15983 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
15984 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
15985 direct use of ipa_edge_args_sum. Removed superficial re-querying the
15987 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
15988 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
15990 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
15991 direct use of ipa_edge_args_sum.
15992 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
15993 use of ipa_node_params_sum.
15994 (ipa_print_node_params): Likewise.
15995 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
15996 direct uses of ipa_edge_args_sum.
15997 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
15999 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16000 ipa_node_params_sum.
16001 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16002 scopes where it is used.
16004 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16006 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16007 <case E_V2SImode>: Force op_true to register.
16009 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16011 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16012 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16013 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16015 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16016 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16017 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16018 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16019 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16020 (mve_vcmpneq_n_f<mode>): Remove.
16021 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16022 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16023 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16025 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16027 * config/arm/iterators.md (MVE_COMPARISONS): New.
16029 (mve_cmp_type): New.
16030 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16032 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16033 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16034 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16035 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16036 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16037 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16038 (mve_vcmpneq_n_<mode>): Remove.
16040 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16042 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16043 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16044 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16047 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16049 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16050 (vcmpneq_n_u): Likewise.
16051 (vcmpeqq_u,): Likewise.
16052 (vcmpeqq_n_u): Likewise.
16053 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16054 VCMPEQQ_N_U and VCMPNEQ_N_U.
16055 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16056 (mve_vcmpeqq_n): Likewise.
16057 (mve_vcmpeqq): Likewise.
16058 (mve_vcmpneq_n): Likewise.
16060 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16062 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16063 the 's' version of the builtin.
16065 2021-05-10 Richard Biener <rguenther@suse.de>
16067 PR tree-optimization/100492
16068 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16069 Find nothing when the loop contains an irreducible region.
16071 2021-05-10 Richard Biener <rguenther@suse.de>
16073 PR middle-end/100464
16075 * gimple-fold.c (canonicalize_constructor_val): Do not set
16078 2021-05-10 Richard Biener <rguenther@suse.de>
16080 PR tree-optimization/100434
16081 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16083 (dse_optimize_stmt): Handle call LHS by dropping the
16084 LHS or the whole call if it doesn't have other
16086 (pass_dse::execute): Adjust.
16088 2021-05-10 Martin Liska <mliska@suse.cz>
16090 * Makefile.in: Add missing genversion rule.
16092 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16095 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16096 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16098 2021-05-10 Martin Liska <mliska@suse.cz>
16100 * builtins.c (is_builtin_name): Use startswith
16101 function instead of strncmp.
16102 * collect2.c (main): Likewise.
16103 (has_lto_section): Likewise.
16104 (scan_libraries): Likewise.
16105 * coverage.c (coverage_checksum_string): Likewise.
16106 (coverage_init): Likewise.
16107 * dwarf2out.c (is_cxx): Likewise.
16108 (gen_compile_unit_die): Likewise.
16109 * gcc-ar.c (main): Likewise.
16110 * gcc.c (init_spec): Likewise.
16111 (read_specs): Likewise.
16112 (execute): Likewise.
16113 (check_live_switch): Likewise.
16114 * genattrtab.c (write_attr_case): Likewise.
16115 (IS_ATTR_GROUP): Likewise.
16116 * gencfn-macros.c (main): Likewise.
16117 * gengtype.c (type_for_name): Likewise.
16118 (gen_rtx_next): Likewise.
16119 (get_file_langdir): Likewise.
16120 (write_local): Likewise.
16121 * genmatch.c (get_operator): Likewise.
16122 (get_operand_type): Likewise.
16123 (expr::gen_transform): Likewise.
16124 * genoutput.c (validate_optab_operands): Likewise.
16125 * incpath.c (add_sysroot_to_chain): Likewise.
16126 * langhooks.c (lang_GNU_C): Likewise.
16127 (lang_GNU_CXX): Likewise.
16128 (lang_GNU_Fortran): Likewise.
16129 (lang_GNU_OBJC): Likewise.
16130 * lto-wrapper.c (run_gcc): Likewise.
16131 * omp-general.c (omp_max_simt_vf): Likewise.
16132 * omp-low.c (omp_runtime_api_call): Likewise.
16133 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16134 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16135 * real.c (real_from_string): Likewise.
16136 * selftest.c (assert_str_startswith): Likewise.
16137 * timevar.c (timer::validate_phases): Likewise.
16138 * tree.c (get_file_function_name): Likewise.
16139 * ubsan.c (ubsan_use_new_style_p): Likewise.
16140 * varasm.c (default_function_rodata_section): Likewise.
16141 (incorporeal_function_p): Likewise.
16142 (default_section_type_flags): Likewise.
16143 * system.h (startswith): Define startswith.
16145 2021-05-10 Martin Liska <mliska@suse.cz>
16147 * bitmap.h (class auto_bitmap): Remove
16148 __cplusplus >= 201103.
16149 * config/aarch64/aarch64.c: Likewise.
16150 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16152 * sbitmap.h: Likewise.
16154 2021-05-10 Martin Liska <mliska@suse.cz>
16156 * Makefile.in: Rename gcov-iov to genversion and depend
16157 on version.h (instead of gcov-iov.h).
16158 * gcov-io.h: Include version.h instread of gcov-iov.h.
16159 * gengtype-state.c (read_state_version): Likewise.
16160 * gcov-iov.c: Moved to...
16161 * genversion.c: ...here.
16162 * lto-streamer.h (LTO_major_version): Define it with
16164 * version.c: Removed.
16165 * version.h: Removed.
16167 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16169 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16170 * config/arc/simdext.md (VCT): Add predicates for iterator
16173 (voptab): Likewise.
16174 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16175 (<voptab>v2si3): New patterns.
16177 (reduc_plus_scal_v4hi): Likewise.
16178 (reduc_plus_scal_v2si): Likewise.
16179 (vec_duplicatev2si): Likewise.
16180 (vec_duplicatev4hi): Likewise.
16182 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16184 * config/arc/simdext.md: Format and cleanup file.
16186 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16188 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16189 only when munaligned-access option is on.
16190 (movmisalign<mode>): Likewise.
16192 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16194 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16196 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16198 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16200 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16201 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16202 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16203 (norm_f): Likewise.
16206 (clzsi2): Use fls instruction when available.
16207 (arc_clzsi2): Likewise.
16209 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16211 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16213 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16215 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16217 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16219 PR middle-end/100467
16220 * toplev.c (compile_file): Call insn_locations_init before
16221 targetm.asm_out.code_end.
16223 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16226 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16228 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16230 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16231 Andrew Stubbs <amd@codesourcery.com>
16234 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16235 emit_move_insn operands.
16237 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16239 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16240 location for the outgoing edges of an empty block.
16241 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16242 type by means of the get_array_descr_info langhook, if it is set and
16243 returns true. Remove obsolete code dealing with unnamed subtypes.
16245 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16247 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16248 (sbr_vector): Renamed from ssa_block_cache.
16249 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16250 (ssa_block_ranges::~ssa_block_ranges): Remove.
16251 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16252 (ssa_block_ranges::set_bb_varying): Remove.
16253 (sbr_vector::get_bb_range): Adjust assert.
16254 (sbr_vector::bb_range_p): Adjust assert.
16255 (~block_range_cache): No freeing loop required.
16256 (block_range_cache::get_block_ranges): Remove.
16257 (block_range_cache::set_bb_range): Inline get_block_ranges.
16258 (block_range_cache::set_bb_varying): Remove.
16259 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16260 * value-range.h (irange_allocator::get_memory): New.
16262 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16264 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16265 dominator tree is available and requested.
16266 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16267 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16268 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16270 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16272 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16273 only PHI nodes better.
16275 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16277 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16279 (gcond_edge_range): Export prototype.
16280 * gimple-range-edge.cc (gcond_edge_range): New.
16281 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16282 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16284 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16286 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16287 default range into a temp and allocate only what is needed.
16289 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16291 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16293 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16295 * gimple-range.h (gimple_range_global): Pick up parameter initial
16296 values, and use-before defined locals are UNDEFINED.
16298 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16300 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16302 * tree.h (reverse_storage_order_for_component_p): Return false if
16303 the type is a pointer.
16305 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16307 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16309 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16312 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16313 Handle V8QI, V4HI and V2SI modes.
16314 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16315 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16316 * config/i386/i386.md (unspec): ... here.
16318 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16319 Tom de Vries <tdevries@suse.de>
16321 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16322 a truth_value_p reduction variable is nonintegral.
16324 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16327 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16328 Return false for mode sizes < 16.
16330 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16333 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
16335 2021-05-06 Martin Jambor <mjambor@suse.cz>
16337 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
16338 when there is no function summary.
16339 (ipa_sra_summarize_function): produce edge summaries even when
16342 2021-05-06 Tom Tromey <tom@tromey.com>
16344 * godump.c (string_hash_eq): Remove.
16345 (go_finish): Use htab_eq_string.
16347 2021-05-06 Tom Tromey <tom@tromey.com>
16349 * gengtype-state.c (read_state): Use htab_eq_string.
16350 (string_eq): Remove.
16352 2021-05-06 Tom Tromey <tom@tromey.com>
16354 * gensupport.c (htab_eq_string): Remove.
16356 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16359 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
16360 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
16361 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
16362 set_ignored_loc callbacks.
16363 * debug.c (do_nothing_debug_hooks): Likewise.
16364 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
16365 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
16366 (in_text_section_p, last_text_label, last_cold_label,
16367 switch_text_ranges, switch_cold_ranges): New data items.
16368 (dwarf2out_note_section_used): Remove.
16369 (dwarf2out_begin_prologue): Set fde->ignored_debug and
16371 (mark_ignored_debug_section): New helper function.
16372 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
16373 mark_ignored_debug_section.
16374 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
16375 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
16376 (size_of_aranges): Adjust formula for multi-part text ranges size.
16377 (output_aranges): Output multi-part text ranges.
16378 (dwarf2out_set_ignored_loc): New callback function.
16379 (dwarf2out_finish): Output multi-part text ranges.
16380 (dwarf2out_c_finalize): Clear new data items.
16381 * final.c (final_start_function_1): Call set_ignored_loc callback.
16382 (final_scan_insn_1): Likewise.
16383 * ggc-page.c (gt_ggc_mx): New helper function.
16384 * stringpool.c (gt_pch_nx): Likewise.
16386 2021-05-06 Richard Biener <rguenther@suse.de>
16388 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
16389 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
16390 (TV_TREE_INTO_SSA): New.
16391 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
16392 (rewrite_blocks): Likewise.
16393 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
16395 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16397 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
16398 abs_replacement, xor_replacement,
16399 cond_removal_in_popcount_clz_ctz_pattern,
16400 replace_phi_edge_with_variable): Change type of phi argument from
16401 gimple * to gphi *.
16403 2021-05-06 Richard Biener <rguenther@suse.de>
16405 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
16406 Output an opt-info message.
16407 (do_split_loop_on_cond): Likewise.
16408 (tree_ssa_split_loops): Update SSA form here.
16410 2021-05-06 Richard Biener <rguenther@suse.de>
16412 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
16413 return variable removal.
16415 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
16417 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
16418 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
16419 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
16421 * config/s390/s390.c (s390_const_operand_ok): Remove unused
16424 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16426 PR tree-optimization/94589
16427 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
16428 spaceship_replacement.
16429 (cond_only_block_p, spaceship_replacement): New functions.
16431 2021-05-06 Richard Biener <rguenther@suse.de>
16434 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
16435 dealing with a debug use and only query existing addresses
16437 (lower_emutls_1): Avoid splitting out addresses for debug
16438 stmts, reset the debug stmt when we fail to find existing
16440 (lower_emutls_phi_arg): Set wi.stmt.
16442 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
16445 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
16446 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
16447 (stack_protect_test): Use cbranch helper.
16449 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16452 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
16453 always return the establisher frame for __builtin_frame_address (0).
16455 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
16458 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
16459 (ix86_cpu_model_var): Likewise.
16460 (ix86_cpu_features2_type_node): Likewise.
16461 (ix86_cpu_features2_var): Likewise.
16462 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
16465 2021-05-05 Martin Sebor <msebor@redhat.com>
16467 * passes.def (pass_warn_printf): Run after SSA.
16469 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16471 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
16472 * config/arm/predicates.md (minus_one_operand): New predicate.
16474 2021-05-05 Jeff Law <jlaw@tachyum.com>
16476 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
16478 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
16480 PR rtl-optimization/100263
16481 * postreload.c (move2add_valid_value_p): Ensure register can
16484 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16486 PR rtl-optimization/100411
16487 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
16488 and beginning of function markers.
16490 2021-05-05 Jeff Law <jlaw@tachyum.com>
16492 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
16493 * config/cr16/cr16.c (notice_update_cc): Remove.
16494 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
16496 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
16499 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
16500 Handle V8QI, V4HI and V2SI modes.
16501 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
16502 (ix86_build_signbit_mask): Ditto.
16503 * config/i386/mmx.md (MMXMODE14): New mode iterator.
16504 (<smaxmin:code><MMXMODE14:mode>3): New expander.
16505 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
16506 (<umaxmin:code><MMXMODE24:mode>3): New expander.
16507 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
16508 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
16509 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16510 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16511 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16512 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16514 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16516 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
16517 not expand the VALUE_EXPR of variables put in the non-local frame.
16518 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
16519 to be ignored for debug info, ensure its variable offsets are not.
16521 2021-05-05 Richard Biener <rguenther@suse.de>
16523 PR tree-optimization/79333
16524 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
16525 Fold stmt following SSA edges.
16527 2021-05-05 Richard Biener <rguenther@suse.de>
16529 PR middle-end/100394
16530 * calls.c (expand_call): Preserve possibly throwing calls.
16531 * cfgexpand.c (expand_call_stmt): When a call can throw signal
16532 RTL expansion there are side-effects.
16533 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
16534 mark all possibly throwing stmts necessary unless we can elide
16536 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
16537 -fdelete-dead-exceptions.
16538 * tree.h (DECL_PURE_P): Add note about exceptions.
16540 2021-05-05 Alexandre Oliva <oliva@adacore.com>
16542 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
16545 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
16547 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
16548 get_fnname_from_decl for name of thunk.
16549 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
16550 and ASM_OUTPUT_LABEL.
16551 (rs6000_xcoff_declare_function_name): Use assemble_name and
16553 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
16554 (rs6000_xcoff_encode_section_info): Don't add mapping class
16555 for aliases. Always add [DS] mapping class to primary
16557 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
16559 2021-05-04 Martin Sebor <msebor@redhat.com>
16561 PR middle-end/100307
16562 * builtins.c (compute_objsize_r): Clear base0 for pointers.
16564 2021-05-04 Jeff Law <jlaw@tachyum.com>
16566 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
16568 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16570 * caller-save.c: Remove CC0.
16571 * cfgcleanup.c: Remove CC0.
16572 * cfgrtl.c: Remove CC0.
16573 * combine.c: Remove CC0.
16574 * compare-elim.c: Remove CC0.
16575 * conditions.h: Remove CC0.
16576 * config/h8300/h8300.h: Remove CC0.
16577 * config/h8300/h8300-protos.h: Remove CC0.
16578 * config/h8300/peepholes.md: Remove CC0.
16579 * config/i386/x86-tune-sched.c: Remove CC0.
16580 * config/m68k/m68k.c: Remove CC0.
16581 * config/rl78/rl78.c: Remove CC0.
16582 * config/sparc/sparc.c: Remove CC0.
16583 * config/xtensa/xtensa.c: Remove CC0.
16584 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
16585 RTL where that is used as a placeholder only.
16586 * cprop.c: Remove CC0.
16587 * cse.c: Remove CC0.
16588 * cselib.c: Remove CC0.
16589 * df-problems.c: Remove CC0.
16590 * df-scan.c: Remove CC0.
16591 * doc/md.texi: Remove CC0. Adjust an example.
16592 * doc/rtl.texi: Remove CC0. Adjust an example.
16593 * doc/tm.texi: Regenerate.
16594 * doc/tm.texi.in: Remove CC0.
16595 * emit-rtl.c: Remove CC0.
16596 * final.c: Remove CC0.
16597 * fwprop.c: Remove CC0.
16598 * gcse-common.c: Remove CC0.
16599 * gcse.c: Remove CC0.
16600 * genattrtab.c: Remove CC0.
16601 * genconfig.c: Remove CC0.
16602 * genemit.c: Remove CC0.
16603 * genextract.c: Remove CC0.
16604 * gengenrtl.c: Remove CC0.
16605 * genrecog.c: Remove CC0.
16606 * haifa-sched.c: Remove CC0.
16607 * ifcvt.c: Remove CC0.
16608 * ira-costs.c: Remove CC0.
16609 * ira.c: Remove CC0.
16610 * jump.c: Remove CC0.
16611 * loop-invariant.c: Remove CC0.
16612 * lra-constraints.c: Remove CC0.
16613 * lra-eliminations.c: Remove CC0.
16614 * optabs.c: Remove CC0.
16615 * postreload-gcse.c: Remove CC0.
16616 * postreload.c: Remove CC0.
16617 * print-rtl.c: Remove CC0.
16618 * read-rtl-function.c: Remove CC0.
16619 * reg-notes.def: Remove CC0.
16620 * reg-stack.c: Remove CC0.
16621 * reginfo.c: Remove CC0.
16622 * regrename.c: Remove CC0.
16623 * reload.c: Remove CC0.
16624 * reload1.c: Remove CC0.
16625 * reorg.c: Remove CC0.
16626 * resource.c: Remove CC0.
16627 * rtl.c: Remove CC0.
16628 * rtl.def: Remove CC0.
16629 * rtl.h: Remove CC0.
16630 * rtlanal.c: Remove CC0.
16631 * sched-deps.c: Remove CC0.
16632 * sched-rgn.c: Remove CC0.
16633 * shrink-wrap.c: Remove CC0.
16634 * simplify-rtx.c: Remove CC0.
16635 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
16636 CC_STATUS_MDEP, and CC_STATUS.
16637 * target.def: Remove CC0.
16638 * valtrack.c: Remove CC0.
16639 * var-tracking.c: Remove CC0.
16641 2021-05-04 Richard Biener <rguenther@suse.de>
16643 PR tree-optimization/100414
16644 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
16646 (tree_ssa_phiopt_worker): But unconditionally here.
16648 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
16650 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
16651 && and || with floating-point and complex arguments.
16653 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
16655 * tree-inline.c (insert_debug_decl_map): Delete.
16656 (copy_debug_stmt): Minor tweak.
16657 (setup_one_parameter): Do not use a variable if the value is either
16658 a read-only DECL or a non-addressable local variable in the caller.
16659 In this case, insert the debug-only variable in the map manually.
16660 (expand_call_inline): Do not generate a CLOBBER for these values.
16661 * tree-inline.h (debug_map): Minor tweak.
16663 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
16665 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
16666 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
16668 2021-05-04 Richard Biener <rguenther@suse.de>
16670 PR tree-optimization/100329
16671 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
16673 (insert_stmt_after): Assert we're not running into asm goto.
16675 2021-05-04 Richard Biener <rguenther@suse.de>
16677 PR tree-optimization/100398
16678 * tree-ssa-dse.c (pass_dse::execute): Preserve control
16681 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16683 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
16685 2021-05-04 Alexandre Oliva <oliva@adacore.com>
16687 * builtins.c (try_store_by_multiple_pieces): New.
16688 (expand_builtin_memset_args): Use it. If target_char_cast
16689 fails, proceed as for non-constant val. Pass len's ctz to...
16690 * expr.c (clear_storage_hints): ... this. Try store by
16691 multiple pieces after setmem.
16692 (clear_storage): Adjust.
16693 * expr.h (clear_storage_hints): Likewise.
16694 (try_store_by_multiple_pieces): Declare.
16695 * passes.def: Replace the last copy_prop with ccp.
16697 2021-05-03 Tom de Vries <tdevries@suse.de>
16700 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
16703 2021-05-03 Richard Biener <rguenther@suse.de>
16705 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
16707 2021-05-03 Richard Biener <rguenther@suse.de>
16709 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
16710 (dse_dom_walker): Remove.
16711 (dse_dom_walker::dse_optimize_stmt): Rename...
16712 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
16713 (dse_dom_walker::before_dom_children): Inline ...
16714 (pass_dse::execute): ... here. Perform a reverse program
16717 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
16720 * configure: Regenerated.
16722 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
16725 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
16726 (s390_md_asm_adjust): Handle hard registers.
16728 2021-05-03 Jakub Jelinek <jakub@redhat.com>
16730 PR tree-optimization/100382
16731 * tree-ssa-dse.c: Include tree-eh.h.
16732 (dse_dom_walker::before_dom_children): Don't remove stmts if
16733 stmt_unremovable_because_of_non_call_eh_p is true.
16735 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
16737 * varasm.c (compute_reloc_for_var): Split out from...
16738 (get_variable_section): Use it.
16739 * output.h (compute_reloc_for_var): Declare.
16740 * config/rs6000/rs6000-protos.h
16741 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
16743 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
16744 Don't append storage mapping class to symbol.
16745 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
16746 Don't convert TLS BSS to common.
16747 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
16748 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
16750 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
16752 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
16753 If align is 0 from TLS class, use the same rules as varasm.c
16754 If not common, switch to BSS section manually.
16755 If common, emit appropriate comm or lcomm directive.
16756 (rs6000_xcoff_encode_section_info): Add logic to append all
16757 storage mapping classes.
16758 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
16759 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
16760 rs6000_xcoff_asm_output_aligned_decl_common.
16761 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
16762 rs6000_xcoff_asm_output_aligned_decl_common.
16763 (ASM_OUTPUT_TLS_COMMON): Use
16764 rs6000_xcoff_asm_output_aligned_decl_common.
16766 2021-05-02 Jakub Jelinek <jakub@redhat.com>
16769 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
16770 as first argument of pseudo_node_t constructors.
16772 2021-05-02 Jakub Jelinek <jakub@redhat.com>
16775 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
16777 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
16779 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
16780 (gt_pch_nx (int_range<1> *&)): New.
16781 (gt_ggc_mx (int_range<1> *&)): New.
16782 * value-range.h (class irange): Add GTY support for
16785 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
16787 * doc/options.texi (Negative): Change either or to both and.
16789 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16791 * config/aarch64/aarch64-simd-builtins.def: Add
16792 float_ml[as][q]_laneq builtin generator macros.
16793 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
16794 (aarch64_float_mla_laneq<mode>): Define.
16795 (aarch64_float_mls_laneq<mode>): Define.
16796 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
16797 instead of GCC vector extensions.
16798 (vmlaq_laneq_f32): Likewise.
16799 (vmls_laneq_f32): Likewise.
16800 (vmlsq_laneq_f32): Likewise.
16802 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16804 * config/aarch64/aarch64-simd-builtins.def: Add
16805 float_ml[as]_lane builtin generator macros.
16806 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
16808 (mul_lane<mode>3): This, and re-order arguments.
16809 (aarch64_float_mla_lane<mode>): Define.
16810 (aarch64_float_mls_lane<mode>): Define.
16811 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
16812 instead of GCC vector extensions.
16813 (vmlaq_lane_f32): Likewise.
16814 (vmls_lane_f32): Likewise.
16815 (vmlsq_lane_f32): Likewise.
16817 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16819 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
16820 builtin generator macros.
16821 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
16823 (aarch64_float_mls<mode>): Define.
16824 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
16825 instead of relying on GCC vector extensions.
16826 (vmla_f64): Likewise.
16827 (vmlaq_f32): Likewise.
16828 (vmlaq_f64): Likewise.
16829 (vmls_f32): Likewise.
16830 (vmls_f64): Likewise.
16831 (vmlsq_f32): Likewise.
16832 (vmlsq_f64): Likewise.
16833 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
16835 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16837 * config/aarch64/aarch64-simd-builtins.def: Add
16838 float_ml[as]_n_builtin generator macros.
16839 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
16841 (mul_n<mode>3): This, and re-order arguments.
16842 (aarch64_float_mla_n<mode>): Define.
16843 (aarch64_float_mls_n<mode>): Define.
16844 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
16845 instead of inline asm.
16846 (vmlaq_n_f32): Likewise.
16847 (vmls_n_f32): Likewise.
16848 (vmlsq_n_f32): Likewise.
16850 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
16852 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
16853 builtin generator macros.
16854 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
16855 (aarch64_pmull_hiv16qi_insn): Define.
16856 (aarch64_pmull_hiv16qi): Define.
16857 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
16858 instead of inline asm.
16859 (vmull_p8): Likewise.
16861 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
16863 * config/avr/avr.md: Adjust peepholes to match and
16864 generate parallels with clobber of REG_CC.
16865 (mov<mode>_insn): Rename to mov<mode>_insn_split.
16866 (*mov<mode>_insn): Rename to mov<mode>_insn.
16868 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
16870 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
16871 for VAR_DECLs if -fdata-sections enabled.
16873 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
16875 PR bootstrap/100327
16876 * config/rs6000/rs6000.c
16877 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
16878 (rs6000_libgcc_floating_mode_supported_p): New target hook.
16880 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
16882 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
16884 (class back_threader_registry): ...to here...
16885 (class back_threader_profitability): ...and here...
16886 (thread_jumps::thread_through_all_blocks): Remove argument.
16887 (back_threader_registry::back_threader_registry): New.
16888 (back_threader_registry::~back_threader_registry): New.
16889 (back_threader_registry::thread_through_all_blocks): New.
16890 (thread_jumps::profitable_jump_thread_path): Move from here...
16891 (back_threader_profitability::profitable_path_p): ...to here.
16892 (thread_jumps::find_taken_edge): New.
16893 (thread_jumps::convert_and_register_current_path): Move...
16894 (back_threader_registry::register_path): ...to here.
16895 (thread_jumps::register_jump_thread_path_if_profitable): Move...
16896 (thread_jumps::maybe_register_path): ...to here.
16897 (thread_jumps::handle_phi): Call find_taken_edge and
16898 maybe_register_path.
16899 (thread_jumps::handle_assignment): Same.
16900 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
16901 tree argument to handle_phi and handle_assignment.
16902 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
16903 set of m_speed_p and m_max_threaded_paths.
16904 (pass_thread_jumps::execute): Remove second argument from
16905 find_jump_threads_backwards.
16906 (pass_early_thread_jumps::execute): Same.
16908 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
16910 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
16911 (class dom_opt_dom_walker): Initialize some class variables.
16912 (pass_dominator::execute): Pass evrp_range_analyzer and
16913 dom_jump_threader_simplifier to dom_opt_dom_walker.
16914 Adjust for some functions moving into classes.
16915 (simplify_stmt_for_jump_threading): Adjust and move to...
16916 (jump_threader_simplifier::simplify): ...here.
16917 (dom_opt_dom_walker::before_dom_children): Adjust for
16918 m_evrp_range_analyzer.
16919 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
16920 (test_for_singularity): Place in dom_opt_dom_walker class.
16921 (dom_opt_dom_walker::optimize_stmt): The argument
16922 evrp_range_analyzer is now a class field.
16923 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
16924 (thread_jumps::thread_through_all_blocks): New.
16925 (thread_jumps::convert_and_register_current_path): Use m_registry.
16926 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
16927 being in the threader class.
16928 (pass_early_thread_jumps::execute): Same.
16929 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
16930 (jump_threader::jump_threader): ...here.
16931 (threadedge_finalize_values): Move...
16932 (jump_threader::~jump_threader): ...here.
16933 (jump_threader::remove_jump_threads_including): New.
16934 (jump_threader::thread_through_all_blocks): New.
16935 (record_temporary_equivalences_from_phis): Move...
16936 (jump_threader::record_temporary_equivalences_from_phis): ...here.
16937 (record_temporary_equivalences_from_stmts_at_dest): Move...
16938 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
16940 (simplify_control_stmt_condition_1): Move to jump_threader class.
16941 (simplify_control_stmt_condition): Move...
16942 (jump_threader::simplify_control_stmt_condition): ...here.
16943 (thread_around_empty_blocks): Move...
16944 (jump_threader::thread_around_empty_blocks): ...here.
16945 (thread_through_normal_block): Move...
16946 (jump_threader::thread_through_normal_block): ...here.
16947 (thread_across_edge): Move...
16948 (jump_threader::thread_across_edge): ...here.
16949 (thread_outgoing_edges): Move...
16950 (jump_threader::thread_outgoing_edges): ...here.
16951 * tree-ssa-threadedge.h: Move externally facing functings...
16952 (class jump_threader): ...here...
16953 (class jump_threader_simplifier): ...and here.
16954 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
16955 (jump_thread_path_allocator::jump_thread_path_allocator): New.
16956 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
16957 (jump_thread_path_allocator::allocate_thread_edge): New.
16958 (jump_thread_path_allocator::allocate_thread_path): New.
16959 (jump_thread_path_registry::jump_thread_path_registry): New.
16960 (jump_thread_path_registry::~jump_thread_path_registry): New.
16961 (jump_thread_path_registry::allocate_thread_edge): New.
16962 (jump_thread_path_registry::allocate_thread_path): New.
16963 (dump_jump_thread_path): Make extern.
16964 (debug (const vec<jump_thread_edge *> &path)): New.
16965 (struct removed_edges): Move to tree-ssa-threadupdate.h.
16966 (struct thread_stats_d): Remove.
16967 (remove_ctrl_stmt_and_useless_edges): Make static.
16968 (lookup_redirection_data): Move...
16969 (jump_thread_path_registry::lookup_redirection_data): ...here.
16970 (ssa_redirect_edges): Make static.
16971 (thread_block_1): Move...
16972 (jump_thread_path_registry::thread_block_1): ...here.
16973 (thread_block): Move...
16974 (jump_thread_path_registry::thread_block): ...here.
16975 (thread_through_loop_header): Move...
16976 (jump_thread_path_registry::thread_through_loop_header): ...here.
16977 (mark_threaded_blocks): Move...
16978 (jump_thread_path_registry::mark_threaded_blocks): ...here.
16979 (debug_path): Move...
16980 (jump_thread_path_registry::debug_path): ...here.
16981 (debug_all_paths): Move...
16982 (jump_thread_path_registry::dump): ..here.
16983 (rewire_first_differing_edge): Move...
16984 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
16985 (adjust_paths_after_duplication): Move...
16986 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
16987 (duplicate_thread_path): Move...
16988 (jump_thread_path_registry::duplicate_thread_path): ..here.
16989 (remove_jump_threads_including): Move...
16990 (jump_thread_path_registry::remove_jump_threads_including): ...here.
16991 (thread_through_all_blocks): Move to...
16992 (jump_thread_path_registry::thread_through_all_blocks): ...here.
16993 (delete_jump_thread_path): Remove.
16994 (register_jump_thread): Move...
16995 (jump_thread_path_registry::register_jump_thread): ...here.
16996 * tree-ssa-threadupdate.h: Move externally facing functions...
16997 (class jump_thread_path_allocator): ...here...
16998 (class jump_thread_path_registry): ...and here.
16999 (thread_through_all_blocks): Remove.
17000 (struct removed_edges): New.
17001 (register_jump_thread): Remove.
17002 (remove_jump_threads_including): Remove.
17003 (delete_jump_thread_path): Remove.
17004 (remove_ctrl_stmt_and_useless_edges): Remove.
17005 (free_dom_edge_info): New prototype.
17006 * tree-vrp.c: Remove x_vr_values hack.
17007 (class vrp_jump_threader_simplifier): New.
17008 (vrp_jump_threader_simplifier::simplify): New.
17009 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17010 Remove m_dummy_cond.
17011 Instantiate m_simplifier and m_threader.
17012 (vrp_jump_threader::thread_through_all_blocks): New.
17013 (vrp_jump_threader::simplify_stmt): Remove.
17014 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17015 Remove x_vr_values hack.
17016 (execute_vrp): Adjust for thread_through_all_blocks being in a
17019 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17021 * genflags.c (gen_insn): Print failed expansion string.
17023 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17025 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17026 without limit to MAX_FIXED_MODE_SIZE.
17028 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17030 PR middle-end/90773
17031 * builtins.c (builtin_memset_gen_str): Don't use return from
17032 simplify_gen_subreg.
17034 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17037 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17038 (*addsi3_carry_zext_0r): Ditto.
17039 (*sub<mode>3_carry_0): Ditto.
17040 (*subsi3_carry_zext_0r): Ditto.
17041 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17043 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17044 Also consider ix86_carry_flag_unset_operator to calculate
17045 the cost of adc/sbb insn.
17047 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17049 PR rtl-optimization/100225
17050 PR rtl-optimization/84878
17051 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17052 where we have an instruction which touches (writes) any hard
17053 register from df->regular_block_artificial_uses set.
17054 Allow not-single-set instruction only right before basic block
17057 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17059 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17061 2021-04-30 LevyHsu <admin@levyhsu.com>
17063 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17064 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17065 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17066 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17068 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17070 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17072 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17074 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17075 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17076 * config/i386/i386.c: Adjust.
17077 * config/i386/i386.md: Adjust.
17078 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17079 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17080 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17081 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17082 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17083 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17084 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17085 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17086 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17087 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17088 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17090 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17092 * config/i386/i386-expand.c (ix86_expand_int_compare):
17093 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17094 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17095 predicate to allow more combine opportunities with memory operands.
17096 (*sub<mode>3_carry_0): Ditto.
17098 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17100 PR rtl-optimization/100303
17101 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17102 boolean that indicates whether the use will only be used in
17103 debug instructions. Treat it in the same way that existing
17104 cross-EBB debug references would be handled if so.
17105 (function_info::make_uses_available): Likewise.
17106 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17107 prototype accordingly.
17108 (function_info::make_uses_available): Likewise.
17109 * fwprop.c (try_fwprop_subst): Update call accordingly.
17111 2021-04-29 Jeff Law <jlaw@tachyum.com>
17113 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17116 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17117 Richard Biener <rguenther@suse.de>
17120 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17121 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17122 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17123 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17124 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17125 to PURE_ARGS category.
17126 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17127 Handle PURE_ARGS category.
17128 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17130 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17132 * configure.ac: Check for the presence of sys/locking.h header and
17133 for whether _LK_LOCK is supported by _locking.
17134 * configure: Regenerate.
17135 * config.in: Likewise.
17136 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17137 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17138 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17140 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17142 * config/i386/predicates.md (fcmov_comparison_operator):
17143 Do not check for trivial FP comparison operator.
17144 <case GEU, case LTU>: Allow CCGZmode.
17145 <case GTU, case LEU>: Do not allow CCCmode.
17146 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17147 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17148 Do not check for trivial FP comparison operator. Allow CCGZmode.
17150 2021-04-29 Tom de Vries <tdevries@suse.de>
17152 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17153 fd->loop.step by either step or orig_step.
17155 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17157 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17158 (load_got_register): Do the PIC dance here.
17159 (sparc_legitimize_tls_address): Simplify.
17160 (sparc_emit_probe_stack_range): Likewise.
17161 (sparc32_initialize_trampoline): Likewise.
17162 (sparc64_initialize_trampoline): Likewise.
17163 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17164 (probe_stack_range<P:mode>): Likewise.
17165 (flush<P:mode>): Likewise.
17166 (tgd_hi22<P:mode>): Likewise.
17167 (tgd_lo10<P:mode>): Likewise.
17168 (tgd_add<P:mode>): Likewise.
17169 (tgd_call<P:mode>): Likewise.
17170 (tldm_hi22<P:mode>): Likewise.
17171 (tldm_lo10<P:mode>): Likewise.
17172 (tldm_add<P:mode>): Likewise.
17173 (tldm_call<P:mode>): Likewise.
17174 (tldo_hix22<P:mode>): Likewise.
17175 (tldo_lox10<P:mode>): Likewise.
17176 (tldo_add<P:mode>): Likewise.
17177 (tie_hi22<P:mode>): Likewise.
17178 (tie_lo10<P:mode>): Likewise.
17179 (tie_add<P:mode>): Likewise.
17180 (tle_hix22<P:mode>): Likewise.
17181 (tle_lox10<P:mode>): Likewise.
17182 (stack_protect_setsi): Rename to...
17183 (stack_protect_set32): ...this.
17184 (stack_protect_setdi): Rename to...
17185 (stack_protect_set64): ...this.
17186 (stack_protect_set): Adjust calls to above.
17187 (stack_protect_testsi): Rename to...
17188 (stack_protect_test32): ...this.
17189 (stack_protect_testdi): Rename to...
17190 (stack_protect_test64): ...this.
17191 (stack_protect_test): Adjust calls to above.
17193 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17195 PR middle-end/90773
17196 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17197 (builtin_strncpy_read_str): Likewise.
17198 (builtin_memset_read_str): Add an argument for the previous RTL
17199 information and generate the new RTL from the previous RTL info.
17200 (builtin_memset_gen_str): Likewise.
17201 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17202 (builtin_memset_read_str): Likewise.
17203 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17204 returns true, round up size and alignment to the widest integer
17205 mode for maximum size.
17206 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17207 and pass it to m_constfn.
17208 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17209 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17210 initialize m_push. Initialize m_overlap_op_by_pieces with
17211 targetm.overlap_op_by_pieces_p ().
17212 (op_by_pieces_d::run): Pass the previous RTL information to
17213 pieces_addr::adjust and generate overlapping operations if
17214 m_overlap_op_by_pieces is true.
17216 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17218 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17220 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17221 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17222 for op_by_pieces_d change.
17223 (clear_by_pieces_1): Add a dummy argument.
17224 (clear_by_pieces): Updated for op_by_pieces_d change.
17225 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17226 (string_cst_read_str): Add a dummy argument.
17227 * expr.h (by_pieces_constfn): Add a dummy argument.
17228 (by_pieces_prev): New.
17229 * target.def (overlap_op_by_pieces_p): New target hook.
17230 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17231 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17232 * doc/tm.texi: Regenerated.
17234 2021-04-29 Richard Biener <rguenther@suse.de>
17236 PR tree-optimization/100253
17237 * tree-vect-stmts.c (vectorizable_load): Do not assume
17238 element alignment when DR_MISALIGNMENT is -1.
17239 (vectorizable_store): Likewise.
17241 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17244 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17245 absu_hwi instead of abs_hwi.
17247 2021-04-29 Richard Biener <rguenther@suse.de>
17249 PR middle-end/38474
17250 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17251 forwarding when indirect forwarding through ESCAPED
17254 2021-04-29 Tom de Vries <tdevries@suse.de>
17257 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17258 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17259 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17260 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17262 2021-04-29 Richard Biener <rguenther@suse.de>
17264 PR tree-optimization/99912
17265 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17266 (dse_dom_walker::todo): Likewise.
17267 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17269 (dse_dom_walker::before_dom_children): Remove trivially
17270 dead SSA defs and schedule CFG cleanup if we removed all
17272 (pass_dse::execute): Get TODO as computed by the DOM walker
17273 and return it. Wipe dominator info earlier.
17275 2021-04-29 Richard Biener <rguenther@suse.de>
17278 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17279 Track blocks to cleanup EH in new m_need_eh_cleanup.
17280 (ipcp_modif_dom_walker::cleanup_eh): New.
17281 (ipcp_transform_function): Release dominator info before
17284 2021-04-29 Martin Sebor <msebor@redhat.com>
17286 PR middle-end/100250
17287 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17288 a pointer when it's null.
17290 2021-04-29 Martin Sebor <msebor@redhat.com>
17292 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17293 * ipa-free-lang-data.cc: New file.
17294 * tree.c: Move pass free_lang_data to file above.
17295 (build_array_type_1): Declare extern.
17296 * tree.h (build_array_type_1): Declare.
17298 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17300 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17301 make consistent with updated RTL pattern.
17302 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17303 Implement using ss_truncate and us_truncate rather than
17305 * config/aarch64/iterators.md: Remove redundant unspecs and
17306 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17308 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17310 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17311 attributes consistent with those defined in arm_neon.h.
17313 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17315 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17316 attributes consistent with those defined in arm_neon.h.
17318 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17320 * config/aarch64/aarch64-simd-builtins.def: Add
17321 float_trunc_rodd builtin generator macros.
17322 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17324 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17325 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17326 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17327 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17328 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17329 instead of inline asm.
17330 (vcvtx_high_f32_f64): Likewise.
17331 (vcvtxd_f32_f64): Likewise.
17332 * config/aarch64/iterators.md: Add FCVTXN unspec.
17334 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17336 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
17338 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
17340 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
17341 instead of inline asm.
17342 (vqtbx1_u8): Likewise.
17343 (vqtbx1_p8): Likewise.
17344 (vqtbx1q_s8): Likewise.
17345 (vqtbx1q_u8): Likewise.
17346 (vqtbx1q_p8): Likewise.
17347 (vtbx2_s8): Likewise.
17348 (vtbx2_u8): Likewise.
17349 (vtbx2_p8): Likewise.
17351 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17353 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
17355 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
17356 instead of inline asm.
17357 (vqtbl1_s8): Likewise.
17358 (vqtbl1_u8): Likewise.
17359 (vqtbl1q_p8): Likewise.
17360 (vqtbl1q_s8): Likewise.
17361 (vqtbl1q_u8): Likewise.
17362 (vtbl1_s8): Likewise.
17363 (vtbl1_u8): Likewise.
17364 (vtbl1_p8): Likewise.
17365 (vtbl2_s8): Likewise.
17366 (vtbl2_u8): Likewise.
17367 (vtbl2_p8): Likewise.
17369 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17371 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
17372 ssri_n buitin generator macro.
17373 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
17374 instead of inline asm.
17375 (vsri_n_p16): Likewise.
17376 (vsri_n_p64): Likewise.
17377 (vsriq_n_p8): Likewise.
17378 (vsriq_n_p16): Likewise.
17379 (vsriq_n_p64): Likewise.
17381 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17383 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
17384 iterator for polynomial ssli_n builtin generator macro.
17385 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
17386 instead of inline asm.
17387 (vsli_n_p16): Likewise.
17388 (vsliq_n_p8): Likewise.
17389 (vsliq_n_p16): Likewise.
17390 * config/aarch64/iterators.md: Define VALLP mode iterator.
17392 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17394 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
17395 iterator to generate [su]adalp RTL builtins.
17396 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
17397 [su]adalp RTL pattern.
17398 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
17399 instead of inline asm.
17400 (vpadal_u32): Likewise.
17402 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17404 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
17405 builtin generator macros.
17406 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
17408 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
17409 instead of inline asm.
17410 (vpaddl_s16): Likewise.
17411 (vpaddl_s32): Likewise.
17412 (vpaddl_u8): Likewise.
17413 (vpaddl_u16): Likewise.
17414 (vpaddl_u32): Likewise.
17415 (vpaddlq_s8): Likewise.
17416 (vpaddlq_s16): Likewise.
17417 (vpaddlq_s32): Likewise.
17418 (vpaddlq_u8): Likewise.
17419 (vpaddlq_u16): Likewise.
17420 (vpaddlq_u32): Liwewise.
17421 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
17422 appropriate attributes.
17424 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17426 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
17427 for aarch64_addp<mode> builtin macro generator.
17428 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
17429 aarch64_addp<mode> RTL pattern.
17430 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
17431 instead of inline asm.
17432 (vpaddq_s16): Likewise.
17433 (vpaddq_s32): Likewise.
17434 (vpaddq_s64): Likewise.
17435 (vpaddq_u8): Likewise.
17436 (vpaddq_u16): Likewise.
17437 (vpaddq_u32): Likewise.
17438 (vpaddq_u64): Likewise.
17440 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17442 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
17443 builtin generator macros.
17444 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
17446 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
17447 instead of inline asm.
17448 (vqdmulh_n_s32): Likewise.
17449 (vqdmulhq_n_s16): Likewise.
17450 (vqdmulhq_n_s32): Likewise.
17451 (vqrdmulh_n_s16): Likewise.
17452 (vqrdmulh_n_s32): Likewise.
17453 (vqrdmulhq_n_s16): Likewise.
17454 (vqrdmulhq_n_s32): Likewise.
17456 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
17458 * doc/install.texi (--enable-offload-defaulted): Document.
17460 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17462 * config/avr/avr-dimode.md: Turn existing patterns into
17463 define_insn_and_split style patterns where the splitter
17464 adds a clobber of the condition code register. Drop "cc"
17465 attribute. Add new patterns to match output of
17467 * config/avr/avr-fixed.md: Likewise.
17468 * config/avr/avr.c (cc_reg_rtx): New.
17469 (avr_parallel_insn_from_insns): Adjust insn count
17470 for removal of set of cc0.
17471 (avr_is_casesi_sequence): Likewise.
17472 (avr_casei_sequence_check_operands): Likewise.
17473 (avr_optimize_casesi): Likewise. Also insert
17474 new insns after jump_insn.
17475 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
17476 for removal of set of cc0.
17477 (avr_init_expanders): Initialize cc_reg_rtx.
17478 (avr_regno_reg_class): Handle REG_CC.
17479 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
17480 (avr_notice_update_cc): Remove function.
17481 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
17482 (compare_condition): Adjust for PARALLEL with
17484 (out_shift_with_cnt): Likewise.
17485 (ashlhi3_out): Likewise.
17486 (ashrhi3_out): Likewise.
17487 (lshrhi3_out): Likewise.
17488 (avr_class_max_nregs): Return single reg for REG_CC.
17489 (avr_compare_pattern): Check for REG_CC instead
17491 (avr_reorg_remove_redundant_compare): Likewise.
17492 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
17493 (avr_hard_regno_nregs): Return single reg for REG_CC.
17494 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
17495 (avr_md_asm_adjust): Clobber REG_CC.
17496 (TARGET_HARD_REGNO_NREGS): Define.
17497 (TARGET_CLASS_MAX_NREGS): Define.
17498 (TARGET_MD_ASM_ADJUST): Define.
17499 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
17501 (enum reg_class): Add CC_REG class.
17502 (NOTICE_UPDATE_CC): Remove.
17503 (CC_OVERFLOW_UNUSABLE): Remove.
17504 (CC_NO_CARRY): Remove.
17505 * config/avr/avr.md: Turn existing patterns into
17506 define_insn_and_split style patterns where the splitter
17507 adds a clobber of the condition code register. Drop "cc"
17508 attribute. Add new patterns to match output of
17510 (sez): Remove unused pattern.
17512 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
17515 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
17518 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
17521 * config/aarch64/constraints.md (Utq): Require the address to
17522 be valid for both the element mode and for V2DImode.
17524 2021-04-28 Jakub Jelinek <jakub@redhat.com>
17525 Tobias Burnus <tobias@codesourcery.com>
17527 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
17528 * gcc.c (process_command): New variable.
17529 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
17530 set it if -foffload is defaulted.
17531 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
17532 (compile_offload_image): If OFFLOAD_DEFAULTED and
17533 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
17534 if corresponding mkoffload can't be found.
17535 (compile_images_for_offload_targets): Likewise. Free and clear
17536 offload_names if no valid offload is found.
17537 * config.in: Regenerate.
17538 * configure: Regenerate.
17540 2021-04-28 Richard Biener <rguenther@suse.de>
17542 PR tree-optimization/100292
17543 * tree-vect-generic.c (expand_vector_condition): Do not fold
17546 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17548 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
17549 * config/rs6000/aix64.opt (m64): New.
17552 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17554 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
17555 (index_term_p): Handle ASHIFT too.
17557 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17559 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
17560 (sync_lock_test_and_set<mode>): Adjust accordingly.
17561 (sync_lock_release<mode>): Likewise.
17563 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17565 * config/vax/vax-protos.h (adjacent_operands_p): Remove
17567 * config/vax/vax.c (adjacent_operands_p): Remove.
17569 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
17571 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
17572 through to the non-conditional execution case if getting the
17573 condition for conditional execution has failed.
17575 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17577 PR middle-end/100284
17578 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
17579 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
17580 than asserting on it.
17582 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17584 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
17585 with TARGET_AIX_OS.
17587 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17590 * calls.c (precompute_register_parameters): Additionally test
17591 targetm.precompute_tls_p to pre-compute argument.
17592 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
17593 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
17594 * target.def (precompute_tls_p): New.
17595 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
17596 * doc/tm.texi: Regenerated.
17598 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17601 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
17602 back to HOST_WIDE_INT.
17604 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17607 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
17608 memory alignment for the outer mode.
17610 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
17612 PR middle-end/90773
17613 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
17614 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
17616 2021-04-27 Alex Coplan <alex.coplan@arm.com>
17619 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
17620 with negative immediates: ensure we expand cbranchsi4_scratch
17621 correctly and ensure we satisfy its constraints.
17622 * config/arm/sync.md
17623 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
17624 attempt to tie two output operands together with constraints;
17625 collapse two alternatives.
17626 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
17627 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
17629 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17632 * config/aarch64/predicates.md (aarch64_sub_immediate,
17633 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
17634 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
17635 * config/aarch64/aarch64.c (aarch64_print_operand,
17636 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
17638 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17640 PR tree-optimization/100239
17641 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
17642 permutations with all indices from the first zero element as vec_shl.
17644 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17646 PR rtl-optimization/100254
17647 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
17648 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
17650 2021-04-27 Richard Biener <rguenther@suse.de>
17652 PR tree-optimization/99912
17653 * passes.def: Add comment about new TODO_remove_unused_locals.
17654 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
17657 2021-04-27 Richard Biener <rguenther@suse.de>
17659 PR tree-optimization/99912
17660 * passes.def (pass_all_optimizations): Add pass_dse before
17661 the first pass_dce, move the first pass_dse before the
17662 pass_dce following pass_pre.
17664 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17666 PR tree-optimization/95527
17667 * generic-match-head.c: Include tm.h.
17668 * gimple-match-head.c: Include tm.h.
17669 * match.pd (CLZ == INTEGER_CST): Don't use
17670 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
17671 if clz == CFN_CLZ. Add missing val declaration.
17672 (CTZ cmp CST): New simplifications.
17674 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17676 PR tree-optimization/96696
17677 * expr.c (expand_expr_divmod): New function.
17678 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
17679 divisions. Formatting fixes.
17680 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
17683 2021-04-27 Martin Jambor <mjambor@suse.cz>
17686 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
17687 If removing a call statement LHS SSA name, release it.
17689 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
17692 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
17693 is valid before including it in the mask.
17695 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17698 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
17701 2021-04-27 Richard Biener <rguenther@suse.de>
17703 PR tree-optimization/100051
17704 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
17705 disambiguator based on access size vs. decl size.
17707 2021-04-27 Richard Biener <rguenther@suse.de>
17709 PR tree-optimization/100278
17710 * tree-ssa-pre.c (compute_avail): Give up when we cannot
17711 adjust TBAA beacuse of mismatching bases.
17713 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17716 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
17717 For any_rotate define_insn_split and following splitters, use
17718 SWI iterator instead of SWI48.
17720 2021-04-27 Richard Biener <rguenther@suse.de>
17722 PR tree-optimization/99776
17723 * match.pd (bit_field_ref (ctor)): Relax element extract
17724 type compatibility checks.
17726 2021-04-27 Cui,Lili <lili.cui@intel.com>
17728 * common/config/i386/i386-common.c (processor_names):
17729 Sync processor_names with processor_type.
17730 * config/i386/i386-options.c (processor_cost_table):
17731 Sync processor_cost_table with processor_type.
17733 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17735 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
17736 (irange::set): Call irange_set_1bit_anti_range for handling all
17737 1-bit ranges. Fall through on ~[MIN,MAX].
17739 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17741 * value-range.cc (irange::legacy_num_pairs): Remove.
17742 (irange::invert): Change gcc_assert to gcc_checking_assert.
17743 * value-range.h (irange::num_pairs): Adjust for a cached
17744 num_pairs(). Also, rename all gcc_assert's to
17745 gcc_checking_assert's.
17747 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17749 * value-range.cc (irange::operator=): Set m_kind.
17750 (irange::copy_to_legacy): Handle varying and undefined sources
17751 as a legacy copy since they can be easily copied.
17752 (irange::irange_set): Set m_kind.
17753 (irange::irange_set_anti_range): Same.
17754 (irange::set): Rename normalize_min_max to normalize_kind.
17755 (irange::verify_range): Adjust for multi-ranges having the
17757 (irange::irange_union): Set m_kind.
17758 (irange::irange_intersect): Same.
17759 (irange::invert): Same.
17760 * value-range.h (irange::kind): Always return m_kind.
17761 (irange::varying_p): Rename to...
17762 (irange::varying_comptaible_p): ...this.
17763 (irange::undefined_p): Only look at m_kind.
17764 (irange::irange): Always set VR_UNDEFINED if applicable.
17765 (irange::set_undefined): Always set VR_UNDEFINED.
17766 (irange::set_varying): Always set m_kind to VR_VARYING.
17767 (irange::normalize_min_max): Rename to...
17768 (irange::normalize_kind): ...this.
17770 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17772 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
17773 Adjust for constant_p including varying_p.
17774 * tree-vrp.c (vrp_prop::finalize): Same.
17775 (determine_value_range): Same.
17776 * vr-values.c (vr_values::range_of_expr): Same.
17777 * value-range.cc (irange::symbolic_p): Do not check varying_p.
17778 (irange::constant_p): Same.
17780 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17782 * value-range.cc (irange::legacy_lower_bound): Replace
17783 !undefined_p check with num_ranges > 0.
17784 (irange::legacy_upper_bound): Same.
17785 * value-range.h (irange::type): Same.
17786 (irange::lower_bound): Same.
17787 (irange::upper_bound): Same.
17789 2021-04-26 Richard Biener <rguenther@suse.de>
17791 PR tree-optimization/99956
17792 * gimple-loop-interchange.cc (compute_access_stride):
17793 Try instantiating the access in a shallower loop nest
17794 if instantiating failed.
17795 (compute_access_strides): Pass adjustable loop_nest
17796 to compute_access_stride.
17798 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
17800 * doc/sourcebuild.texi (arm_cmse_hw): Document.
17802 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17804 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
17806 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
17807 Nathan Sidwell <nathan@codesourcery.com>
17808 Tom de Vries <vries@codesourcery.com>
17809 Julian Brown <julian@codesourcery.com>
17810 Kwok Cheung Yeung <kcy@codesourcery.com>
17812 * omp-offload.c (oacc_validate_dims): Implement
17813 '-Wopenacc-parallelism'.
17814 * doc/invoke.texi (-Wopenacc-parallelism): Document.
17816 2021-04-26 Richard Biener <rguenther@suse.de>
17818 * tree-cfg.h (gimplify_build1): Remove.
17819 (gimplify_build2): Likewise.
17820 (gimplify_build3): Likewise.
17821 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
17822 (gimplify_build2): Likewise.
17823 (gimplify_build3): Likewise.
17824 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
17826 (gimplify_build2): Likewise.
17827 (gimplify_build3): Likewise.
17828 (tree_vec_extract): Use resimplify with following SSA edges.
17829 (expand_vector_parallel): Avoid passing NULL size/bitpos
17830 to tree_vec_extract.
17831 * expr.c (store_constructor): Deal with zero-element CTORs.
17832 * match.pd (bit_field_ref <vector CTOR>): Make sure to
17833 produce vector constants when possible.
17835 2021-04-26 Richard Biener <rguenther@suse.de>
17837 * tree-complex.c: Include gimple-fold.h.
17838 (expand_complex_addition): Use gimple_build.
17839 (expand_complex_multiplication_components): Likewise.
17840 (expand_complex_multiplication): Likewise.
17841 (expand_complex_div_straight): Likewise.
17842 (expand_complex_div_wide): Likewise.
17843 (expand_complex_division): Likewise.
17844 (expand_complex_conjugate): Likewise.
17845 (expand_complex_comparison): Likewise.
17847 2021-04-26 Richard Biener <rguenther@suse.de>
17849 * tree-ssa-phiopt.c (two_value_replacement): Remove use
17850 of legacy gimplify_buildN API.
17852 2021-04-26 Richard Biener <rguenther@suse.de>
17854 PR tree-optimization/99473
17855 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
17858 2021-04-26 Richard Biener <rguenther@suse.de>
17860 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
17861 Use replace_call_with_value.
17863 2021-04-26 Richard Biener <rguenther@suse.de>
17865 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
17866 (update_gimple_call): Likewise.
17867 (update_call_from_tree): Likewise.
17868 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
17869 (valid_gimple_call_p): Likewise.
17870 (move_ssa_defining_stmt_for_defs): Likewise.
17871 (finish_update_gimple_call): Likewise.
17872 (update_gimple_call): Likewise.
17873 (update_call_from_tree): Likewise.
17874 (propagate_tree_value_into_stmt): Use replace_call_with_value.
17875 * gimple-fold.h (update_gimple_call): Declare.
17876 * gimple-fold.c (valid_gimple_rhs_p): Move here from
17877 tree-ssa-propagate.c.
17878 (update_gimple_call): Likewise.
17879 (valid_gimple_call_p): Likewise.
17880 (finish_update_gimple_call): Likewise, and simplify.
17881 (gimplify_and_update_call_from_tree): Implement
17882 update_call_from_tree functionality, avoid excessive
17883 push/pop_gimplify_context.
17884 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
17885 (gimple_fold_call): Likewise.
17886 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
17887 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
17888 (pass_fold_builtins::execute): Likewise.
17889 (optimize_stack_restore): Use replace_call_with_value.
17890 * tree-cfg.c (fold_loop_internal_call): Likewise.
17891 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
17892 only gimplify_and_update_call_from_tree.
17893 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
17894 (handle_builtin_strchr): Likewise.
17895 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
17897 2021-04-26 Jakub Jelinek <jakub@redhat.com>
17900 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
17901 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
17904 2021-04-25 liuhongt <hongtao.liu@intel.com>
17907 * config/i386/i386-builtin.def (BDESC): Change the icode of
17908 the following builtins to CODE_FOR_nothing.
17909 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
17910 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
17911 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
17912 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
17913 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
17914 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
17915 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
17916 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
17917 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
17918 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
17919 (sse2_eq<mode>3): Ditto.
17920 (sse4_1_eqv2di3): Ditto.
17921 (sse2_gt<mode>3): Rename to ..
17922 (*sse2_gt<mode>3): .. this.
17924 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
17927 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
17930 * config/darwin.c (darwin_binds_local_p): Assume that any
17931 public symbol might be interposed for PIC code. Update function
17932 header comment to reflect current Darwin capability.
17934 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
17937 * config/darwin.c (darwin_binds_local_p): Assume that any
17938 public symbol might be interposed for PIC code. Update function
17939 header comment to reflect current Darwin capability.
17941 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
17943 * doc/sourcebuild.texi: Document no-opts and any-opts target
17946 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
17948 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
17950 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
17952 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
17954 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
17957 * config/i386/i386-options.c (ix86_option_override_internal):
17958 Error out when -m96bit-long-double is used with 64bit targets.
17959 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
17961 2021-04-23 Martin Liska <mliska@suse.cz>
17963 * lto-wrapper.c: Remove FIXME about usage of
17964 hardware_concurrency. The function is not on par with
17967 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
17970 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
17971 Copy operand 3 to operand 4. Use sse_reg_operand
17972 as operand 3 predicate.
17973 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
17974 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
17975 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
17976 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
17977 Copy operand 1 to operand 0.
17978 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
17979 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
17980 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
17982 2021-04-23 Alex Coplan <alex.coplan@arm.com>
17984 PR rtl-optimization/100230
17985 * early-remat.c (early_remat::sort_candidates): Use delete[]
17986 instead of delete for array allocated with new[].
17988 2021-04-23 Richard Biener <rguenther@suse.de>
17990 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
17991 (capture_info::capture_info): Likewise.
17992 (capture_info::walk_match): Likewise.
17993 (expr::gen_transform): Likewise.
17994 (dt_simplify::gen_1): Likewise.
17995 * gimple-match-head.c (maybe_resimplify_conditional_op):
17996 Remove VEC_COND_EXPR special-casing.
17997 (gimple_simplify): Likewise.
17998 * gimple.c (gimple_could_trap_p_1): Adjust.
17999 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18000 to participate in PRE.
18002 2021-04-23 Richard Biener <rguenther@suse.de>
18004 * cfganal.c (connect_infinite_loops_to_exit): First call
18005 add_noreturn_fake_exit_edges.
18006 * ipa-sra.c (process_scan_results): Do not call the now redundant
18007 add_noreturn_fake_exit_edges.
18008 * predict.c (tree_estimate_probability): Likewise.
18009 (rebuild_frequencies): Likewise.
18010 * store-motion.c (one_store_motion_pass): Likewise.
18012 2021-04-23 Richard Biener <rguenther@suse.de>
18014 PR tree-optimization/100222
18015 * predict.c (pass_profile::execute): Remove redundant call to
18016 mark_irreducible_loops.
18017 (report_predictor_hitrates): Likewise.
18019 2021-04-23 Richard Biener <rguenther@suse.de>
18021 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18022 valid_gimple_rhs_p by instead gimplifying to one.
18024 2021-04-23 Richard Biener <rguenther@suse.de>
18026 PR tree-optimization/99971
18027 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18028 Always use TBAA for loads.
18030 2021-04-23 liuhongt <hongtao.liu@intel.com>
18033 * config/i386/i386-options.c (ix86_option_override_internal):
18034 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18035 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18036 by target attribute.
18038 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18040 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18042 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18044 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18046 * config.gcc (powerpc-ibm-aix6.*): Remove.
18047 * config/rs6000/aix61.h: Delete.
18049 2021-04-22 Martin Liska <mliska@suse.cz>
18051 PR testsuite/100159
18052 PR testsuite/100192
18053 * builtins.c (expand_builtin): Fix typos and missing comments.
18054 * dwarf2out.c (gen_subprogram_die): Likewise.
18055 (gen_struct_or_union_type_die): Likewise.
18057 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18060 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18061 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18063 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18065 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18066 based on TARGET_ISA_64BIT.
18067 (TARGET_64BIT_P): Remove definition based on
18068 TARGET_ISA_64BIT_P().
18070 2021-04-21 Martin Liska <mliska@suse.cz>
18073 2021-04-21 Martin Liska <mliska@suse.cz>
18075 * lto-wrapper.c (cpuset_popcount): Remove.
18076 (init_num_threads): Remove and use hardware_concurrency.
18078 2021-04-21 Martin Liska <mliska@suse.cz>
18081 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18082 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18083 when incremental LTO linking happens.
18085 2021-04-21 Martin Liska <mliska@suse.cz>
18087 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18088 makeserver cannot be detected, then use -flto=N fallback.
18090 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18092 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18093 default to yes for aarch64-linux-gnu.
18094 * configure: Regenerate.
18096 2021-04-21 Martin Liska <mliska@suse.cz>
18098 * lto-wrapper.c (cpuset_popcount): Remove.
18099 (init_num_threads): Remove and use hardware_concurrency.
18101 2021-04-21 Martin Liska <mliska@suse.cz>
18103 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18104 which remains to be '(... || 0)' and clang complains about it.
18105 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18106 (add_AT_vms_delta): Likewise.
18107 * tree.c (fld_simplified_type): Use rather more common pattern
18108 for disabling of something (#if 0).
18109 (get_tree_code_name): Likewise.
18110 (verify_type_variant): Likewise.
18112 2021-04-21 Martin Liska <mliska@suse.cz>
18114 * config/i386/i386-expand.c (decide_alignment): Use newly named
18115 macro TARGET_CPU_P.
18116 * config/i386/i386.c (ix86_decompose_address): Likewise.
18117 (ix86_address_cost): Likewise.
18118 (ix86_lea_outperforms): Likewise.
18119 (ix86_avoid_lea_for_addr): Likewise.
18120 (ix86_add_stmt_cost): Likewise.
18121 * config/i386/i386.h (TARGET_*): Remove.
18122 (TARGET_CPU_P): New macro.
18123 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18124 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18125 (swap_top_of_ready_list): Likewise.
18126 (ix86_atom_sched_reorder): Likewise.
18127 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18128 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18130 2021-04-21 Martin Liska <mliska@suse.cz>
18132 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18134 (SET_TARGET_NO_SAHF): Likewise.
18135 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18136 (SET_TARGET_PREFETCH_SSE): Likewise.
18137 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18138 (SET_TARGET_NO_TUNE): Likewise.
18139 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18140 (SET_TARGET_NO_80387): Likewise.
18142 * config/i386/i386.h (TARGET_*): Remove.
18143 * opth-gen.awk: Generate new used macros.
18145 2021-04-21 Martin Liska <mliska@suse.cz>
18147 * config/i386/i386.h (PTA_*): Remove.
18148 (enum pta_flag): New.
18149 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18150 * config/i386/i386-isa.def: New file.
18152 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18155 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18156 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18157 jump table targets.
18159 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18161 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18162 x86_64-*-* targets.
18163 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18165 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18166 (ix86_handle_option): Handle -mmwait.
18167 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18168 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18169 __builtin_ia32_monitor and __builtin_ia32_mwait.
18170 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18171 (ix86_valid_target_attribute_inner_p): Likewise.
18172 (ix86_option_override_internal): Enable mwait/monitor
18173 instructions for -msse3.
18174 * config/i386/i386.h (TARGET_MWAIT): New.
18175 (TARGET_MWAIT_P): Likewise.
18176 * config/i386/i386.opt: Add -mmwait.
18177 * config/i386/mwaitintrin.h: New file.
18178 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18179 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18181 (@sse3_monitor_<mode>): Likewise.
18182 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18183 * doc/extend.texi: Document mwait target attribute.
18184 * doc/invoke.texi: Document -mmwait.
18186 2021-04-21 Martin Liska <mliska@suse.cz>
18188 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18189 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18190 * config/i386/stringop.def (DEF_ENUM): Likewise.
18192 2021-04-21 Martin Liska <mliska@suse.cz>
18194 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18195 of print_generic_expr.
18197 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18199 PR rtl-optimization/100148
18200 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18203 2021-04-21 Martin Liska <mliska@suse.cz>
18206 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18207 free_dominance_info calls.
18209 2021-04-21 Richard Biener <rguenther@suse.de>
18211 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18212 parameter (and assume it to be false).
18213 (fold_gimple_assign): Adjust, remove all callers of
18214 maybe_fold_reference calling it with is_lhs true.
18215 (gimple_fold_call): Likewise.
18216 (fold_stmt_1): Likewise.
18218 2021-04-21 Richard Biener <rguenther@suse.de>
18220 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18221 (fold_binary_loc): Adjust.
18222 (fold_ternary_loc): Likewise.
18224 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18226 PR middle-end/100130
18227 * varasm.c (get_block_for_decl): Make sure that any use of the
18228 retain attribute matches the section's retain flag.
18229 (switch_to_section): Check for retain mismatches even when
18230 changing sections, but do not warn if the given decl is the
18231 section's named.decl.
18232 (output_object_block): Pass the first decl in the block (if any)
18233 to switch_to_section.
18235 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18237 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18238 __CRC32__ for -mcrc32.
18239 * config/i386/i386-options.c (ix86_option_override_internal):
18240 Enable crc32 instruction for -msse4.2.
18241 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18243 (sse4_2_crc32di): Likewise.
18244 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18247 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18250 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18253 2021-04-20 Martin Liska <mliska@suse.cz>
18255 * doc/invoke.texi: Fix typo.
18256 * params.opt: Likewise.
18258 2021-04-20 Martin Liska <mliska@suse.cz>
18260 * doc/invoke.texi: Document new param.
18262 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18264 PR tree-optimization/100081
18265 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18266 rather than gori_compute_cache.
18267 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18268 (range_def_chain::m_logical_depth): New member.
18269 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18270 (range_def_chain::get_def_chain): Don't build defchains through more
18271 than LOGICAL_LIMIT logical expressions.
18272 * params.opt (param_ranger_logical_depth): New.
18274 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18277 * config/arm/arm.c (arm_configure_build_target): Do not strip
18278 extended FPU/SIMD feature bits from the target ISA when -mfpu
18279 is specified (partial revert of r11-8168).
18281 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18283 * params.opt (-param=openacc-kernels=): Add.
18284 * omp-oacc-kernels-decompose.cc
18285 (pass_omp_oacc_kernels_decompose::gate): Use it.
18286 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18287 (--param): ... here, 'openacc-kernels'.
18289 2021-04-19 Martin Liska <mliska@suse.cz>
18292 * gengtype.c (finish_root_table): Align function arguments
18293 in between declaration and definition.
18295 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18297 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18298 frames larger than the SEH maximum frame size.
18300 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18302 PR rtl-optimization/99927
18303 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18304 is dead, just drop it.
18306 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18309 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18310 * doc/tm.texi: Regenerate.
18311 * doc/tm.texi.in (D language and ABI): Add @hook for
18312 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18314 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18316 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18318 (darwin_d_register_target_info): New function.
18319 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18320 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18322 (dragonfly_d_register_target_info): New function.
18323 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18324 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18326 (freebsd_d_register_target_info): New function.
18327 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18328 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18330 (glibc_d_register_target_info): New function.
18331 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18332 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
18334 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
18335 as handler for objectFormat key.
18336 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
18338 (winnt_d_register_target_info): New function.
18339 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18340 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
18342 (netbsd_d_register_target_info): New function.
18343 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18344 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
18346 (openbsd_d_register_target_info): New function.
18347 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18348 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
18349 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
18350 handler for objectFormat key.
18351 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
18353 (rs6000_d_register_target_info): Add
18354 rs6000_d_handle_target_object_format as handler for objectFormat key.
18355 * config/sol2-d.c (solaris_d_handle_target_object_format): New
18357 (solaris_d_register_target_info): New function.
18358 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18360 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18363 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
18364 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
18366 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
18367 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
18368 the old and new alignment after applying MIN/MAX to it is different.
18370 2021-04-16 Tamar Christina <tamar.christina@arm.com>
18373 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
18374 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
18376 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
18378 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
18380 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
18381 this section and its subsections.
18383 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18386 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
18387 define_insn patterns.
18389 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18391 PR rtl-optimization/98689
18392 * reg-notes.def (UNTYPED_CALL): New note.
18393 * combine.c (distribute_notes): Handle it.
18394 * emit-rtl.c (try_split): Likewise.
18395 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
18396 that calls with the note implicitly set all return value registers.
18397 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
18400 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18402 PR rtl-optimization/99596
18403 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
18404 register accesses for const calls. Assume that pure functions
18405 can only read from global registers. Ignore cases in which
18406 the stack pointer has been marked global.
18408 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18411 * tree-vect-loop.c (vect_transform_loop): Don't remove just
18412 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
18413 them by their last argument.
18415 2021-04-15 Martin Liska <mliska@suse.cz>
18417 * doc/invoke.texi: Other params don't use it, remove it.
18419 2021-04-15 Richard Biener <rguenther@suse.de>
18421 * gimple-builder.h: Add deprecation note.
18423 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18426 * attribs.h (restrict_type_identity_attributes_to): Declare.
18427 * attribs.c (restrict_type_identity_attributes_to): New function.
18429 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18432 * attribs.h (affects_type_identity_attributes): Declare.
18433 * attribs.c (remove_attributes_matching): New function.
18434 (affects_type_identity_attributes): Likewise.
18436 2021-04-15 Jakub Jelinek <jakub@redhat.com>
18439 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
18440 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
18441 ZERO_EXTEND, SIGN_EXTEND or AND.
18443 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18445 PR rtl-optimization/99929
18446 * rtl.h (same_vector_encodings_p): New function.
18447 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
18448 * cselib.c (rtx_equal_for_cselib_1): Likewise.
18449 * jump.c (rtx_renumbered_equal_p): Likewise.
18450 * lra-constraints.c (operands_match_p): Likewise.
18451 * reload.c (operands_match_p): Likewise.
18452 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
18454 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18456 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
18457 more information about variable-length CONST_VECTORs.
18459 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
18461 PR rtl-optimization/100066
18462 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
18463 ordered modes when choosing splitting mode for hard reg.
18465 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18468 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
18470 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
18472 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
18474 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
18475 for mask operand types.
18476 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18477 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
18479 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
18481 * config/s390/s390.c (s390_const_operand_ok): Check the new
18482 operand types and generate a list of valid values.
18484 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18486 * doc/tm.texi: Regenerate.
18487 * doc/tm.texi.in (D language and ABI): Add @hook for
18488 TARGET_D_REGISTER_OS_TARGET_INFO.
18490 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18492 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
18494 (aarch64_d_register_target_info): New function.
18495 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
18497 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
18499 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
18500 (arm_d_register_target_info): New function.
18501 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
18502 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18503 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
18504 (ix86_d_register_target_info): New function.
18505 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
18506 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18507 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
18508 (mips_d_register_target_info): New function.
18509 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
18510 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18511 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
18512 (pa_d_register_target_info): New function.
18513 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
18514 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18515 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
18517 (riscv_d_register_target_info): New function.
18518 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
18519 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18520 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
18522 (rs6000_d_register_target_info): New function.
18523 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
18525 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18526 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
18527 (s390_d_register_target_info): New function.
18528 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
18529 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18530 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
18532 (sparc_d_register_target_info): New function.
18533 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
18534 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18535 * doc/tm.texi: Regenerate.
18536 * doc/tm.texi.in (D language and ABI): Add @hook for
18537 TARGET_D_REGISTER_CPU_TARGET_INFO.
18539 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18541 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
18542 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
18543 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
18544 * doc/tm.texi: Regenerate.
18545 * doc/tm.texi.in (D language and ABI): Add @hook for
18546 TARGET_D_HAS_STDCALL_CONVENTION.
18548 2021-04-14 Richard Biener <rguenther@suse.de>
18550 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
18551 VEC_COND_EXPRs have a gimple_val condition.
18552 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
18553 can no longer have a GENERIC condition.
18555 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
18558 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
18559 from the isa_delta when -mfpu has been used.
18560 (arm_options_perform_arch_sanity_checks): It's the architecture that
18561 lacks an FPU not the processor.
18563 2021-04-13 Richard Biener <rguenther@suse.de>
18565 PR tree-optimization/100053
18566 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
18567 not use optimistic dominance queries for backedges to validate
18569 (dominated_by_p_w_unex): Add parameter to ignore executable
18570 state on backedges.
18571 (rpo_elim::eliminate_avail): Adjust.
18573 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18576 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
18577 *aarch64_bfxilsi_extrdi): New define_insn patterns.
18579 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18582 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
18583 outermode, return NULL if the result doesn't encode back to the
18584 original byte sequence.
18585 (simplify_gen_subreg): Don't create SUBREGs from constants to
18586 MODE_COMPOSITE_P outermode.
18588 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18590 PR rtl-optimization/99905
18591 * combine.c (expand_compound_operation): If pos + len > modewidth,
18592 perform the right shift by pos in inner_mode and then convert to mode,
18593 instead of trying to simplify a shift of rtx with inner_mode by pos
18594 as if it was a shift in mode.
18596 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18599 * combine.c (simplify_and_const_int_1): Don't optimize varop
18600 away if it has side-effects.
18602 2021-04-12 Martin Liska <mliska@suse.cz>
18604 * doc/extend.texi: Escape @smallexample content.
18606 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
18608 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
18609 alternative in order to load a DFP zero.
18611 2021-04-12 Martin Liska <mliska@suse.cz>
18613 * doc/extend.texi: Be more precise in documentation
18614 of symver attribute.
18616 2021-04-12 Martin Liska <mliska@suse.cz>
18619 * gimplify.c (gimplify_expr): Right now, we unpoison all
18620 variables before a goto <dest>. We should not do it if we are
18623 2021-04-12 Cui,Lili <lili.cui@intel.com>
18625 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
18627 * common/config/i386/i386-common.c (processor_names): Add
18629 (processor_alias_table): Add rocketlake.
18630 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
18631 INTEL_COREI7_ROCKETLAKE.
18632 * config.gcc: Add -march=rocketlake.
18633 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18635 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
18636 (processor_cost_table): Add rocketlake cost.
18637 * config/i386/i386.h (ix86_size_cost) : Define
18639 (processor_type) : Add PROCESSOR_ROCKETLAKE.
18640 (PTA_ROCKETLAKE): Ditto.
18641 * doc/extend.texi: Add rocketlake.
18642 * doc/invoke.texi: Add rocketlake.
18644 2021-04-12 Cui,Lili <lili.cui@intel.com>
18646 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
18647 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
18648 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
18649 * doc/invoke.texi: Change alderlake ISA list.
18651 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
18653 PR middle-end/98088
18654 * omp-expand.c (expand_oacc_collapse_init): Update condition in
18657 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
18660 * config/i386/serializeintrin.h (_serialize): Defined as macro.
18662 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18665 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
18666 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
18668 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18670 PR middle-end/99989
18671 * gimple-ssa-warn-alloca.c
18672 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
18673 0 with integer precision unconditionally.
18675 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18677 PR rtl-optimization/98601
18678 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
18679 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
18680 unaligned_mems handle VOIDmode like BLKmode.
18682 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
18685 * tree.c (free_lang_data_in_decl): Do not release body of
18686 declare_variant_alt.
18688 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
18690 * config/aarch64/aarch64.c (aarch64_option_restore): If the
18691 architecture was specified explicitly and the tuning wasn't,
18692 tune for the architecture rather than the configured default CPU.
18694 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
18696 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
18697 as the temporary register.
18699 2021-04-09 Martin Liska <mliska@suse.cz>
18701 * doc/extend.texi: Move non-target attributes on the top level.
18703 2021-04-09 Martin Liska <mliska@suse.cz>
18705 * doc/invoke.texi: Document minimum and maximum value of the
18706 argument for both supported compression algorithms.
18708 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
18710 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
18711 TLS BSS before TLS data.
18712 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
18714 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
18716 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
18718 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
18720 * match.pd: Extend vec_cond folds to handle shifts.
18722 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
18724 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
18727 2021-04-08 Alex Coplan <alex.coplan@arm.com>
18730 * config/arm/iterators.md (MVE_vecs): New.
18731 (V_elem): Also handle V2DF.
18732 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
18733 (*mve_vdup<mode>): ... this. Remove second alternative since
18734 vec_duplicate of const_int is not canonical RTL, and we don't
18735 want to match symbol_refs.
18736 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
18738 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
18740 * fold-const.c (fold_single_bit_test): Fix typo.
18741 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
18744 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
18746 PR tree-optimization/97513
18747 * tree-vect-slp.c (vect_add_slp_permutation): New function,
18749 (vectorizable_slp_permutation): ...here. Detect cases in which
18750 all VEC_PERM_EXPRs are guaranteed to have the same stepped
18751 permute vector and only generate one permute vector for that case.
18752 Extend that case to handle variable-length vectors.
18754 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
18756 PR tree-optimization/99873
18757 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
18758 (vect_build_slp_instance): Don't split store groups that could
18759 use IFN_STORE_LANES.
18761 2021-04-07 Jakub Jelinek <jakub@redhat.com>
18764 * varasm.c (output_constant_pool_contents): Don't strip name encoding
18765 from XSTR (desc->sym, 0) or from label before passing those to
18768 2021-04-07 Richard Biener <rguenther@suse.de>
18770 PR tree-optimization/99954
18771 * tree-loop-distribution.c: Include tree-affine.h.
18772 (generate_memcpy_builtin): Try using tree-affine to prove
18774 (loop_distribution::classify_builtin_ldst): Always classify
18777 2021-04-07 Richard Biener <rguenther@suse.de>
18779 PR tree-optimization/99947
18780 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
18781 steps vector to avoid pushing elements from the reallocated
18784 2021-04-07 Richard Biener <rguenther@suse.de>
18786 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
18787 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
18789 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
18791 (debug_vn_reference_ops): New.
18793 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
18795 PR tree-optimization/98736
18796 * tree-loop-distribution.c
18797 * (loop_distribution::bb_top_order_init):
18798 Compute RPO with programing order preserved by calling function
18799 rev_post_order_and_mark_dfs_back_seme.
18801 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
18804 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
18805 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
18807 (process_bb_lives): Don't update biggest mode of hard reg for
18808 implicit in multi-register group. Use the new functions for
18809 updating dead_set and unused_set by register notes.
18811 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
18813 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
18816 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
18818 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
18819 (skylake_memset): Likewise.
18820 (skylake_cost): Change CLEAR_RATIO to 17.
18821 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
18822 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
18823 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
18825 2021-04-06 Richard Biener <rguenther@suse.de>
18827 PR tree-optimization/99880
18828 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
18829 set vectorized defs of relevant PHIs.
18831 2021-04-06 Richard Biener <rguenther@suse.de>
18833 PR tree-optimization/99924
18834 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
18835 nodes w/o scalar stmts as visited.
18837 2021-04-06 Alex Coplan <alex.coplan@arm.com>
18840 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
18841 PCS for [su]fix_optab.
18843 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
18845 * config/darwin.c (machopic_legitimize_pic_address): Check
18846 that the current pic register is one of the hard reg set
18847 before setting liveness.
18849 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
18851 * config/darwin.c (machopic_legitimize_pic_address): Fix
18852 whitespace, remove unused code.
18854 2021-04-03 Jakub Jelinek <jakub@redhat.com>
18856 PR tree-optimization/99882
18857 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
18860 2021-04-03 Jakub Jelinek <jakub@redhat.com>
18862 PR rtl-optimization/99863
18863 * dse.c (replace_read): Drop regs_live argument. Instead of
18864 regs_live, use store_insn->fixed_regs_live if non-NULL,
18865 otherwise punt if insns sequence clobbers or sets any hard
18868 2021-04-03 Jakub Jelinek <jakub@redhat.com>
18871 * targhooks.h (default_print_patchable_function_entry_1): Declare.
18872 * targhooks.c (default_print_patchable_function_entry_1): New function,
18873 copied from default_print_patchable_function_entry with an added flags
18875 (default_print_patchable_function_entry): Rewritten into a small
18876 wrapper around default_print_patchable_function_entry_1.
18877 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
18879 (rs6000_print_patchable_function_entry): New function.
18881 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
18883 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
18885 2021-04-01 Jason Merrill <jason@redhat.com>
18888 * common.opt: Document v15 and v16.
18890 2021-04-01 Richard Biener <rguenther@suse.de>
18892 PR tree-optimization/99863
18893 * gimplify.c (gimplify_init_constructor): Recompute vector
18896 2021-04-01 Jakub Jelinek <jakub@redhat.com>
18898 * doc/extend.texi (symver attribute): Fix up syntax errors
18901 2021-04-01 Jakub Jelinek <jakub@redhat.com>
18903 PR tree-optimization/96573
18904 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
18905 also pointer types.
18907 2021-04-01 Richard Biener <rguenther@suse.de>
18909 PR tree-optimization/99856
18910 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
18911 precision to vector element precision.
18913 2021-04-01 Martin Jambor <mjambor@suse.cz>
18915 PR tree-optimization/97009
18916 * tree-sra.c (access_or_its_child_written): New function.
18917 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
18920 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
18923 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
18925 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
18928 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
18929 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
18930 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
18931 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
18932 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
18933 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
18935 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
18936 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
18937 (define_attr "prefixed"): Update initializer.
18939 2021-03-31 Jakub Jelinek <jakub@redhat.com>
18942 * dwarf2out.c (debug_ranges_dwo_section): New variable.
18943 (DW_RANGES_IDX_SKELETON): Define.
18944 (struct dw_ranges): Add begin_entry and end_entry members.
18945 (DEBUG_DWO_RNGLISTS_SECTION): Define.
18946 (add_ranges_num): Adjust r initializer for addition of *_entry
18948 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
18949 set idx to DW_RANGES_IDX_SKELETON.
18950 (use_distinct_base_address_for_range): New function.
18951 (index_rnglists): Don't set r->idx if it is equal to
18952 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
18953 r->end_entry for -gsplit-dwarf if those will be needed by
18955 (output_rnglists): Add DWO argument. If true, switch to
18956 debug_ranges_dwo_section rather than debug_ranges_section.
18957 Adjust l1/l2 label indexes. Only output the offset table when
18958 dwo is true and don't include in there the skeleton range
18959 entry if present. For -gsplit-dwarf, skip ranges that belong
18960 to the other rnglists section. Change return type from void
18961 to bool and return true if there are any range entries for
18962 the other section. For dwarf_split_debug_info use
18963 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
18964 entries instead of DW_RLE_start_end, DW_RLE_start_length and
18965 DW_RLE_base_address. Use use_distinct_base_address_for_range.
18966 (init_sections_and_labels): Initialize debug_ranges_dwo_section
18967 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
18968 and range_base_label indexes.
18969 (dwarf2out_finish): Call index_rnglists earlier before finalizing
18970 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
18971 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
18972 with different dwo arguments.
18973 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
18975 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
18977 PR tree-optimization/98268
18978 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
18979 recompute_tree_invariant_for_addr_expr after successfully
18980 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
18982 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
18984 PR tree-optimization/99726
18985 * tree-data-ref.c (create_intersect_range_checks_index): Bail
18986 out if there is more than one access function SCEV for the loop
18989 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
18991 PR rtl-optimization/97141
18992 PR rtl-optimization/98726
18993 * emit-rtl.c (valid_for_const_vector_p): Return true for
18995 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
18996 poly_wide_int instead of a wide_int.
18997 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
18999 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19000 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19001 false for CONST_VECTORs that cannot be forced to memory.
19002 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19003 is too complex to force to memory, build it up from individual
19006 2021-03-31 Jan Hubicka <jh@suse.cz>
19009 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19011 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19014 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19017 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19019 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19020 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19022 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19023 "rep movsb/stosb" only for known sizes.
19024 * config/i386/i386-options.c (processor_cost_table): Use Ice
19025 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19026 Rapids and Alder Lake.
19027 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19028 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19029 (icelake_memset): Likewise.
19030 (icelake_cost): Likewise.
19031 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19034 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19037 * config/aarch64/aarch64.c
19038 (aarch64_vectorize_preferred_vector_alignment): Query the size
19039 of the provided SVE vector; do not assume that all SVE vectors
19040 have the same size.
19042 2021-03-31 Jan Hubicka <jh@suse.cz>
19045 * cgraph.c (cgraph_node::release_body): Remove all callers and
19047 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19048 * cgraphunit.c (cgraph_node::expand): And here.
19050 2021-03-31 Martin Liska <mliska@suse.cz>
19052 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19053 and one negated condition.
19055 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19056 Richard Sandiford <richard.sandiford@arm.com>
19059 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19060 constraints on operands[2] and similarly 0 and rk constraints
19061 on operands[1] corresponding to that.
19063 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19066 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19067 linker doesn't support DWARF sections new in DWARF5.
19068 * config/i386/i386-options.c (ix86_option_override_internal): Default
19069 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19071 * config.in: Regenerated.
19072 * configure: Regenerated.
19074 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19077 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19078 available issue_info before using it.
19080 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19083 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19086 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19089 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19090 (altivec_lvsl_reg_<mode>): ... this.
19091 (altivec_lvsr_reg): Change to ...
19092 (altivec_lvsr_reg_<mode>): ... this.
19093 * config/rs6000/predicates.md (vec_set_index_operand): New.
19094 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19095 Enable 32bit variable vec_insert for all TARGET_VSX.
19096 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19097 Enable 32bit variable vec_insert for p9 and above.
19098 (rs6000_expand_vector_set_var_p8): Rename to ...
19099 (rs6000_expand_vector_set_var_p7): ... this.
19100 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19102 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19103 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19104 gen_altivec_lvsr_reg_di.
19106 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19109 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19110 (__rdtscp): Likewise.
19112 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19114 PR tree-optimization/99825
19115 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19116 Reject non-mult 2 lanes.
19118 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19121 * config/arm/arm.c (arm_file_start): Fix emission of
19122 Tag_ABI_VFP_args attribute.
19124 2021-03-30 Richard Biener <rguenther@suse.de>
19126 PR tree-optimization/99824
19127 * stor-layout.c (set_min_and_max_values_for_integral_type):
19128 Assert the precision is within the bounds of
19129 WIDE_INT_MAX_PRECISION.
19130 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19131 the outermost component ref only to lower the access size
19132 and initialize that from the access type.
19134 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19137 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19138 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19140 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19142 * config/aarch64/aarch64.md
19143 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19144 attribute to disambiguate between SIMD and FP variants of the
19147 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19149 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19150 (analyze_ssa_name_flags): Fix typo in comment.
19152 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19155 * config/aarch64/aarch64-sve-builtins.cc
19156 (function_builder::add_function): Add placeholder_p argument, use
19157 placeholder decls if this is set.
19158 (function_builder::add_unique_function): Instead of conditionally adding
19159 direct overloads, unconditionally add either a direct overload or a
19161 (function_builder::add_overloaded_function): Set placeholder_p if we're
19162 using C++ overloads. Use the obstack for string storage instead
19163 of relying on the tree nodes.
19164 (function_builder::add_overloaded_functions): Don't return early for
19165 m_direct_overloads: we need to add placeholders.
19166 * config/aarch64/aarch64-sve-builtins.h
19167 (function_builder::add_function): Add placeholder_p argument.
19169 2021-03-29 Richard Biener <rguenther@suse.de>
19171 PR tree-optimization/99807
19172 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19173 assert below VEC_PERM handling.
19175 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19178 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19179 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19180 matching const_int 0.
19181 (move_lo_quad_internal_be_<mode>): Likewise.
19182 (move_lo_quad_<mode>): Update for the above.
19183 * config/aarch64/iterators.md (VQ_2E): Delete.
19185 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19187 PR tree-optimization/99777
19188 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19189 types other than scalar integral types.
19191 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19193 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19194 XCOFF TLS reloc decorations.
19196 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19198 * doc/analyzer.texi (Analyzer Internals): Update link to
19199 "A Memory Model for Static Analysis of C Programs".
19201 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19203 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19204 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19206 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19207 (rs6000_special_round_type_align): Recursively check innermost first
19210 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19213 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19214 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19215 assignment with drap_reg active, queue reg save for hfp with offset 0
19216 and flush queued reg saves. When handling a push with rule18,
19217 defer queueing reg save for hfp and just assert the offset is 0.
19218 (scan_trace): Assert that fde->rule18 is false.
19220 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19223 * ira-costs.c (record_reg_classes): Put case with
19224 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19225 * ira.c (ira_setup_alts): Ditto.
19226 * lra-constraints.c (process_alt_operands): Ditto.
19227 * recog.c (asm_operand_ok): Ditto.
19228 * reload.c (find_reloads): Ditto.
19230 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19232 * config/aarch64/aarch64-protos.h
19233 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19234 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19235 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19236 accordingly, using the same costs as for post_modify.
19237 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19238 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19239 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19240 (a64fx_addrcost_table): Likewise.
19241 (neoversev1_addrcost_table): New.
19242 (neoversev1_tunings): Use neoversev1_addrcost_table.
19243 (aarch64_address_cost): Use the new post_modify costs for CImode
19246 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19248 * config/aarch64/aarch64.opt
19249 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19250 * doc/invoke.texi: Document it.
19251 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19252 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19253 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19254 (aarch64_vec_issue_info): New structures.
19255 (cpu_vector_cost): Write comments above the variables rather
19257 (cpu_vector_cost::issue_info): New member variable.
19258 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19259 and tree-ssa-loop-niter.h.
19260 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19261 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19262 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19263 (thunderx3t110_vector_cost): Initialize issue_info to null.
19264 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19265 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19266 (neoversev1_vector_cost): Use them.
19267 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19268 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19269 (aarch64_vector_costs::num_vector_iterations): Likewise.
19270 (aarch64_vector_costs::scalar_ops): Likewise.
19271 (aarch64_vector_costs::advsimd_ops): Likewise.
19272 (aarch64_vector_costs::sve_ops): Likewise.
19273 (aarch64_vector_costs::seen_loads): Likewise.
19274 (aarch64_simd_vec_costs_for_flags): New function.
19275 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19276 Count the number of predicate operations required by SVE WHILE
19278 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19279 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19280 (aarch64_count_ops): Likewise.
19281 (aarch64_add_stmt_cost): Record whether see an SVE operation
19282 that cannot currently be implementing using Advanced SIMD.
19283 Record issue information about the scalar, Advanced SIMD
19284 and (where relevant) SVE versions of a loop.
19285 (aarch64_vec_op_count::dump): New function.
19286 (aarch64_sve_op_count::dump): Likewise.
19287 (aarch64_estimate_min_cycles_per_iter): Likewise.
19288 (aarch64_adjust_body_cost): If issue information is available,
19289 try to compare the issue rates of the various loop implementations
19290 and increase or decrease the vector body cost accordingly.
19292 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19294 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19295 Assume a zero cost for induction phis.
19297 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19299 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19301 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19302 vector comparisons.
19304 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19306 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19308 (aarch64_add_stmt_cost): Call it.
19310 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19312 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19313 New tuning parameter.
19314 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19315 (aarch64_estimated_sve_vq): New function.
19316 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19317 (aarch64_vector_costs::is_loop): Likewise.
19318 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19319 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19320 (aarch64_record_potential_advsimd_unrolling): New function.
19321 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19322 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19323 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19324 Detect whether we're vectorizing a loop for SVE that might be
19325 completely unrolled if it used Advanced SIMD instead.
19326 (aarch64_adjust_body_cost_for_latency): New function.
19327 (aarch64_finish_cost): Call it.
19329 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19331 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
19332 (aarch64_init_cost): New function.
19333 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
19334 the default unsigned[3].
19335 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
19336 (TARGET_VECTORIZE_INIT_COST): Override.
19337 (TARGET_VECTORIZE_FINISH_COST): Likewise.
19338 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
19340 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19342 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
19343 (neoversev1_sve_vector_cost): New cost structures.
19344 (neoversev1_vector_cost): Likewise.
19345 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
19347 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19349 * config/aarch64/aarch64-protos.h
19350 (sve_vec_cost::scatter_store_elt_cost): New member variable.
19351 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19352 accordingly, taking the cost from the cost of a scalar_store.
19353 (a64fx_sve_vector_cost): Likewise.
19354 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
19356 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19358 * config/aarch64/aarch64-protos.h
19359 (simd_vec_cost::store_elt_extra_cost): New member variable.
19360 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19361 accordingly, using the vec_to_scalar cost for the new field.
19362 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19363 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19364 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19365 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19366 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19367 (thunderx3t110_advsimd_vector_cost): Likewise.
19368 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
19370 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19372 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
19373 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
19374 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
19375 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19376 accordingly, using zero for the new costs.
19377 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19378 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19379 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19380 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19381 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19382 (thunderx3t110_advsimd_vector_cost): Likewise.
19383 (aarch64_ld234_st234_vectors): New function.
19384 (aarch64_adjust_stmt_cost): Likewise.
19385 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
19386 the new vector costs.
19388 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19390 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
19391 derived class of simd_vec_cost. Add information about CLAST[AB]
19392 and FADDA instructions.
19393 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19394 accordingly, using the vec_to_scalar costs for the new fields.
19395 (a64fx_sve_vector_cost): Likewise.
19396 (aarch64_reduc_type): New function.
19397 (aarch64_sve_in_loop_reduction_latency): Likewise.
19398 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
19399 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
19400 that occur in the loop body.
19401 (aarch64_add_stmt_cost): Update call accordingly.
19403 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19405 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
19407 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
19408 above the fields rather than to the right.
19409 (simd_vec_cost::reduc_i8_cost): New member variable.
19410 (simd_vec_cost::reduc_i16_cost): Likewise.
19411 (simd_vec_cost::reduc_i32_cost): Likewise.
19412 (simd_vec_cost::reduc_i64_cost): Likewise.
19413 (simd_vec_cost::reduc_f16_cost): Likewise.
19414 (simd_vec_cost::reduc_f32_cost): Likewise.
19415 (simd_vec_cost::reduc_f64_cost): Likewise.
19416 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19417 accordingly, using the vec_to_scalar_cost for the new fields.
19418 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19419 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19420 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19421 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19422 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19423 (thunderx3t110_advsimd_vector_cost): Likewise.
19424 (aarch64_use_new_vector_costs_p): New function.
19425 (aarch64_simd_vec_costs): New function, split out from...
19426 (aarch64_builtin_vectorization_cost): ...here.
19427 (aarch64_is_reduction): New function.
19428 (aarch64_detect_vector_stmt_subtype): Likewise.
19429 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
19430 using the new vector costs.
19432 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19435 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
19436 TLS declarations as public.
19438 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19440 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
19441 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
19442 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
19443 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
19444 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
19445 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
19446 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
19447 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
19448 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
19450 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19453 * config.gcc (*-*-cygwin*): Add winnt-d.o
19454 (*-*-mingw*): Likewise.
19455 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
19456 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
19457 * config/i386/t-cygming: Add winnt-d.o.
19458 * config/i386/winnt-d.c: New file.
19460 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19462 * config/freebsd-d.c: Include memmodel.h.
19464 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19467 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
19468 * config/t-openbsd: Add openbsd-d.o.
19469 * config/openbsd-d.c: New file.
19471 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
19473 PR tree-optimization/96974
19474 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
19475 with graceful exit.
19477 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19480 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19484 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19485 always_inline in system headers.
19487 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
19489 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
19491 2021-03-25 Jakub Jelinek <jakub@redhat.com>
19494 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
19495 * fold-const.c (operand_compare::operand_equal_p): Don't compare
19496 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
19498 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19502 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19503 always_inline in system headers.
19505 2021-03-25 Richard Biener <rguenther@suse.de>
19507 PR tree-optimization/99746
19508 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
19509 the scalar stmt as patterned. Instead set up required things
19512 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
19514 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
19517 2021-03-24 Martin Liska <mliska@suse.cz>
19520 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
19522 * config/i386/i386-options.c (ix86_option_override_internal):
19523 Add run-time assert.
19525 2021-03-24 Martin Jambor <mjambor@suse.cz>
19528 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
19529 parameters with unknown type.
19530 (ipacp_value_safe_for_type): New function.
19531 (propagate_vals_across_arith_jfunc): Verify that the constant type
19532 can be used for a type of the formal parameter.
19533 (propagate_vals_across_ancestor): Likewise.
19534 (propagate_scalar_across_jump_function): Likewise. Pass the type
19535 also to propagate_vals_across_ancestor.
19537 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
19540 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
19542 (movmisalign<mode>_mve_load): Likewise.
19544 2021-03-24 Jakub Jelinek <jakub@redhat.com>
19547 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
19548 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
19550 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19552 * doc/sourcebuild.texi (sysconf): New effective target.
19554 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19556 * config/i386/predicates.md (reg_or_const_vec_operand): New.
19557 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
19558 the now *-prefixed insn_and_split, turn the splitter const vec
19559 into an input for the insn, making it an ignored immediate for
19560 non-split cases, and loaded into the scratch register
19563 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
19566 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
19567 Use define_relaxed_memory_constraint for them.
19569 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
19572 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
19573 colon to the diagnostic message.
19575 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
19577 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
19579 (try_fwprop_subst_note): Use set_info instead of insn_info.
19580 (try_fwprop_subst_pattern): Likewise.
19581 (try_fwprop_subst_notes): Likewise.
19582 (try_fwprop_subst): Likewise.
19583 (forward_propagate_subreg): Likewise.
19584 (forward_propagate_and_simplify): Likewise.
19585 (forward_propagate_into): Likewise.
19586 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
19588 (set_info::single_nondebug_insn_use): Likewise.
19589 (set_info::single_phi_use): Likewise.
19590 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
19592 (set_info::single_nondebug_insn_use): Likewise.
19593 (set_info::single_phi_use): Likewise.
19595 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
19597 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
19599 2021-03-23 Jakub Jelinek <jakub@redhat.com>
19602 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
19603 expand_mult to perform an unsigned rather than a signed
19606 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
19609 * config/i386/cpuid.h (__cpuid): Add __volatile__.
19610 (__cpuid_count): Likewise.
19612 2021-03-23 Richard Biener <rguenther@suse.de>
19614 PR tree-optimization/99721
19615 * tree-vect-slp.c (vect_slp_analyze_node_operations):
19616 Make sure we can schedule the node.
19618 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19620 * config/riscv/riscv.c (riscv_subword): Take endianness into
19621 account when calculating the byte offset.
19623 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19625 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
19626 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
19627 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
19628 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
19629 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
19630 new predicate "subreg_lowpart_operator"
19632 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19634 * config/riscv/riscv.c (riscv_swap_instruction): New function
19635 to byteswap an SImode rtx containing an instruction.
19636 (riscv_trampoline_init): Byteswap the generated instructions
19639 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19641 * common/config/riscv/riscv-common.c
19642 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
19643 * config.gcc (riscv32be-*, riscv64be-*): Set
19644 TARGET_BIG_ENDIAN_DEFAULT to 1.
19645 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
19646 depending on default endianness.
19647 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
19648 * config/riscv/linux.h (LINK_SPEC): Likewise.
19649 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
19650 default endianness.
19651 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
19653 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19655 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
19656 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
19657 * config/riscv/linux.h (LINK_SPEC): Likewise.
19658 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
19660 (BYTES_BIG_ENDIAN): Handle big endian.
19661 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
19662 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
19664 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
19666 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
19668 * regcprop.c (find_oldest_value_reg): Ask target whether
19669 different mode is fine for replacement register.
19671 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
19673 PR tree-optimization/99296
19674 * value-range.cc (irange::irange_set_1bit_anti_range): New.
19675 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
19676 * value-range.h (irange::irange_set_1bit_anti_range): New.
19678 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
19681 * config/aarch64/constraints.md (UtQ): Use
19682 define_relaxed_memory_constraint for it.
19683 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
19684 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19685 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
19686 (have_relaxed_memory_constraints): New static var.
19687 (relaxed_memory_start, relaxed_memory_end): Ditto.
19688 (add_constraint): Add arg is_relaxed_memory. Check name for
19689 relaxed memory. Set up is_relaxed_memory in constraint_data and
19690 have_relaxed_memory_constraints. Adjust calls.
19691 (choose_enum_order): Process relaxed memory.
19692 (write_tm_preds_h): Ditto.
19693 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19694 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19695 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
19696 * ira-lives.c (single_reg_class): Use
19697 insn_extra_relaxed_memory_constraint.
19698 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
19699 * lra-constraints.c (valid_address_p): Use
19700 insn_extra_relaxed_memory_constraint instead of other memory
19702 (process_alt_operands): Process CT_RELAXED_MEMORY.
19703 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
19704 * recog.c (asm_operand_ok, preprocess_constraints): Process
19706 * reload.c (find_reloads): Ditto.
19707 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
19708 * stmt.c (parse_input_constraint): Use
19709 insn_extra_relaxed_memory_constraint.
19711 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
19714 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
19717 2021-03-22 Alex Coplan <alex.coplan@arm.com>
19720 * config/arm/arm-protos.h (neon_make_constant): Add generate
19721 argument to guard emitting insns, default to true.
19722 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
19723 CONST_VECTORs which neon_make_constant can't handle.
19724 (neon_vdup_constant): Add generate argument, avoid emitting
19725 insns if it's not set.
19726 (neon_make_constant): Plumb new generate argument through.
19727 * config/arm/constraints.md (Ui): New. Use it...
19728 * config/arm/mve.md (*mve_mov<mode>): ... here.
19729 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
19730 synthesize constants.
19732 2021-03-22 Richard Biener <rguenther@suse.de>
19734 * debug.h: Add deprecation warning.
19736 2021-03-22 Richard Biener <rguenther@suse.de>
19738 PR tree-optimization/99694
19739 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
19742 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
19745 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
19746 after type checking.
19748 2021-03-22 Jakub Jelinek <jakub@redhat.com>
19752 * dwarf2out.c (get_full_len): Use get_precision rather than
19754 (add_const_value_attribute): Make sure add_AT_wide argument has
19755 precision prec rather than some very wide one.
19757 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
19759 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
19760 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
19761 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
19762 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
19763 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
19764 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
19765 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
19766 *fix_trunc<mode>si2_internal): Fix empty split condition.
19767 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
19768 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
19769 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
19770 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
19772 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
19775 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19776 Convert idx to DImode.
19777 (rs6000_expand_vector_set_var_p8): Likewise.
19779 2021-03-21 Jakub Jelinek <jakub@redhat.com>
19782 * dwarf2out.c (insert_float): Change return type from void to
19783 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
19784 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
19787 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
19790 * config/i386/i386.c (construct_container): Check cfun != NULL
19791 before accessing silent_p.
19793 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
19795 * asan.c: Fix typos in comments.
19797 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
19799 PR rtl-optimization/99680
19800 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
19801 (process_address_1): Check empty constraint before using
19804 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
19806 * config/rs6000/rs6000.c (power10_cost): New.
19807 (rs6000_option_override_internal): Set Power10 costs.
19808 (rs6000_issue_rate): Set Power10 issue rate.
19809 * config/rs6000/power10.md: Rewrite for Power10.
19811 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
19814 * lra-constraints.c (process_address_1): Don't use unknown
19815 constraint for address constraint.
19817 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
19820 * config.gcc (powerpc-*-darwin8): Delete the reference to
19821 the now removed darwin8.h.
19823 2021-03-19 Olivier Hainque <hainque@adacore.com>
19826 * config/vxworksae.h (VX_CPU_PREFIX): Define.
19828 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
19830 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
19832 2021-03-19 Tamar Christina <tamar.christina@arm.com>
19834 PR tree-optimization/99656
19835 * tree-vect-slp-patterns.c (linear_loads_p,
19836 complex_add_pattern::matches, is_eq_or_top,
19837 vect_validate_multiplication, complex_mul_pattern::matches,
19838 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
19839 * tree-vectorizer.h: (complex_load_perm_t): Removed.
19840 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
19841 complex_load_perm_t.
19843 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
19846 * config/i386/i386-options.c (ix86_init_machine_status): Set
19848 * config/i386/i386.c (init_cumulative_args): Set silent_p to
19850 (construct_container): Return early for return and argument
19851 errors if silent_p is true.
19852 * config/i386/i386.h (machine_function): Add silent_p.
19854 2021-03-19 Jakub Jelinek <jakub@redhat.com>
19857 * config/arm/constraints.md (Ds): New constraint.
19858 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
19859 constraint instead of w,Dm.
19861 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
19863 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
19866 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
19868 PR middle-end/99641
19869 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
19870 array type, do the computation of the current position in sizetype.
19872 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
19875 * lra-constraints.c (process_address_1): Use lookup_constraint
19876 only for a single constraint.
19878 2021-03-18 Martin Sebor <msebor@redhat.com>
19880 PR middle-end/99502
19881 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
19882 (inbounds_memaccess_p): ...to this. Check the ending offset of
19883 the accessed member.
19885 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
19887 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
19888 %> quote markers to error messages.
19889 (gcn_goacc_validate_dims): Likewise.
19890 (gcn_conditional_register_usage): Remove exclaimation mark from error
19892 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
19894 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
19896 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
19899 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
19900 Kito Cheng <kito.cheng@sifive.com>
19902 * config/riscv/riscv.c (riscv_block_move_straight): Change type
19903 to unsigned HOST_WIDE_INT for parameter and local variable with
19904 HOST_WIDE_INT type.
19905 (riscv_adjust_block_mem): Ditto.
19906 (riscv_block_move_loop): Ditto.
19907 (riscv_expand_block_move): Ditto.
19909 2021-03-18 Nick Clifton <nickc@redhat.com>
19911 * config/v850/v850.c (construct_restore_jr): Increase static
19913 (construct_save_jarl): Likewise.
19914 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
19916 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19918 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
19919 (aarch64_override_options_internal): Use it.
19920 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
19923 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
19925 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
19926 error message format issues.
19927 (nios2_option_override): Likewise.
19928 (nios2_expand_fpu_builtin): Likewise.
19929 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
19930 truncation warning.
19931 (nios2_expand_custom_builtin): More error message format fixes.
19932 (nios2_expand_rdwrctl_builtin): Likewise.
19933 (nios2_expand_rdprs_builtin): Likewise.
19934 (nios2_expand_eni_builtin): Likewise.
19935 (nios2_expand_builtin): Likewise.
19936 (nios2_register_custom_code): Likewise.
19937 (nios2_valid_target_attribute_rec): Likewise.
19938 (nios2_add_insn_asm): Fix uninitialized variable warning.
19940 2021-03-17 Jan Hubicka <jh@suse.cz>
19942 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
19943 of gather to match reality.
19944 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
19946 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19948 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
19949 to compare against CC_REG rather than NE.
19951 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
19954 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
19955 inline assembly statements.
19956 (ix86_print_operand): Update 'P' handling for -fno-plt.
19958 2021-03-17 Tamar Christina <tamar.christina@arm.com>
19961 * config/aarch64/aarch64.c
19962 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
19964 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
19967 * config/rs6000/predicates.md (branch_comparison_operator): Allow
19968 ordered and unordered for CCFPmode, if flag_finite_math_only.
19970 2021-03-16 Jakub Jelinek <jakub@redhat.com>
19973 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
19974 rather than ASHIFT.
19975 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
19977 2021-03-16 Martin Liska <mliska@suse.cz>
19980 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
19981 cl_optimization_compare function.
19983 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
19985 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
19988 2021-03-16 Jakub Jelinek <jakub@redhat.com>
19991 * config/i386/i386.h (struct machine_function): Add
19992 has_explicit_vzeroupper bitfield.
19993 * config/i386/i386-expand.c (ix86_expand_builtin): Set
19994 cfun->machine->has_explicit_vzeroupper when expanding
19995 IX86_BUILTIN_VZEROUPPER.
19996 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
19997 Do the mode switching only when TARGET_VZEROUPPER, expensive
19998 optimizations turned on and not optimizing for size.
19999 (pass_insert_vzeroupper::gate): Enable even when
20000 cfun->machine->has_explicit_vzeroupper is set.
20002 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20005 * config/aarch64/aarch64.c
20006 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20007 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20008 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20010 2021-03-15 Richard Biener <rguenther@suse.de>
20012 PR tree-optimization/98834
20013 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20014 subsetting by truncating the access size.
20016 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20018 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20019 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20022 2021-03-15 Martin Liska <mliska@suse.cz>
20024 * spellcheck.c: Add missing comma in initialization.
20026 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20028 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20029 alternative 2 and alternative 1 with alternative 3 using
20030 YW register constraint.
20031 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20032 using YW register constraint.
20033 (*vec_extractv16qi_zext): Ditto.
20034 (*vec_extractv4si): Merge alternatives 4 and 5
20035 using Yw register constraint.
20036 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20038 2021-03-13 Martin Sebor <msebor@redhat.com>
20040 PR tree-optimization/99489
20041 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20042 is not a call statement.
20044 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20046 PR tree-optimization/99544
20047 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20048 if for vector types multiplication can't be done in type's mode.
20050 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20053 * config/sparc/constraints.md (w): Rename to...
20054 (W): ... this and ditch previous implementation.
20055 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20056 (*movdf_insn_sp64): Likewise.
20057 (*mov<VM64:mode>_insn_sp64): Likewise.
20058 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20060 (atomic_compare_and_swap_leon3_1): Likewise.
20061 (*atomic_compare_and_swapdi_v8plus): Likewise.
20062 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20063 architecture and add missing address validity check during LRA.
20065 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20068 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20069 occuring for assumed-size arrays in use_device_{ptr,addr}.
20071 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20074 * config/i386/constraints.md (YW): New internal constraint.
20075 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20076 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20077 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20078 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20080 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20081 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20082 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20083 into one, use Yw instead of former x,v.
20084 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20085 the last alternative.
20086 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20087 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20088 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20089 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20090 into one, use <v_Yw> instead of former x,v.
20091 (avx2_interleave_highv32qi<mask_name>,
20092 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20093 constraints. Add && <mask_avx512bw_condition> to condition.
20094 (avx2_interleave_lowv32qi<mask_name>,
20095 vec_interleave_lowv16qi<mask_name>,
20096 avx2_interleave_highv16hi<mask_name>,
20097 vec_interleave_highv8hi<mask_name>,
20098 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20099 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20100 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20101 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20102 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20103 Yw instead of v in constraints.
20104 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20105 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20106 instead of Yv in constraints.
20107 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20108 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20109 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20110 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20111 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20113 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20114 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20115 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20116 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20117 two, one with just x, another isa avx512vl with v.
20119 2021-03-12 Martin Liska <mliska@suse.cz>
20121 * doc/invoke.texi: Add missing param documentation.
20123 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20126 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20127 analyzer/trimmed-graph.o.
20128 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20129 feasibility checking to reflect new implementation.
20130 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20132 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20134 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20136 * digraph.cc (selftest::test_shortest_paths): Update
20137 shortest_paths init for new param. Add test of
20138 SPS_TO_GIVEN_TARGET.
20139 * shortest-paths.h (enum shortest_path_sense): New.
20140 (shortest_paths::shortest_paths): Add "sense" param.
20141 Update for renamings. Generalize to use "sense" param.
20142 (shortest_paths::get_shortest_path): Rename param.
20143 (shortest_paths::m_sense): New field.
20144 (shortest_paths::m_prev): Rename...
20145 (shortest_paths::m_best_edge): ...to this.
20146 (shortest_paths::get_shortest_path): Update for renamings.
20147 Conditionalize flipping of path on sense of traversal.
20149 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20151 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20152 for paths from B and C.
20153 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20154 unreachable nodes, rather than asserting.
20156 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20159 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20160 xcoff_tbss_section_name.
20161 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20162 * xcoffout.c (xcoff_tbss_section_name): Delete.
20163 * xcoffout.h (xcoff_tbss_section_name): Delete.
20165 2021-03-11 Richard Biener <rguenther@suse.de>
20167 PR tree-optimization/99523
20168 * tree-cfg.c (dump_function_to_file): Dump SSA names
20169 w/o identifier to the decls section as well, not only those
20170 without a VAR_DECL.
20172 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20175 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20176 function calls with lhs fail if the lhs don't have compatible types.
20178 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20180 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20181 Change FRAME_POINTER_REGNUM to correspond to a new faked
20182 register faked_fp, part of GENNONACR_REGS like faked_ap.
20183 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20184 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20185 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20186 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20187 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20189 (CRIS_REAL_FP_REGNUM): New constant.
20190 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20191 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20192 (cris_initial_elimination_offset): Handle elimination changes
20193 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20194 and add one from FRAME_POINTER_REGNUM to
20195 HARD_FRAME_POINTER_REGNUM.
20196 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20197 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20199 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20202 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20203 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20205 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20208 * lra-constraints.c (process_address_1): Don't check unknown
20209 constraint, use X for empty constraint.
20211 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20213 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20214 Fix typo in comment describing "is_ha" argument.
20216 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20218 * doc/sourcebuild.texi: Document LRA target selector.
20220 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20222 * doc/ux.texi: Add subsection contrasting interactive versus
20223 batch usage of GCC.
20225 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20228 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20230 (vectorizable_load): Fix gather load mask check condition.
20232 2021-03-10 Richard Biener <rguenther@suse.de>
20234 PR tree-optimization/99510
20235 * tree.c (check_aligned_type): Check that the candidate
20236 has TYPE_USER_ALIGN set instead of matching with the
20239 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20241 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20242 float and vector integer modes only if the mode is not larger.
20244 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20246 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20248 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20250 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20252 * ira-lives.c (single_reg_class): Ditto.
20254 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20256 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20257 the architecture-specific rtems.h.
20258 (aarch64-*-rtems*): Likewise.
20259 (arm*-*-rtems*): Likewise.
20260 (epiphany-*-rtems*): Likewise.
20261 (riscv*-*-rtems*): Likewise.
20263 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20265 PR tree-optimization/99305
20266 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20267 before integer_all_onesp instead of vice versa.
20269 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20271 * common/config/arm/arm-common.c (arm_config_default): Change type
20272 of 'i' to unsigned.
20274 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20277 * lra-constraints.c (process_address_1): Process constraint 'g'
20278 separately and digital constraints containing more one digit.
20280 2021-03-09 Nick Clifton <nickc@redhat.com>
20282 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20283 (DWARF"_DEBUGGING_INFO): Define.
20285 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20288 * calls.c (initialize_argument_information): When the argument
20289 is passed by reference, do not make a copy in a thunk only if
20290 the argument is already in memory. Remove redundant test for
20291 the case of callee copy.
20293 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20296 * lra-constraints.c (process_address_1): Process 0..9 constraints
20297 in process_address_1.
20299 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20301 * config/s390/s390.c (struct s390_processor processor_table):
20302 Binutils name string must not be empty.
20304 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20306 * config/arc/arc.c (arc_attr_type): Remove function.
20308 2021-03-09 Martin Liska <mliska@suse.cz>
20311 * config/i386/i386-options.c (ix86_option_override_internal):
20312 Set isa_flags for OPTS argument and not for the global
20315 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20317 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20320 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20323 * config/rs6000/predicates.md (ds_form_mem_operand) New
20325 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20326 ds_form_mem_operand in ld/lwa patterns.
20327 * config/rs6000/fusion.md: Regenerate file.
20329 2021-03-08 Martin Sebor <msebor@redhat.com>
20331 PR middle-end/98266
20332 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
20333 (array_bounds_checker::check_array_bounds): Call it.
20335 2021-03-08 Martin Sebor <msebor@redhat.com>
20337 PR middle-end/97631
20338 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
20339 (handle_builtin_stxncpy_strncat): Rename locals. Determine
20340 destination size from allocation calls. Issue a more appropriate
20342 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
20343 (handle_builtin_memset): Same.
20345 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
20348 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
20349 to ensure we do not have an Altivec style address.
20350 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
20351 an Altivec style address.
20352 (*vsx_le_perm_store_<mode>): Likewise.
20353 (splitters after *vsx_le_perm_store_<mode>): Likewise.
20354 (vsx_load_<mode>): Disable special expander if passed an Altivec
20356 (vsx_store_<mode>): Likewise.
20358 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20361 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
20362 (aarch64_simd_shift_imm_vec_hi): Likewise.
20363 (aarch64_simd_shift_imm_vec_si): Likewise.
20364 (aarch64_simd_shift_imm_vec_di): Likewise.
20365 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
20366 predicate from above.
20367 (aarch64_shrn<mode>_insn_be): Likewise.
20368 (aarch64_rshrn<mode>_insn_le): Likewise.
20369 (aarch64_rshrn<mode>_insn_be): Likewise.
20370 (aarch64_shrn2<mode>_insn_le): Likewise.
20371 (aarch64_shrn2<mode>_insn_be): Likewise.
20372 (aarch64_rshrn2<mode>_insn_le): Likewise.
20373 (aarch64_rshrn2<mode>_insn_be): Likewise.
20375 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
20378 * lra-constraints.c (skip_contraint_modifiers): New function.
20379 (process_address_1): Use it before lookup_constraint call.
20381 2021-03-08 Martin Liska <mliska@suse.cz>
20384 * config/i386/i386-options.c (ix86_option_override_internal):
20385 Enable UINTR and HRESET for -march that supports it.
20387 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
20389 * config/s390/s390.c (f_constraint_p): New function.
20390 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
20391 (TARGET_MD_ASM_ADJUST): Likewise.
20393 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
20396 * tree-nested.c (convert_local_reference_stmt): Avoid calling
20397 lookup_field_for_decl for Fortran module (= namespace context).
20399 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
20401 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
20402 comparison with arithmetic right shift.
20403 (s390_expand_vcond): No need for a force_reg anymore.
20404 s390_vec_compare will do it.
20405 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
20406 immediate operands.
20408 2021-03-07 Jakub Jelinek <jakub@redhat.com>
20411 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
20412 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
20414 * config/i386/sse.md (v_Yw): New define_mode_attr.
20415 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
20416 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
20418 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
20419 xYw in constraints.
20421 2021-03-06 Julian Brown <julian@codesourcery.com>
20423 * tree-pretty-print.c (dump_generic_node): Emit non-generic
20424 address space info for aggregates.
20426 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
20428 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
20430 2021-03-05 Jakub Jelinek <jakub@redhat.com>
20432 PR middle-end/99322
20433 * tree-cfg.c (bb_to_omp_idx): New variable.
20434 (execute_build_cfg): Release the bb_to_omp_idx vector after
20435 cleanup_tree_cfg returns.
20436 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
20437 for bb_to_omp_idx being a vec<int> instead of pointer to array
20439 (make_edges): Remove bb_to_omp_idx local variable, don't pass
20440 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
20441 vec<int> instead of pointer to array of ints and don't free/release
20443 (remove_bb): When removing a bb and placing forced label somewhere
20444 else, ensure it is put into the same OpenMP region during cfg
20445 pass if possible or to entry successor as fallback. Unregister
20446 bb from bb_to_omp_idx.
20448 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
20451 * lra-constraints.c (process_address_1): Skip decomposing address
20452 for asm insn operand with unknown constraint.
20454 2021-03-05 Martin Jambor <mjambor@suse.cz>
20457 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
20458 corresponding speculative edges if we are about to resolve
20459 sepculation. Make edge direct (and so resolve speculations) before
20460 removing it from call_site_hash.
20461 (cgraph_edge::make_direct): Relax the initial assert to allow calling
20462 the function on speculative direct edges.
20464 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
20466 PR rtl-optimization/99376
20467 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
20468 of low-order zero bits is too large, set the result to 0 directly.
20470 2021-03-04 Jakub Jelinek <jakub@redhat.com>
20472 PR middle-end/93235
20473 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
20474 SUBREG and a SUBREG to op_mode can't be created.
20476 2021-03-04 Alex Coplan <alex.coplan@arm.com>
20479 * config/aarch64/aarch64-sve-builtins.cc
20480 (function_resolver::require_vector_type): Handle error_mark_node.
20482 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
20484 * cfgexpand.c (expand_asm_loc): Pass new parameter.
20485 (expand_asm_stmt): Likewise.
20486 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
20488 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20489 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
20490 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
20491 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
20492 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
20493 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
20494 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
20495 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
20496 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
20497 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
20498 * doc/tm.texi (md_asm_adjust): Likewise.
20499 * target.def (md_asm_adjust): Likewise.
20501 2021-03-04 Richard Biener <rguenther@suse.de>
20503 PR middle-end/97855
20504 * tree-pretty-print.c: Poison pp_printf.
20505 (dump_decl_name): Avoid use of pp_printf.
20506 (dump_block_node): Likewise.
20507 (dump_generic_node): Likewise.
20509 2021-03-04 Martin Sebor <msebor@redhat.com>
20511 PR middle-end/96963
20512 PR middle-end/94655
20513 * builtins.c (handle_array_ref): New helper.
20514 (handle_mem_ref): New helper.
20515 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
20516 into new helper functions. Correct a workaround for vectorized
20519 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
20521 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
20522 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
20523 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
20524 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
20525 attribute for Power10.
20526 * config/rs6000/mma.md (*movoo): Likewise.
20527 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
20528 (define_mode_attr bits): Add DD/TD modes.
20529 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
20530 store_conditionalpti): Update size attribute for Power10.
20532 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20535 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
20536 -Wuninitialized, -Wmaybe-uninitialized.
20537 (wide-int.o-warn): Likewise.
20539 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
20541 * common/config/arm/arm-common.c: Include configargs.h.
20542 (arm_config_default): New function.
20543 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
20544 processors that do not support Thumb. Take into account the
20545 --with-mode configuration setting for selecting the default.
20546 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
20547 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
20549 2021-03-03 Martin Liska <mliska@suse.cz>
20551 PR gcov-profile/97461
20552 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
20554 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
20557 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20558 point back the hard frame pointer to its default location when the
20559 frame is larger than SEH_MAX_FRAME_SIZE.
20561 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20564 * config/i386/predicates.md (logic_operator): New define_predicate.
20565 * config/i386/i386.md (mov + mem using comm arith peephole2):
20566 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
20567 and the inner mode is [QH]Imode.
20569 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20572 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
20573 (new_loc_list): Clear end_entry.
20574 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
20575 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
20577 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
20578 initialize also end_entry.
20580 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20583 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
20584 partitions, if in non-layout mode after reorder_blocks also move
20585 affected blocks to ensure a single partition transition.
20587 2021-03-03 Jason Merrill <jason@redhat.com>
20590 * cgraphunit.c (process_function_and_variable_attributes): Don't
20591 warn about flatten on an alias if the target also has it.
20592 * cgraph.h (symtab_node::get_alias_target_tree): New.
20594 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
20596 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
20597 period to symbol name.
20598 (tls_get_addr_internal<mode>): Same.
20600 2021-03-02 David Malcolm <dmalcolm@redhat.com>
20603 * diagnostic-show-locus.c
20604 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
20607 2021-03-02 Martin Sebor <msebor@redhat.com>
20609 PR middle-end/99276
20610 * builtins.c (warn_for_access): Remove stray warning text.
20612 2021-03-02 Martin Sebor <msebor@redhat.com>
20614 PR middle-end/99295
20615 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
20618 2021-03-02 Jakub Jelinek <jakub@redhat.com>
20621 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
20622 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
20623 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
20624 DW_MACRO_define_strx and DW_MACRO_undef_strx.
20625 (save_macinfo_strings): Use DW_MACRO_*_str* even with
20626 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
20627 DW_MACRO_undef_strx.
20629 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
20631 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
20633 (BT_FN_V8HI_V8HI_UINT): Likewise.
20634 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
20635 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
20636 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
20637 New builtin definitions.
20638 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
20639 vector extension version.
20640 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
20641 available with current -march level.
20642 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
20643 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
20644 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
20645 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
20646 (vec_extend_to_fp32_lo): Likewise.
20647 (vec_round_from_fp32): Likewise.
20648 (vec_convert_to_fp16): Likewise.
20649 (vec_convert_from_fp16): Likewise.
20650 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
20651 (vclfnls_v8hi): Likewise.
20652 (vcrnfs_v8hi): Likewise.
20653 (vcfn_v8hi): Likewise.
20654 (vcnf_v8hi): Likewise.
20656 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
20658 * common/config/s390/s390-common.c (processor_flags_table): New entry.
20659 * config.gcc: Enable arch14 for --with-arch and --with-tune.
20660 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
20661 arch14 for unknown CPU models.
20662 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
20663 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
20664 (s390_get_sched_attrmask): Likewise.
20665 (s390_get_unit_mask): Likewise.
20666 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
20667 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
20668 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
20669 (TARGET_NNPA_P): New macro definitions.
20670 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
20671 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
20673 2021-03-02 Jakub Jelinek <jakub@redhat.com>
20675 PR middle-end/95757
20676 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
20677 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
20678 != 1 comparisons if name is lhs of a comparison.
20680 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
20684 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
20685 * config/darwin.c (darwin_should_restore_cfa_state): New.
20686 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
20687 * doc/tm.texi: Regenerated.
20688 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
20689 * dwarf2cfi.c (connect_traces): If the target requests, restore
20690 the CFA expression after a DW_CFA_restore.
20691 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
20693 2021-03-01 Martin Liska <mliska@suse.cz>
20696 * optc-save-gen.awk: Add 4 more exceptions.
20698 2021-03-01 Nathan Sidwell <nathan@acm.org>
20701 * tree.h (TYPE_ALIGN_RAW): New accessor.
20702 (TYPE_ALIGN): Use it.
20704 2021-03-01 Jan Hubicka <jh@suse.cz>
20707 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
20709 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
20712 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20713 point the hard frame pointer to the SSE register save area instead
20714 of the general register save area. Perform only minimal adjustment
20715 for small frames if it is initially not correctly aligned.
20716 (ix86_expand_prologue): Remove early saves for a SEH target.
20717 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
20719 2021-02-28 Jakub Jelinek <jakub@redhat.com>
20722 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
20723 typo - referneced -> referenced.
20724 * tree.c (component_ref_size): Fix comment typo -
20725 refernce -> reference.
20726 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
20727 traling -> trailing.
20728 (aliasing_component_refs_p): Fix comment typos -
20729 refernce -> reference and refernece -> reference and
20730 traling -> trailing.
20731 (nonoverlapping_refs_since_match_p): Fix comment typo -
20732 referneces -> references.
20733 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
20734 referneces -> references.
20736 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
20738 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
20739 diagnostic message to avoid use of a contraction and format
20742 2021-02-27 Jakub Jelinek <jakub@redhat.com>
20745 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
20746 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
20747 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
20748 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
20751 2021-02-26 Richard Biener <rguenther@suse.de>
20753 PR middle-end/99281
20754 * expr.c (store_field): For calls with return-slot optimization
20755 and addressable return type expand the store directly.
20757 2021-02-26 Richard Biener <rguenther@suse.de>
20760 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
20762 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
20765 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
20768 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
20770 * config.gcc: Add rs6000-pcrel-opt.o.
20771 * config/rs6000/rs6000-pcrel-opt.c: New file.
20772 * config/rs6000/pcrel-opt.md: New file.
20773 * config/rs6000/predicates.md: Add d_form_memory predicate.
20774 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
20775 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
20776 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
20777 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
20778 and make_pass_pcrel_opt().
20779 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
20780 (rs6000_option_override_internal): Add pcrel-opt.
20781 (rs6000_delegitimize_address): Support pcrel-opt.
20782 (rs6000_opt_masks): Add pcrel-opt.
20783 (pcrel_opt_valid_mem_p): New function.
20784 (reg_to_non_prefixed): Make global.
20785 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
20786 (output_pcrel_opt_reloc): New function.
20787 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
20788 (pcrel_extern_addr): Set loads_extern_addr.
20789 Add include for pcrel-opt.md.
20790 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
20791 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
20794 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
20797 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
20798 If TARGET_64BIT and dest is SUBREG, we check the width, if it
20799 equal to SImode, we use SImode operation, just like what we are
20802 2021-02-26 Marek Polacek <polacek@redhat.com>
20804 * builtins.c (warn_for_access): Fix typos.
20806 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
20808 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
20809 mark in front of the immediate quantity.
20810 (<optab>_rolsi3_uxtw): Likewise.
20812 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
20815 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
20816 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
20817 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
20818 address and disable when the FPCXT is not available.
20819 (nonsecure_call_value_reg_thumb2): Likewise.
20821 2021-02-25 Nathan Sidwell <nathan@acm.org>
20824 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
20826 2021-02-25 Tamar Christina <tamar.christina@arm.com>
20828 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
20830 2021-02-25 Richard Biener <rguenther@suse.de>
20832 PR tree-optimization/99253
20833 * tree-vect-loop.c (check_reduction_path): First compute
20834 code, then verify out-of-loop uses.
20836 2021-02-25 Jakub Jelinek <jakub@redhat.com>
20839 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
20841 2021-02-25 Jakub Jelinek <jakub@redhat.com>
20843 PR tree-optimization/80635
20844 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
20845 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
20846 has mode precision.
20848 2021-02-25 Richard Biener <rguenther@suse.de>
20850 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
20851 load_map population.
20852 (vect_match_slp_patterns_2): Revert part of last change.
20853 (vect_analyze_slp): Do not interleave optimize_load_redistribution
20854 with pattern detection but do it afterwards. Dump the
20855 whole SLP graph after pattern recognition and load
20856 redistribution optimization finished.
20858 2021-02-24 Jakub Jelinek <jakub@redhat.com>
20861 * omp-low.c (struct omp_context): Add teams_nested_p and
20862 nonteams_nested_p members.
20863 (scan_omp_target): Diagnose teams nested inside of target with other
20864 directives strictly nested inside of the same target.
20865 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
20866 ctx->nonteams_nested_p as needed.
20868 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
20870 PR inline-asm/99123
20871 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
20873 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
20875 * config/cris/cris.c (cris_expand_prologue): Set
20876 current_function_static_stack_size, if flag_stack_usage_info.
20878 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
20880 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
20881 (rs6000_final_prescan_insn): Adjust.
20882 (rs6000_asm_output_opcode): Likewise.
20884 2021-02-24 Martin Sebor <msebor@redhat.com>
20886 PR middle-end/97172
20887 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
20888 from function arguments.
20890 2021-02-24 Tamar Christina <tamar.christina@arm.com>
20892 PR tree-optimization/99220
20893 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
20894 node from cache when it's about to be deleted.
20896 2021-02-24 Jakub Jelinek <jakub@redhat.com>
20898 PR tree-optimization/99225
20899 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
20900 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
20901 build_int_cst (..., 1). Formatting fixes.
20903 2021-02-24 Tamar Christina <tamar.christina@arm.com>
20905 PR tree-optimization/99149
20906 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
20908 (vect_slp_reset_pattern): Remove.
20909 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
20910 (complex_mul_pattern::build, complex_fma_pattern::build,
20911 complex_fms_pattern::build): Fix ref counts.
20912 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
20913 when node is being deleted.
20914 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
20915 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
20917 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
20919 2021-02-24 Matthias Klose <doko@ubuntu.com>
20922 2020-12-07 Matthias Klose <doko@ubuntu.com>
20924 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
20925 and ENABLE_RTL_FLAG_CHECKING.
20927 2021-02-24 Richard Biener <rguenther@suse.de>
20930 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
20932 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
20934 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
20935 (vsx_assemble_pair): ...to this.
20936 (*mma_assemble_pair): Rename from this...
20937 (*vsx_assemble_pair): ...to this.
20938 (mma_disassemble_pair): Rename from this...
20939 (vsx_disassemble_pair): ...to this.
20940 (*mma_disassemble_pair): Rename from this...
20941 (*vsx_disassemble_pair): ...to this.
20942 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
20943 BU_COMPAT): New macros.
20944 (mma_assemble_pair): Rename from this...
20945 (vsx_assemble_pair): ...to this.
20946 (mma_disassemble_pair): Rename from this...
20947 (vsx_disassemble_pair): ...to this.
20948 (mma_assemble_pair): New compatibility built-in.
20949 (mma_disassemble_pair): Likewise.
20950 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
20951 (RS6000_BUILTIN_COMPAT): Define.
20952 (bdesc_compat): New.
20953 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
20954 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
20955 and VSX_BUILTIN_ASSEMBLE_PAIR.
20956 (rs6000_init_builtins): Register compatibility built-ins.
20957 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
20958 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
20959 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
20960 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
20961 (__builtin_vsx_assemble_pair): ...to this.
20962 (__builtin_mma_disassemble_pair): Rename from this...
20963 (__builtin_vsx_disassemble_pair): ...to this.
20965 2021-02-23 Martin Liska <mliska@suse.cz>
20968 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
20969 with different alignment. That leads to an invalid red zone
20970 size allocated in runtime.
20972 2021-02-23 Jakub Jelinek <jakub@redhat.com>
20974 PR tree-optimization/99204
20975 * fold-const.c (fold_read_from_constant_string): Check that
20976 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
20978 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
20979 Kewen Lin <linkw@gcc.gnu.org>
20981 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
20982 (rotl<mode>3_insert_3): ...this.
20983 (plus_ior_xor): New code_iterator.
20984 (define_split for GPR rl*imi): New splitter.
20985 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
20986 for integer merging.
20988 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20990 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
20992 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
20993 into a register when the above is enabled.
20994 * config/aarch64/aarch64.c (neoversev1_tunings):
20995 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
20996 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
20998 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21000 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21001 valid operand from is now an addi mult-value to shift-value.
21002 * config/cris/cris.md (*addi): Change expression of scaled
21003 operand from mult to ashift.
21004 * config/cris/cris.md (*addi_reload): New insn_and_split.
21006 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21009 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21010 hook_bool_const_tree_hwi_hwi_const_tree_true.
21011 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21013 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21015 PR rtl-optimization/98791
21016 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21017 for unordered modes.
21019 2021-02-22 Martin Liska <mliska@suse.cz>
21021 * tree-inline.c (inline_forbidden_p): Set
21022 inline_forbidden_reason.
21024 2021-02-22 Richard Biener <rguenther@suse.de>
21026 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21029 2021-02-22 Richard Biener <rguenther@suse.de>
21031 PR tree-optimization/99165
21032 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21033 Accumulate changed to ret.
21035 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21038 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21040 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21042 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21045 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21047 (trunctf<DFP_ALL:mode>2): Likewise.
21048 (trunctdtf2_vr): Likewise.
21049 (trunctdtf2): Likewise.
21050 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21051 (extend<DFP_ALL:mode>tf2): Likewise.
21052 (extendtftd2_vr): Likewise.
21053 (extendtftd2): Likewise.
21055 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21057 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21058 add memory alternative.
21059 (tf_to_fprx2): New pattern.
21061 2021-02-19 Martin Sebor <msebor@redhat.com>
21064 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21065 (attr_access::free_lang_data): Remove a spurious test.
21067 2021-02-19 Nathan Sidwell <nathan@acm.org>
21069 * doc/invoke.texi (flang-info-module-read): Document.
21071 2021-02-19 Martin Liska <mliska@suse.cz>
21073 PR translation/99167
21074 * params.opt: Fix typo.
21076 2021-02-19 Richard Biener <rguenther@suse.de>
21078 PR middle-end/99122
21079 * tree-inline.c (inline_forbidden_p): Do not inline functions
21080 with VLA arguments or return value.
21082 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21085 * config/arm/arm.md (*stack_protect_combined_set_insn,
21086 *stack_protect_combined_test_insn): If force_const_mem result
21087 is not valid general operand, force its address into the destination
21090 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21093 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21094 pad or non-local label, put FORCED_LABELs from bb b after that label
21095 rather than before it.
21097 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21100 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21101 expand_vector_broadcast' to emit the vec_duplicate operand.
21103 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21105 PR rtl-optimization/96264
21106 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21109 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21112 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21113 looking up the retain attribute.
21114 (resolve_unique_section): Likewise.
21115 (get_variable_section): Likewise.
21116 (switch_to_section): Likewise. Warn when a symbol without the
21117 retain attribute and a symbol with the retain attribute are
21118 placed in the section with the same name, instead of the used
21120 * doc/extend.texi: Document the "retain" attribute.
21122 2021-02-18 Nathan Sidwell <nathan@acm.org>
21125 * doc/invoke.texi (flang-info-include-translate): Document header
21128 2021-02-18 Richard Biener <rguenther@suse.de>
21130 PR middle-end/99122
21131 * ipa-fnsummary.c (analyze_function_body): Set
21132 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21133 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21134 error_mark_node values.
21135 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21137 (setup_one_parameter): Delay force_value_to_type until when
21140 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21142 PR tree-optimization/99142
21143 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21145 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21147 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21148 wide_int_bitmask::wide_int_bitmask (uint64_t),
21149 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21150 wide_int_bitmask::operator ~ () const,
21151 wide_int_bitmask::operator | (wide_int_bitmask) const,
21152 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21154 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21155 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21156 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21157 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21158 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21159 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21160 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21161 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21162 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21163 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21164 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21165 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21166 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21167 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21168 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21169 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21170 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21171 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21172 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21173 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21174 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21175 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21176 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21177 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21179 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21181 PR middle-end/99109
21182 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21183 (build_printable_array_type): ... this. Add nelts argument. For
21184 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21185 nelts, call build_array_type_nelts.
21186 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21187 instead of build_zero_elt_array_type and build_array_type_nelts.
21189 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21192 * config/i386/i386.c (distance_non_agu_define): Don't call
21193 extract_insn_cached here.
21194 (ix86_lea_outperforms): Save and restore recog_data around call
21195 to distance_non_agu_define and distance_agu_use.
21196 (ix86_ok_to_clobber_flags): Remove.
21197 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21198 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21199 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21200 into define_insn. Move the splitting to define_peephole2 and
21201 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21203 2021-02-17 Julian Brown <julian@codesourcery.com>
21205 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21208 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21211 * config/mips/mips.c (mips_symbol_insns): Do not use
21212 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21214 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21216 PR inline-asm/98096
21217 * stmt.c (resolve_operand_name_1): Take inout operands into account
21218 for access to labels by names.
21219 * doc/extend.texi: Describe counting operands for accessing labels.
21221 2021-02-16 Richard Biener <rguenther@suse.de>
21223 PR tree-optimization/38474
21224 * tree-ssa-structalias.c (variable_info::address_taken): New.
21225 (new_var_info): Initialize address_taken.
21226 (process_constraint): Set address_taken.
21227 (solve_constraints): Use the new address_taken flag rather
21228 than is_reg_var for sorting variables.
21229 (dump_constraint): Dump the variable number if the name
21232 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21235 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21236 multiply by 4096 and for inbranch by 8192.
21237 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21238 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21240 2021-02-15 Maya Rashish <coypu@sdf.org>
21242 * config/aarch64/aarch64.c (aarch64_init_builtins):
21243 Call SUBTARGET_INIT_BUILTINS.
21245 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21247 PR rtl-optimization/98872
21248 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21249 if CONST0_RTX is NULL.
21251 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21253 PR rtl-optimization/98863
21254 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21255 (function_info::build_info): Turn into a declaration, moving the
21256 definition to internals.h.
21257 (function_info::bb_walker): Declare.
21258 (function_info::create_reg_use): Likewise.
21259 (function_info::calculate_potential_phi_regs): Take a build_info
21261 (function_info::place_phis, function_info::create_ebbs): Declare.
21262 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21263 (function_info::populate_backedge_phis): Delete.
21264 (function_info::start_block, function_info::end_block): Declare.
21265 (function_info::populate_phi_inputs): Delete.
21266 (function_info::m_potential_phi_regs): Move information to build_info.
21267 * rtl-ssa/internals.h: New file.
21268 (function_info::bb_phi_info): New class.
21269 (function_info::build_info): Moved from functions.h.
21270 Add a constructor and destructor.
21271 (function_info::build_info::ebb_use): Delete.
21272 (function_info::build_info::ebb_def): Likewise.
21273 (function_info::build_info::bb_live_out): Likewise.
21274 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21275 (function_info::build_info::potential_phi_regs): Likewise.
21276 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21277 (function_info::build_info::ebb_def_regs): Likewise.
21278 (function_info::build_info::bb_phis): Likewise.
21279 (function_info::build_info::bb_mem_live_out): Likewise.
21280 (function_info::build_info::bb_to_rpo): Likewise.
21281 (function_info::build_info::def_stack): Likewise.
21282 (function_info::build_info::old_def_stack_limit): Likewise.
21283 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21284 Remove the regno argument. Push the previous definition onto the
21285 definition stack where necessary.
21286 * rtl-ssa/accesses.cc: Include internals.h.
21287 * rtl-ssa/changes.cc: Likewise.
21288 * rtl-ssa/blocks.cc: Likewise.
21289 (function_info::build_info::build_info): Define.
21290 (function_info::build_info::~build_info): Likewise.
21291 (function_info::bb_walker): New class.
21292 (function_info::bb_walker::bb_walker): Define.
21293 (function_info::add_live_out_use): Convert a logarithmic-complexity
21294 test into a linear one. Allow the same definition to be passed
21296 (function_info::calculate_potential_phi_regs): Moved from
21297 functions.cc. Take a build_info parameter and store the
21298 information there instead.
21299 (function_info::place_phis): New function.
21300 (function_info::add_entry_block_defs): Update call to record_reg_def.
21301 (function_info::calculate_ebb_live_in_for_debug): New function.
21302 (function_info::add_phi_nodes): Use bb_phis to decide which
21303 registers need phi nodes and initialize ebb_def_regs accordingly.
21304 Do not add degenerate phis here.
21305 (function_info::add_artificial_accesses): Use create_reg_use.
21306 Assert that all definitions are listed in the DF LR sets.
21307 Update call to record_reg_def.
21308 (function_info::record_block_live_out): Record live-out register
21309 values in the phis of successor blocks. Use the live-out set
21310 when processing the last block in an EBB, instead of always
21311 using the live-in sets of successor blocks. AND the live sets
21312 with the set of registers that have been defined in the EBB,
21313 rather than with all potential phi registers. Cope correctly
21314 with branches back to the start of the current EBB.
21315 (function_info::start_block): New function.
21316 (function_info::end_block): Likewise.
21317 (function_info::populate_phi_inputs): Likewise.
21318 (function_info::create_ebbs): Likewise.
21319 (function_info::process_all_blocks): Rewrite into a multi-phase
21321 * rtl-ssa/functions.cc: Include internals.h.
21322 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21323 (function_info::init_function_data): Remove caller.
21324 * rtl-ssa/insns.cc: Include internals.h
21325 (function_info::create_reg_use): New function. Lazily any
21326 degenerate phis needed by the linear RPO view.
21327 (function_info::record_use): Use create_reg_use. When processing
21328 debug uses, use potential_phi_regs and test it before checking
21329 whether the register is live on entry to the current EBB. Lazily
21330 calculate ebb_live_in_for_debug.
21331 (function_info::record_call_clobbers): Update call to record_reg_def.
21332 (function_info::record_def): Likewise.
21334 2021-02-15 Martin Liska <mliska@suse.cz>
21336 * toplev.c (init_asm_output): Free output of
21337 gen_command_line_string function.
21338 (process_options): Likewise.
21340 2021-02-15 Martin Liska <mliska@suse.cz>
21342 * params.opt: Add 2 missing Param keywords.
21344 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
21346 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
21348 2021-02-15 Jakub Jelinek <jakub@redhat.com>
21350 PR tree-optimization/99079
21351 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
21352 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
21353 require both type and TREE_TYPE (@1) to be integral types and either
21354 type having smaller or equal precision, or TREE_TYPE (@1) being
21355 unsigned type, or type being signed type. If TREE_TYPE (@1)
21356 doesn't have wrapping overflow, perform the subtraction of one in
21359 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
21360 Richard Biener <rguether@suse.de>
21363 * ipa-reference.c (ipa_init): Only conditinally initialize
21364 reference_vars_to_consider.
21365 (propagate): Conditionally deninitialize reference_vars_to_consider.
21366 (ipa_reference_write_optimization_summary): Sanity check that
21367 reference_vars_to_consider is not allocated.
21369 2021-02-13 Levy Hsu <admin@levyhsu.com>
21372 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
21373 extend parameter to get_si_mem_base_reg declaration.
21374 (get_si_mem_base_reg): Add extend parameter. Set it.
21375 (analyze): Pass extend arg to get_si_mem_base_reg.
21376 (transform): Likewise. Use it when rewriting mems.
21377 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
21378 loads and emit sign/zero extending load followed by subreg move.
21380 2021-02-13 Jim Wilson <jimw@sifive.com>
21383 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
21384 exit when !reload_completed. Only perform check for compressed reg
21385 if reload_completed.
21386 (riscv_rtx_costs): In MEM case, when optimizing for size and
21387 shorten memrefs, if not compressible, then increase cost.
21389 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21391 PR rtl-optimization/98439
21392 * recog.c (pass_split_before_regstack::gate): Enable even when
21393 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
21396 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21399 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
21400 swap of V2SImode elements in memory into DImode memory rotate by 32.
21402 2021-02-12 Martin Sebor <msebor@redhat.com>
21404 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
21406 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21408 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
21409 m_temp_obstack rather than m_obstack to allocate the temporary use.
21411 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21413 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
21414 as read-modify operations.
21416 2021-02-12 Richard Biener <rguenther@suse.de>
21418 PR middle-end/38474
21419 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
21420 fbi->aa_walk_budget is bigger than zero. Update
21421 fbi->aa_walk_budget.
21422 (param_change_prob): Likewise.
21423 * ipa-prop.c (detect_type_change_from_memory_writes):
21424 Properly account walk_aliased_vdefs.
21425 (parm_preserved_before_stmt_p): Canonicalize updates.
21426 (parm_ref_data_preserved_p): Likewise.
21427 (parm_ref_data_pass_through_p): Likewise.
21428 (determine_known_aggregate_parts): Account own alias queries.
21430 2021-02-12 Martin Liska <mliska@suse.cz>
21432 * opts-common.c (decode_cmdline_option): Release werror_arg.
21433 * opts.c (gen_producer_string): Release output of
21434 gen_command_line_string.
21436 2021-02-12 Richard Biener <rguenther@suse.de>
21438 PR tree-optimization/38474
21439 * params.opt (-param=max-store-chains-to-track=): New param.
21440 (-param=max-stores-to-track=): Likewise.
21441 * doc/invoke.texi (max-store-chains-to-track): Document.
21442 (max-stores-to-track): Likewise.
21443 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
21445 (pass_store_merging::m_n_stores): Likewise.
21446 (pass_store_merging::terminate_and_process_chain): Update
21447 m_n_stores and m_n_chains.
21448 (pass_store_merging::process_store): Likewise. Terminate
21449 oldest chains if the number of stores or chains get too large.
21450 (imm_store_chain_info::terminate_and_process_chain): Dump
21453 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
21455 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
21456 the cold section, emit a nop before the directive if the previous
21457 active instruction can throw.
21459 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
21462 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
21463 memory addresses that are legal for quad word accesses.
21465 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
21468 * config/arm/thumb2.md (*doloop_end_internal): Generate
21469 alternative sequence to handle long range branches.
21471 2021-02-11 Joel Hutton <joel.hutton@arm.com>
21473 PR tree-optimization/98772
21474 * optabs-tree.c (supportable_half_widening_operation): New function
21475 to check for supportable V8QI->V8HI widening patterns.
21476 * optabs-tree.h (supportable_half_widening_operation): New function.
21477 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
21478 to create promotion stmts for V8QI->V8HI widening patterns.
21479 (vectorizable_conversion): Add case for V8QI->V8HI.
21481 2021-02-11 Richard Biener <rguenther@suse.de>
21483 * sparseset.h (SPARSESET_ELT_BITS): Remove.
21484 (SPARSESET_ELT_TYPE): Use unsigned int.
21485 * fwprop.c: Do not include sparseset.h.
21487 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21490 * varasm.c (declare_weak): For -fsyntax-only, allow even
21491 TREE_ASM_WRITTEN function decls.
21493 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21496 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
21497 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
21498 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
21499 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
21500 calling simplify_gen_subreg on it.
21502 2021-02-10 Martin Liska <mliska@suse.cz>
21504 * config/nvptx/nvptx.c (nvptx_option_override): Use
21505 flag_patchable_function_entry instead of the removed
21506 function_entry_patch_area_size.
21508 2021-02-10 Martin Liska <mliska@suse.cz>
21510 PR tree-optimization/99002
21511 PR tree-optimization/99026
21512 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
21513 leak when adjacent cases are merged.
21514 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
21516 (make_pass_lower_switch): Remove trailing whitespace.
21517 * tree-switch-conversion.h (release_clusters): New.
21519 2021-02-10 Richard Biener <rguenther@suse.de>
21521 PR rtl-optimization/99054
21522 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
21523 (fixup_partitions): Adjust.
21524 (rtl_verify_edges): Likewise.
21526 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21528 PR middle-end/99007
21529 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
21530 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
21533 2021-02-10 Richard Biener <rguenther@suse.de>
21536 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
21539 2021-02-10 Richard Biener <rguenther@suse.de>
21541 PR tree-optimization/99024
21542 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
21543 clear loop->aux if it is associated with the destroyed loop_vinfo.
21545 2021-02-10 Martin Liska <mliska@suse.cz>
21547 PR tree-optimization/99002
21548 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
21551 2021-02-10 Martin Liska <mliska@suse.cz>
21554 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
21555 a reference exists.
21557 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21560 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
21561 at class scope for DWARF5+.
21563 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
21565 PR rtl-optimization/96015
21566 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
21567 (relax_delay_slots): When deleting a jump to the next active
21568 instruction over a barrier, first delete the barrier if the
21569 jump is the only way to reach the target label.
21571 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
21573 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
21574 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
21575 vector multiplies and vect.alu for SSRA.
21576 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
21577 vect.mul cost field.
21578 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
21579 * config/arm/arm.c: Likewise.
21581 2021-02-09 Richard Biener <rguenther@suse.de>
21583 PR tree-optimization/98863
21584 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
21585 * tree-ssa-sccvn.c (last_pushed_avail): New global.
21586 (rpo_elim::eliminate_push_avail): Chain pushed avails.
21587 (unwind_state::avail_top): Add.
21588 (do_unwind): Rewrite unwinding of avail entries.
21589 (do_rpo_vn): Initialize last_pushed_avail and
21590 avail_top of the undo state.
21592 2021-02-09 Jakub Jelinek <jakub@redhat.com>
21594 PR middle-end/99004
21595 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
21596 const char * to char * and free those pointers after use.
21598 2021-02-09 Richard Biener <rguenther@suse.de>
21600 PR tree-optimization/99017
21601 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
21602 zero vector cost entries.
21604 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
21606 PR middle-end/98974
21607 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
21608 parameter in vectorizable_condition.
21610 2021-02-08 Richard Biener <rguenther@suse.de>
21613 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
21615 2021-02-08 Martin Liska <mliska@suse.cz>
21618 * cfgexpand.c (pass_expand::execute): Parse per-function option
21619 flag_patchable_function_entry and use it.
21620 * common.opt: Remove function_entry_patch_area_size and
21621 function_entry_patch_area_start global variables.
21622 * opts.c (parse_and_check_patch_area): New function.
21623 (common_handle_option): Use it.
21624 * opts.h (parse_and_check_patch_area): New function.
21625 * toplev.c (process_options): Parse and use
21626 function_entry_patch_area_size.
21628 2021-02-08 Martin Sebor <msebor@redhat.com>
21630 * doc/extend.texi (attribute malloc): Correct typos.
21632 2021-02-05 Nathan Sidwell <nathan@acm.org>
21635 * gcc.c (driver::maybe_run_linker): Check for input file
21636 accessibility if not linking.
21638 2021-02-05 Richard Biener <rguenther@suse.de>
21640 PR tree-optimization/98855
21641 * tree-vectorizer.h (add_stmt_cost): New overload.
21642 * tree-vect-slp.c (li_cost_vec_cmp): New.
21643 (vect_bb_slp_scalar_cost): Cost individual loop regions
21644 separately. Account for the scalar instance root stmt.
21646 2021-02-05 Tom de Vries <tdevries@suse.de>
21649 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
21651 (bit_test_cluster::emit): Reuse location_t for newly created
21653 (switch_decision_tree::try_switch_expansion): Preserve
21655 * tree-switch-conversion.h: Change function signatures.
21657 2021-02-05 Jakub Jelinek <jakub@redhat.com>
21660 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
21661 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
21662 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
21663 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
21665 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21667 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
21668 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
21669 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
21670 (vget_high_f16): Reimplement using new builtin.
21671 (vget_high_f32): Likewise.
21672 (vget_high_f64): Likewise.
21673 (vget_high_p8): Likewise.
21674 (vget_high_p16): Likewise.
21675 (vget_high_p64): Likewise.
21676 (vget_high_s8): Likewise.
21677 (vget_high_s16): Likewise.
21678 (vget_high_s32): Likewise.
21679 (vget_high_s64): Likewise.
21680 (vget_high_u8): Likewise.
21681 (vget_high_u16): Likewise.
21682 (vget_high_u32): Likewise.
21683 (vget_high_u64): Likewise.
21685 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21687 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
21688 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
21689 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
21690 (vget_low_f16): Reimplement using new builtin.
21691 (vget_low_f32): Likewise.
21692 (vget_low_f64): Likewise.
21693 (vget_low_p8): Likewise.
21694 (vget_low_p16): Likewise.
21695 (vget_low_p64): Likewise.
21696 (vget_low_s8): Likewise.
21697 (vget_low_s16): Likewise.
21698 (vget_low_s32): Likewise.
21699 (vget_low_s64): Likewise.
21700 (vget_low_u8): Likewise.
21701 (vget_low_u16): Likewise.
21702 (vget_low_u32): Likewise.
21703 (vget_low_u64): Likewise.
21705 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
21707 * gcc.c (print_multilib_info): Check all required argument is provided
21710 2021-02-05 liuhongt <hongtao.liu@intel.com>
21713 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
21714 generate integer mask comparison for 128/256-bits vector when
21715 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
21716 delete redundant !maskcmp condition.
21717 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
21719 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
21720 condition directly to if (maskcmp), add extra check for
21721 cmpmode, it should be MODE_INT.
21722 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
21723 parameters op_true/op_false.
21724 (ix86_use_mask_cmp_p): New.
21726 2021-02-05 liuhongt <hongtao.liu@intel.com>
21729 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
21730 Remove m_GENERIC from ~list.
21731 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
21733 2021-02-04 David Malcolm <dmalcolm@redhat.com>
21736 * diagnostic-show-locus.c (compatible_locations_p): Require
21737 locations in the same macro map to be either both from the
21738 macro definition, or both from the macro arguments.
21740 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
21742 * config/aarch64/aarch64-simd-builtins.def: Add
21743 [su]mull_hi_lane[q] builtin generator macros.
21744 * config/aarch64/aarch64-simd.md
21745 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
21746 (aarch64_<su>mull_hi_lane<mode>): Define.
21747 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
21748 (aarch64_<su>mull_hi_laneq<mode>): Define.
21749 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
21750 builtin instead of inline asm.
21751 (vmull_high_lane_s32): Likewise.
21752 (vmull_high_lane_u16): Likewise.
21753 (vmull_high_lane_u32): Likewise.
21754 (vmull_high_laneq_s16): Likewise.
21755 (vmull_high_laneq_s32): Likewise.
21756 (vmull_high_laneq_u16): Likewise.
21757 (vmull_high_laneq_u32): Liekwise.
21759 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
21761 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
21762 builtin generator macros.
21763 * config/aarch64/aarch64-simd.md
21764 (aarch64_<su>mull_hi_n<mode>_insn): Define.
21765 (aarch64_<su>mull_hi_n<mode>): Define.
21766 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
21767 instead of inline asm.
21768 (vmull_high_n_s32): Likewise.
21769 (vmull_high_n_u16): Likewise.
21770 (vmull_high_n_u32): Likewise.
21772 2021-02-04 Richard Biener <rguenther@suse.de>
21774 PR tree-optimization/98855
21775 * tree-vect-loop.c (vectorizable_phi): Do not cost
21776 single-argument PHIs.
21777 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
21778 * tree-vect-stmts.c (vectorizable_bswap): Also perform
21779 costing for SLP operation.
21781 2021-02-04 Martin Liska <mliska@suse.cz>
21783 * doc/extend.texi: Mention -mprefer-vector-width in target
21786 2021-02-03 Martin Sebor <msebor@redhat.com>
21788 PR tree-optimization/98937
21789 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
21790 Flush pointer_query cache.
21792 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
21794 * config/rs6000/genfusion.pl (gen_2logical): Add missing
21795 fixes based on patch review.
21796 * config/rs6000/fusion.md: Regenerate file.
21798 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
21800 * config/rs6000/t-rs6000: Comment out auto generation of
21803 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
21805 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
21806 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
21807 (output_file_start): Add gfx908.
21808 * config/gcn/gcn.opt (gpu_type): Add gfx908.
21809 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
21810 (MULTILIB_DIRNAMES): Add gfx908.
21811 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
21812 (main): Recognize gfx908.
21813 * config/gcn/t-omp-device: Add gfx908.
21815 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
21817 * config/aarch64/aarch64-simd-builtins.def: Add
21818 [su]mlsl_hi_lane[q] builtin macro generators.
21819 * config/aarch64/aarch64-simd.md
21820 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
21821 (aarch64_<su>mlsl_hi_lane<mode>): Define.
21822 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
21823 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
21824 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
21825 builtin instead of inline asm.
21826 (vmlsl_high_lane_s32): Likewise.
21827 (vmlsl_high_lane_u16): Likewise.
21828 (vmlsl_high_lane_u32): Likewise.
21829 (vmlsl_high_laneq_s16): Likewise.
21830 (vmlsl_high_laneq_s32): Likewise.
21831 (vmlsl_high_laneq_u16): Likewise.
21832 (vmlsl_high_laneq_u32): Likewise.
21833 (vmlal_high_laneq_u32): Likewise.
21835 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
21837 * config/aarch64/aarch64-simd-builtins.def: Add
21838 [su]mlal_hi_lane[q] builtin generator macros.
21839 * config/aarch64/aarch64-simd.md
21840 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
21841 (aarch64_<su>mlal_hi_lane<mode>): Define.
21842 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
21843 (aarch64_<su>mlal_hi_laneq<mode>): Define.
21844 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
21845 builtin instead of inline asm.
21846 (vmlal_high_lane_s32): Likewise.
21847 (vmlal_high_lane_u16): Likewise.
21848 (vmlal_high_lane_u32): Likewise.
21849 (vmlal_high_laneq_s16): Likewise.
21850 (vmlal_high_laneq_s32): Likewise.
21851 (vmlal_high_laneq_u16): Likewise.
21852 (vmlal_high_laneq_u32): Likewise.
21854 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
21856 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
21857 builtin generator macros.
21858 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
21860 (aarch64_<su>mlsl_hi_n<mode>): Define.
21861 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
21862 instead of inline asm.
21863 (vmlsl_high_n_s32): Likewise.
21864 (vmlsl_high_n_u16): Likewise.
21865 (vmlsl_high_n_u32): Likewise.
21867 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
21869 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
21870 builtin generator macros.
21871 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
21873 (aarch64_<su>mlal_hi_n<mode>): Define.
21874 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
21875 instead of inline asm.
21876 (vmlal_high_n_s32): Likewise.
21877 (vmlal_high_n_u16): Likewise.
21878 (vmlal_high_n_u32): Likewise.
21880 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
21882 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
21884 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
21886 (aarch64_<su>mlal_hi<mode>_insn): This.
21887 (aarch64_<su>mlal_hi<mode>): Define.
21888 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
21889 instead of inline asm.
21890 (vmlal_high_s16): Likewise.
21891 (vmlal_high_s32): Likewise.
21892 (vmlal_high_u8): Likewise.
21893 (vmlal_high_u16): Likewise.
21894 (vmlal_high_u32): Likewise.
21896 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
21898 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
21899 after calling alter_subreg() on a (mem).
21901 2021-02-03 Martin Liska <mliska@suse.cz>
21904 * lto-streamer-out.c (produce_lto_section): Fill up missing
21906 * lto-streamer.h (struct lto_section): Add _padding field.
21908 2021-02-03 Richard Biener <rguenther@suse.de>
21910 * lto-streamer.c (lto_get_section_name): Free temporary
21912 * tree-loop-distribution.c
21913 (loop_distribution::merge_dep_scc_partitions): Free edge data.
21915 2021-02-03 Jakub Jelinek <jakub@redhat.com>
21917 PR middle-end/97487
21918 * ifcvt.c (noce_can_force_operand): New function.
21919 (noce_emit_move_insn): Use it.
21920 (noce_try_sign_mask): Likewise. Formatting fix.
21922 2021-02-03 Jakub Jelinek <jakub@redhat.com>
21924 PR middle-end/97971
21925 * lra-constraints.c (process_alt_operands): For inline asm, don't call
21926 fatal_insn, but instead return false.
21928 2021-02-03 Jakub Jelinek <jakub@redhat.com>
21930 PR tree-optimization/98287
21931 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
21934 2021-02-03 Tamar Christina <tamar.christina@arm.com>
21936 PR tree-optimization/98928
21937 * tree-vect-loop.c (vect_analyze_loop_2): Change
21938 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
21939 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
21940 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
21941 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
21943 2021-02-02 Richard Biener <rguenther@suse.de>
21945 * gimple-loop-interchange.cc (prepare_data_references):
21947 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
21948 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
21949 * tree-vect-stmts.c (vectorizable_condition): Do not
21951 (vectorizable_comparison): Likewise.
21953 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21955 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
21956 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
21957 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
21958 (vrsqrteq_u32): Likewise.
21960 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21962 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
21963 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
21964 (aarch64_sqxtun2<mode>_be): Likewise.
21965 (aarch64_sqxtun2<mode>): Likewise.
21966 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
21967 (vqmovun_high_s32): Likewise.
21968 (vqmovun_high_s64): Likewise.
21969 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
21971 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21973 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
21975 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
21977 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21979 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
21980 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
21981 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
21982 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
21983 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
21984 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
21985 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
21986 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
21988 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21990 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
21991 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
21992 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
21994 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21996 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
21997 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
21999 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22001 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22002 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22003 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22004 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22005 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22006 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22007 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22008 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22009 NONE builtin flags.
22011 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22013 PR tree-optimization/98848
22014 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22015 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22017 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22020 * expr.c: Check mode before calling store_expr.
22022 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22024 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22026 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22027 instruction using expression ior.
22028 (mve_vornq_u<mode>): New expander.
22029 (mve_vornq_f<mode>): Use ior code instead of unspec.
22030 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22032 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22034 * tree-nested.c (convert_nonlocal_reference_op): Move
22035 current_function_decl restore after re-gimplification.
22036 (convert_local_reference_op): Likewise.
22038 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22040 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22042 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22044 (aarch64_rshrn<mode>_insn_be): Likewise.
22045 (aarch64_rshrn<mode>): Likewise.
22046 (aarch64_rshrn2<mode>_insn_le): Likewise.
22047 (aarch64_rshrn2<mode>_insn_be): Likewise.
22048 (aarch64_rshrn2<mode>): Likewise.
22049 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22050 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22052 (vrshrn_high_n_s32): Likewise.
22053 (vrshrn_high_n_s64): Likewise.
22054 (vrshrn_high_n_u16): Likewise.
22055 (vrshrn_high_n_u32): Likewise.
22056 (vrshrn_high_n_u64): Likewise.
22057 (vrshrn_n_s16): Likewise.
22058 (vrshrn_n_s32): Likewise.
22059 (vrshrn_n_s64): Likewise.
22060 (vrshrn_n_u16): Likewise.
22061 (vrshrn_n_u32): Likewise.
22062 (vrshrn_n_u64): Likewise.
22064 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22066 PR tree-optimization/98499
22067 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22068 conservatively and assume all possible side-effects.
22070 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22072 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22073 vec_unpacku_hi_): Define builtins.
22074 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22076 (vmovl_high_s16): Likewise.
22077 (vmovl_high_s32): Likewise.
22078 (vmovl_high_u8): Likewise.
22079 (vmovl_high_u16): Likewise.
22080 (vmovl_high_u32): Likewise.
22082 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22084 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22086 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22088 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22090 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22092 (vabdl_s16): Likewise.
22093 (vabdl_s32): Likewise.
22094 (vabdl_u8): Likewise.
22095 (vabdl_u16): Likewise.
22096 (vabdl_u32): Likewise.
22097 * config/aarch64/iterators.md (ABDL): New int iterator.
22098 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22100 2021-02-01 Martin Sebor <msebor@redhat.com>
22102 * tree.h (BLOCK_VARS): Add comment.
22103 (BLOCK_SUBBLOCKS): Same.
22104 (BLOCK_SUPERCONTEXT): Same.
22105 (BLOCK_ABSTRACT_ORIGIN): Same.
22106 (inlined_function_outer_scope_p): Same.
22108 2021-02-01 Martin Sebor <msebor@redhat.com>
22110 PR middle-end/97172
22111 * attribs.c (attr_access::free_lang_data): Define new function.
22112 * attribs.h (attr_access::free_lang_data): Declare new function.
22114 2021-02-01 Richard Biener <rguenther@suse.de>
22116 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22118 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22120 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22122 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22123 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22125 2021-02-01 Richard Biener <rguenther@suse.de>
22127 PR rtl-optimization/98863
22128 * config/i386/i386-features.c (convert_scalars_to_vector):
22129 Set DF_RD_PRUNE_DEAD_DEFS.
22131 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22133 * system.h (SIZE_MAX): Define if not already defined.
22135 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22137 * config/rs6000/genfusion.pl (gen_2logical): New function to
22138 generate patterns for logical-logical fusion.
22139 * config/rs6000/fusion.md: Regenerated patterns.
22140 * config/rs6000/rs6000-cpus.def: Add
22141 OPTION_MASK_P10_FUSION_2LOGICAL.
22142 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22143 Enable logical-logical fusion for p10.
22144 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22146 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22148 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22150 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22152 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22153 (mabi=vec-default): New.
22154 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22155 __EXTABI__ for AIX Vector extended ABI.
22156 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22158 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22160 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22162 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22164 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22165 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22167 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22170 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22173 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22175 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22176 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22177 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22179 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22181 PR tree-optimization/98866
22182 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22183 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22184 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22185 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22186 (gori_map::is_export_p): Ditto.
22187 (gori_map::calculate_gori): Ditto.
22188 (gori_compute::set_range_invariant): New.
22189 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22190 invariant for pointers evaluating to [1, +INF].
22192 2021-01-29 Richard Biener <rguenther@suse.de>
22194 PR rtl-optimization/98863
22195 * config/i386/i386-features.c (remove_partial_avx_dependency):
22196 Do not perform DF analysis.
22197 (pass_data_remove_partial_avx_dependency): Remove
22200 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22202 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22203 builtin generator macros.
22204 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22206 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22207 instead of inline asm.
22208 (vmull_n_s32): Likewise.
22209 (vmull_n_u16): Likewise.
22210 (vmull_n_u32): Likewise.
22212 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22214 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22216 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22218 (aarch64_<sur>abdl2<mode>): ... This.
22219 (<sur>sadv16qi): Adjust use of above.
22220 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22222 (vabdl_high_s16): Likewise.
22223 (vabdl_high_s32): Likewise.
22224 (vabdl_high_u8): Likewise.
22225 (vabdl_high_u16): Likewise.
22226 (vabdl_high_u32): Likewise.
22228 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22230 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22232 (uabal2): Likewise.
22233 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22235 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22237 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22239 (vabal_high_s16): Likewise.
22240 (vabal_high_s32): Likewise.
22241 (vabal_high_u8): Likewise.
22242 (vabal_high_u16): Likewise.
22243 (vabal_high_u32): Likewise.
22244 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22245 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22247 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22249 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22252 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22254 (aarch64_<sur>abal<mode>): ... This
22255 (<sur>sadv16qi): Adust use of the above.
22256 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22258 (vabal_s16): Likewise.
22259 (vabal_s32): Likewise.
22260 (vabal_u8): Likewise.
22261 (vabal_u16): Likewise.
22262 (vabal_u32): Likewise.
22264 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22266 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22268 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22270 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22272 (vaddlv_s16): Likewise.
22273 (vaddlv_u8): Likewise.
22274 (vaddlv_u16): Likewise.
22275 (vaddlvq_s8): Likewise.
22276 (vaddlvq_s16): Likewise.
22277 (vaddlvq_s32): Likewise.
22278 (vaddlvq_u8): Likewise.
22279 (vaddlvq_u16): Likewise.
22280 (vaddlvq_u32): Likewise.
22281 (vaddlv_s32): Likewise.
22282 (vaddlv_u32): Likewise.
22283 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22284 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22285 (Vwstype): New mode attribute.
22287 (VWIDE_S): Likewise.
22288 (USADDLV): New int iterator.
22289 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22291 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22293 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22294 builtin generator macros.
22295 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22297 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22298 instead of inline asm.
22299 (vmlsl_lane_s32): Likewise.
22300 (vmlsl_lane_u16): Likewise.
22301 (vmlsl_lane_u32): Likewise.
22302 (vmlsl_laneq_s16): Likewise.
22303 (vmlsl_laneq_s32): Likewise.
22304 (vmlsl_laneq_u16): Likewise.
22305 (vmlsl_laneq_u32): Likewise.
22307 2021-01-29 Richard Biener <rguenther@suse.de>
22309 * doc/invoke.texi (--param max-gcse-memory): Document unit
22311 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22312 * params.opt (--param max-gcse-memory): Adjust default and
22313 document unit of size.
22315 2021-01-29 Richard Biener <rguenther@suse.de>
22317 PR rtl-optimization/98863
22318 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22319 HOST_WIDE_INT for the memory estimate.
22321 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22322 Richard Biener <rguenther@suse.de>
22324 PR tree-optimization/97627
22325 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22326 Do not analyze fake edges.
22328 2021-01-29 Richard Biener <rguenther@suse.de>
22330 PR rtl-optimization/98144
22331 * df.h (df_mir_bb_info): Add con_visited member.
22332 * df-problems.c (df_mir_alloc): Initialize con_visited,
22333 do not fully populate IN and OUT.
22334 (df_mir_reset): Likewise.
22335 (df_mir_confluence_0): Set con_visited.
22336 (df_mir_confluence_n): Properly handle implicitely
22337 fully populated IN and OUT as designated by con_visited
22338 and update con_visited accordingly.
22340 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22343 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
22344 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
22345 && !TARGET_REALLY_IWMMXT to conditions.
22347 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22350 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
22353 2021-01-28 Marek Polacek <polacek@redhat.com>
22356 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
22357 the main variant, maybe reset it in its variants too.
22358 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
22359 (check_aligned_type): Check if TYPE_USER_ALIGN match.
22361 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
22364 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
22365 of constant zero for comparisons.
22367 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
22369 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
22370 support for mapping built-in function names for long double
22371 built-in functions if long double is IEEE 128-bit.
22373 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22375 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
22376 builtin generator macros.
22377 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
22379 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
22380 instead of inline asm.
22381 (vmlsl_n_s32): Likewise.
22382 (vmlsl_n_u16): Likewise.
22383 (vmlsl_n_u32): Likewise.
22385 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22387 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
22388 builtin generator macros.
22389 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
22391 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
22392 instead of inline asm.
22393 (vmlal_n_s32): Likewise.
22394 (vmlal_n_u16): Likewise.
22395 (vmlal_n_u32): Likewise.
22397 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22399 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
22401 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
22403 (aarch64_shrn2<mode>_insn_be): Likewise.
22404 (aarch64_shrn2<mode>): Likewise.
22405 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
22407 (vshrn_high_n_s32): Likewise.
22408 (vshrn_high_n_s64): Likewise.
22409 (vshrn_high_n_u16): Likewise.
22410 (vshrn_high_n_u32): Likewise.
22411 (vshrn_high_n_u64): Likewise.
22413 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22415 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
22417 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
22419 (aarch64_shrn<mode>_insn_be): Likewise.
22420 (aarch64_shrn<mode>): Likewise.
22421 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
22423 (vshrn_n_s32): Likewise.
22424 (vshrn_n_s64): Likewise.
22425 (vshrn_n_u16): Likewise.
22426 (vshrn_n_u32): Likewise.
22427 (vshrn_n_u64): Likewise.
22428 * config/aarch64/iterators.md (vn_mode): New mode attribute.
22430 2021-01-28 Richard Biener <rguenther@suse.de>
22432 PR rtl-optimization/80960
22433 * dse.c (check_mem_read_rtx): Call get_addr on the
22436 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
22437 David Edelsohn <dje.gcc@gmail.com>
22440 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22441 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
22443 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22445 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
22446 wrapper call rs6000_expand_vector_set_var for cleanup. Call
22447 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
22449 (rs6000_expand_vector_set_var): Delete.
22450 (rs6000_expand_vector_set_var_p9): Make static.
22451 (rs6000_expand_vector_set_var_p8): Make static.
22453 2021-01-28 Xing GUO <higuoxing@gmail.com>
22455 * common/config/riscv/riscv-common.c
22456 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
22457 when `p` extension exists.
22459 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
22461 PR rtl-optimization/97684
22462 * ira.c (ira): Call ira_set_pseudo_classes before
22463 update_equiv_regs when it is necessary.
22465 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22468 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
22469 %w0, %w1 and %2 instead of %0, %1 and %2.
22471 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
22473 * config/rs6000/genfusion.pl: New script to generate
22474 define_insn_and_split patterns so combine can arrange fused
22475 instructions next to each other.
22476 * config/rs6000/fusion.md: New file, generated fused instruction
22477 patterns for combine.
22478 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
22479 (non_update_memory_operand): New predicate.
22480 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
22481 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
22483 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
22485 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22486 Automatically set OPTION_MASK_P10_FUSION and
22487 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
22488 (rs600_opt_masks): Allow -mpower10-fusion
22489 in function attributes.
22490 (address_is_non_pfx_d_or_x): New function.
22491 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
22492 * config/rs6000/rs6000.md: Include fusion.md.
22493 * config/rs6000/rs6000.opt: Add -mpower10-fusion
22494 and -mpower10-fusion-ld-cmpi.
22495 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
22497 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22499 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
22500 builtin generator macros.
22501 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
22503 (aarch64_<su>mlal<mode>): This.
22504 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
22505 instead of inline asm.
22506 (vmlal_s16): Likewise.
22507 (vmlal_s32): Likewise.
22508 (vmlal_u8): Likewise.
22509 (vmlal_u16): Likewise.
22510 (vmlal_u32): Likewise.
22512 2021-01-27 Richard Biener <rguenther@suse.de>
22514 PR tree-optimization/98854
22515 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
22516 PHIs from scalars when the number of CTORs matches the
22517 number of children.
22519 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22521 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
22523 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
22525 (aarch64_mls_n<mode>): This.
22526 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
22528 (vmls_n_s32): Likewise.
22529 (vmls_n_u16): Likewise.
22530 (vmls_n_u32): Likewise.
22531 (vmlsq_n_s16): Likewise.
22532 (vmlsq_n_s32): Likewise.
22533 (vmlsq_n_u16): Likewise.
22534 (vmlsq_n_u32): Likewise.
22536 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22538 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
22540 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
22542 (vmls_s16): Likewise.
22543 (vmls_s32): Likewise.
22544 (vmls_u8): Likewise.
22545 (vmls_u16): Likewise.
22546 (vmls_u32): Likewise.
22547 (vmlsq_s8): Likewise.
22548 (vmlsq_s16): Likewise.
22549 (vmlsq_s32): Likewise.
22550 (vmlsq_u8): Likewise.
22551 (vmlsq_u16): Likewise.
22552 (vmlsq_u32): Likewise.
22554 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22556 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
22558 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
22560 (aarch64_mla_n<mode>): This.
22561 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
22563 (vmla_n_s32): Likewise.
22564 (vmla_n_u16): Likewise.
22565 (vmla_n_u32): Likewise.
22566 (vmlaq_n_s16): Likewise.
22567 (vmlaq_n_s32): Likewise.
22568 (vmlaq_n_u16): Likewise.
22569 (vmlaq_n_u32): Likewise.
22571 2021-01-27 liuhongt <hongtao.liu@intel.com>
22574 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
22575 (*sse2_eq<mode>3): Ditto.
22577 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22579 * tree-pass.h (PROP_trees): Rename to ...
22580 (PROP_gimple): ... this.
22581 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
22582 * passes.c (execute_function_dump, execute_function_todo,
22583 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
22584 * varpool.c (ctor_for_folding): Likewise.
22586 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22588 PR tree-optimization/97260
22589 * varpool.c: Include tree-pass.h.
22590 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
22591 non-TREE_SIDE_EFFECTS automatic variables.
22593 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
22595 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
22597 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
22598 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
22599 (gen_compile_unit_die): Recognise C++23.
22601 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22604 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
22607 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22610 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
22611 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
22612 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
22614 2021-01-26 Richard Biener <rguenther@suse.de>
22616 * gimple-pretty-print.c (dump_binary_rhs): Handle
22617 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
22619 2021-01-26 Richard Biener <rguenther@suse.de>
22621 PR middle-end/98726
22622 * tree.h (vector_cst_int_elt): Remove.
22623 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
22626 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
22628 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
22629 for V64DFmode min/max reductions.
22631 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22633 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
22634 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
22635 two halves, one with x and the other with const0_rtx, ordered
22636 depending on endianity.
22638 2021-01-26 Alexandre Oliva <oliva@adacore.com>
22640 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
22641 temporaries not seen in binding block, and not about to be
22642 added as gimple variables.
22644 2021-01-25 Martin Sebor <msebor@redhat.com>
22647 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
22649 2021-01-25 Martin Liska <mliska@suse.cz>
22651 * value-prof.c (get_nth_most_common_value): Use %s instead
22654 2021-01-25 Jakub Jelinek <jakub@redhat.com>
22657 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
22658 readelf -wi is able to read the emitted .debug_info back.
22659 * configure: Regenerated.
22661 2021-01-25 Martin Liska <mliska@suse.cz>
22663 PR gcov-profile/98739
22664 * common.opt: Add missing sign symbol.
22665 * value-prof.c (get_nth_most_common_value): Restore handling
22666 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
22667 PROFILE_REPRODUCIBILITY_MULTITHREADED.
22669 2021-01-25 Richard Biener <rguenther@suse.de>
22671 PR middle-end/98807
22672 * tree.c (vector_element_bits): Always use precision of
22673 the element type for boolean vectors.
22675 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
22677 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
22678 (ENDFILE_SPEC): Evaluate qnolinkcmds.
22680 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
22682 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
22683 nostartfiles handling since this is already done by
22684 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
22685 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
22686 is already done by LINK_COMMAND_SPEC.
22687 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
22688 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
22691 2021-01-25 Jakub Jelinek <jakub@redhat.com>
22694 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
22695 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
22696 value rather than host size_t.
22697 (fold_const_call): Change type of s2 from size_t to
22698 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
22699 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
22700 instead of s2 as last argument.
22702 2021-01-25 Tamar Christina <tamar.christina@arm.com>
22704 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
22705 VCMLA_OP, VCMUL_OP): New.
22706 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
22707 * config/arm/neon.md (cmul<conj_op><mode>3): New.
22708 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
22709 UNSPEC_VCMUL_CONJ): New.
22710 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
22711 cml<fcmac1><conj_op><mode>4): New.
22713 2021-01-23 Jakub Jelinek <jakub@redhat.com>
22716 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
22718 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
22720 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
22722 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
22724 (vmla_s16): Likewise.
22725 (vmla_s32): Likewise.
22726 (vmla_u8): Likewise.
22727 (vmla_u16): Likewise.
22728 (vmla_u32): Likewise.
22729 (vmlaq_s8): Likewise.
22730 (vmlaq_s16): Likewise.
22731 (vmlaq_s32): Likewise.
22732 (vmlaq_u8): Likewise.
22733 (vmlaq_u16): Likewise.
22734 (vmlaq_u32): Likewise.
22736 2021-01-22 David Malcolm <dmalcolm@redhat.com>
22738 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
22741 2021-01-22 Jakub Jelinek <jakub@redhat.com>
22744 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
22745 filenames to emit, still emit the required 0 index directory and
22746 filename entries that match DW_AT_comp_dir and DW_AT_name of the
22749 2021-01-22 Marek Polacek <polacek@redhat.com>
22752 * doc/invoke.texi: Update C++ ABI Version 15 description.
22754 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22756 PR tree-optimization/98766
22757 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
22758 comparing against type size with param_avoid_fma_max_bits.
22760 2021-01-22 Richard Biener <rguenther@suse.de>
22762 PR middle-end/98793
22763 * tree.c (vector_element_bits): Key single-bit bool vector on
22764 integer mode rather than not vector mode.
22766 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
22769 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22770 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
22772 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
22773 to call different path for P8 and P9.
22774 (rs6000_expand_vector_set_var_p9): New function.
22775 (rs6000_expand_vector_set_var_p8): New function.
22777 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
22781 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22782 Ajdust variable index vec_insert from address dereference to
22783 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
22784 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22786 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
22788 2021-01-22 Martin Liska <mliska@suse.cz>
22790 PR gcov-profile/98739
22791 * profile.c (compute_value_histograms): Drop time profile for
22792 -fprofile-reproducible=multithreaded.
22794 2021-01-22 Nathan Sidwell <nathan@acm.org>
22796 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
22799 2021-01-22 Richard Biener <rguenther@suse.de>
22801 PR middle-end/98773
22802 * tree-data-ref.c (initalize_matrix_A): Revert previous
22803 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
22805 2021-01-22 Jakub Jelinek <jakub@redhat.com>
22807 PR tree-optimization/90248
22808 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
22809 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
22811 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
22812 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
22814 2021-01-22 Jakub Jelinek <jakub@redhat.com>
22816 PR tree-optimization/98255
22817 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
22818 extend index - low_bound from sizetype's precision rather than index
22820 (get_addr_base_and_unit_offset_1): Likewise.
22821 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
22822 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
22824 2021-01-22 Richard Biener <rguenther@suse.de>
22826 PR tree-optimization/98786
22827 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
22828 adding new uses of abnormals. Verify we deal with a conditional
22831 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22834 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
22836 2021-01-22 liuhongt <hongtao.liu@intel.com>
22840 * config/i386/sse.md (VI_128_256): New mode iterator.
22841 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
22842 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
22843 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
22844 define_insn_and_split to lower avx512 vector comparison to avx
22845 version when dest is vector.
22846 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
22847 define_insn_and_split for negating the comparison result.
22848 * config/i386/predicates.md (float_vector_all_ones_operand):
22850 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
22851 general NOT operator without UNSPEC_MASKOP.
22853 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
22855 PR rtl-optimization/98777
22856 * lra-int.h (lra_pmode_pseudo): New extern.
22857 * lra.c (lra_pmode_pseudo): New global.
22859 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
22861 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
22863 * fwprop.c (fwprop_propagation::classify_result): Allow
22864 (subreg (mem)) simplifications.
22866 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22868 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
22870 (aarch64_sqdmlal<mode>): ... This...
22871 (aarch64_sqdmlsl<mode>): ... And this.
22872 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
22873 (aarch64_sqdmlal_lane<mode>): ... This...
22874 (aarch64_sqdmlsl_lane<mode>): ... And this.
22875 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
22876 (aarch64_sqdmlsl_laneq<mode>): ... This...
22877 (aarch64_sqdmlal_laneq<mode>): ... And this.
22878 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
22879 (aarch64_sqdmlsl_n<mode>): ... This...
22880 (aarch64_sqdmlal_n<mode>): ... And this.
22881 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
22882 (aarch64_sqdmlal2<mode>_internal): ... This...
22883 (aarch64_sqdmlsl2<mode>_internal): ... And this.
22885 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
22887 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
22889 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
22892 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
22894 2021-01-21 liuhongt <hongtao.liu@intel.com>
22896 PR rtl-optimization/98694
22897 * regcprop.c (copy_value): If SRC had been assigned a mode
22898 narrower than the copy, we can't link DEST into the chain even
22899 they have same hard_regno_nregs(i.e. HImode/SImode in i386
22902 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22904 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
22905 Convert to define_insn_and_split. Split into simple move when moving
22908 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
22910 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
22911 Adjust comment. Simplify code.
22913 2021-01-20 Jakub Jelinek <jakub@redhat.com>
22916 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
22917 with DW_FORM_line_strp form.
22918 (prune_unused_types_update_strings): Don't add into debug_str_hash
22919 indirect strings with DW_FORM_line_strp form.
22920 (adjust_name_comp_dir): New function.
22921 (dwarf2out_finish): Call it on CU DIEs after resetting
22922 debug_line_str_hash.
22924 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
22926 PR rtl-optimization/98722
22927 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
22928 has no 3-op add insn to transform insns containing two pluses.
22930 2021-01-20 Richard Biener <rguenther@suse.de>
22932 * hwint.h (add_hwi): New function.
22933 (mul_hwi): Likewise.
22934 * tree-data-ref.c (initialize_matrix_A): Properly translate
22935 tree constants and avoid HOST_WIDE_INT_MIN.
22936 (lambda_matrix_row_add): Avoid undefined integer overflow
22937 and return true on such overflow.
22938 (lambda_matrix_right_hermite): Handle overflow from
22939 lambda_matrix_row_add gracefully. Simplify previous fix.
22940 (analyze_subscript_affine_affine): Likewise.
22942 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
22944 PR tree-optimization/96674
22945 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
22946 x >= y && y != XXX_MIN --> x > y - 1
22948 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
22950 PR tree-optimization/98535
22951 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
22952 If the high and low permutes are the same, remove the high permutes
22953 from the working set and only continue with the low ones.
22955 2021-01-20 Jakub Jelinek <jakub@redhat.com>
22957 PR tree-optimization/98721
22958 * builtins.c (access_ref::inform_access): Don't assume
22959 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
22960 object whenever allocfn is NULL, rather than only when DECL_P
22961 is true. Use %qE instead of %qD for that. Formatting fixes.
22963 2021-01-20 Richard Biener <rguenther@suse.de>
22965 PR tree-optimization/98758
22966 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
22967 (lambda_matrix_right_hermite): Avoid undefinedness with
22968 signed integer abs and multiplication.
22969 (analyze_subscript_affine_affine): Use lambda_int.
22971 2021-01-20 David Malcolm <dmalcolm@redhat.com>
22974 * dwarf2out.c (output_line_info): Rename static variable
22975 "generation", moving it out of the function to...
22976 (output_line_info_generation): New.
22977 (init_sections_and_labels): Likewise, renaming the variable to...
22978 (init_sections_and_labels_generation): New.
22979 (dwarf2out_c_finalize): Reset the new variables.
22981 2021-01-19 Martin Sebor <msebor@redhat.com>
22983 PR middle-end/98664
22984 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
22985 all functions, even if they're not declared artificial or inline.
22986 * tree.c (tree_inlined_location): Use macro expansion location
22987 only if scope traversal fails to expose one.
22989 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
22991 PR rtl-optimization/92294
22992 * alias.c (compare_base_symbol_refs): Take an extra parameter
22993 and add the distance between two symbols to it. Enshrine in
22994 comments that -1 means "either 0 or 1, but we can't tell
22995 which at compile time".
22996 (memrefs_conflict_p): Update call accordingly.
22997 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23000 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23002 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23003 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23004 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23005 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23006 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23007 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23008 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23009 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23010 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23011 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23012 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23013 sqmovun): Use NONE flags.
23015 2021-01-19 Richard Biener <rguenther@suse.de>
23018 * ipa-modref.c (analyze_stmt): Only record a summary for a
23021 2021-01-19 Richard Biener <rguenther@suse.de>
23023 PR middle-end/98638
23024 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23026 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23028 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23029 built-in define __FIX_LEON3FT_TN0018.
23031 2021-01-19 Richard Biener <rguenther@suse.de>
23034 * tree-inline.c (tree_function_versioning): Set input_location
23035 to UNKNOWN_LOCATION throughout the function.
23037 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23040 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23042 2021-01-19 Martin Jambor <mjambor@suse.cz>
23045 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23046 whether non-call exceptions allow removal of a statement.
23047 (isra_analyze_call): Pass the appropriate function to
23048 ssa_name_only_returned_p.
23050 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23052 * config/riscv/arch-canonicalize (longext_sort): New function for
23053 sorting 'multi-letter'.
23054 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23055 'alts'. The 'arch' may not be the first of 'alts'.
23056 (_expand_combination): Add underline for the 'ext' without '*'.
23057 This is because, a single-letter extension can always be treated well
23058 with a '_' prefix, but it cannot be separated out if it is appended
23061 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23064 * ira.c (ira): Skip abnormal critical edge splitting.
23066 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23068 PR tree-optimization/98727
23069 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23070 second .MUL_OVERFLOW operand for signed multiplication with overflow
23071 checking if the second operand of multiplication is not constant.
23073 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23075 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23076 defaults to version 4.
23078 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23080 * attribs.h (fndecl_dealloc_argno): New decl.
23081 * builtins.c (call_dealloc_argno): Split out second half of
23083 (fndecl_dealloc_argno): New.
23084 * doc/extend.texi (Common Function Attributes): Document the
23085 interaction between the analyzer and the malloc attribute.
23086 * doc/invoke.texi (Static Analyzer Options): Likewise.
23088 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23090 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23091 dwarf_version to 4.
23092 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23094 2021-01-17 Martin Jambor <mjambor@suse.cz>
23097 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23100 2021-01-17 Mark Wielaard <mark@klomp.org>
23102 * common.opt (gdwarf-): Init(5).
23103 * doc/invoke.texi (-gdwarf): Document default to 5.
23105 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23107 * builtin-types.def
23108 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23110 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23111 ...this. Add extra argument.
23112 * gimplify.c (omp_default_clause): Ensure that event handle is
23113 firstprivate in a task region.
23114 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23115 (gimplify_adjust_omp_clauses): Likewise.
23116 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23117 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23118 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23119 if detach clause specified. Add detach argument when generating
23121 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23123 (finish_taskreg_scan): Move field for variable containing the event
23124 handle to the front of the struct.
23125 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23127 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23128 OMP_CLAUSE_DETACH clause.
23129 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23130 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23131 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23133 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23135 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23137 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23139 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23140 multilib directory names. Use MULTILIB_REQUIRED instead of
23141 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23142 -mcustom-fpu-cfg=fph2 multilib.
23144 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23146 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23147 (nios2_init_fpu_configs): Provide register values for new
23148 -mcustom-fpu-cfg=fph2 option variant.
23149 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23152 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23154 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23155 custom instruction warnings.
23157 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23159 PR tree-optimization/96669
23160 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23162 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23164 PR tree-optimization/96271
23165 * passes.def: Pass false argument to first two pass_cd_dce
23166 instances and true to last instance. Add comment that
23167 last instance rewrites no longer addressed locals.
23168 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23170 (pass_cd_dce::set_pass_param): New method.
23171 (pass_cd_dce::execute): Return TODO_update_address_taken from
23172 last cd_dce instance.
23174 2021-01-15 Carl Love <cel@us.ibm.com>
23176 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23178 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23179 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23180 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23181 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23182 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23183 Add builtin define.
23184 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23185 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23186 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23187 New overloaded definitions.
23188 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23189 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23190 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23191 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23192 P10V_BUILTIN_MULHU_V4SI]: Add case
23193 statement for builtins.
23194 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23195 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23196 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23197 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23198 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23199 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23200 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23201 Add define_insn, mode is VIlong.
23202 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23203 Add builtin descriptions.
23205 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23207 * final.c (final_start_function_1): Reset force_source_line.
23209 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23211 PR tree-optimization/96669
23212 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23213 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23214 possibly different power of two constants and to right shift too.
23216 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23218 PR tree-optimization/96681
23219 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23220 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23221 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23222 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23224 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23226 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23228 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23230 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23231 cmul<conj_op><mode>3): New.
23232 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23233 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23234 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23235 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23236 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23237 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23238 (rot_op): Renamed to conj_op.
23239 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23240 cmul<conj_op><mode>3): New.
23241 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23242 cmul<conj_op><mode>3): New.
23244 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23248 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23249 Escape the tempfile name when constructing the expected output.
23251 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23253 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23255 (aarch64_<su>mlsl_hi<mode>): ... This.
23256 (aarch64_<su>mlsl_hi<mode>): Define.
23257 (*aarch64_<su>mlsl<mode): Rename to...
23258 (aarch64_<su>mlsl<mode): ... This.
23259 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23260 smlsl_hi, umlsl_hi): Define builtins.
23261 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23262 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23263 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23264 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23266 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23268 * config/i386/i386-c.c (ix86_target_macros):
23269 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23271 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23274 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23275 * Makefile.in (RTL_SSA_H): New variable.
23276 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23277 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23278 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23280 * config/aarch64/aarch64-cc-fusion.cc: New file.
23282 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23284 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23285 calling cancel_changes for changes that no longer exist.
23287 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23289 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23290 (function_info::reg_defs): ...this.
23291 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23292 (function_info::reg_defs): ...this.
23294 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23297 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23299 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23302 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23305 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23307 2021-01-15 Richard Biener <rguenther@suse.de>
23309 PR tree-optimization/96376
23310 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23311 for VMAT_INVARIANT.
23313 2021-01-15 Martin Liska <mliska@suse.cz>
23315 * doc/install.texi: Document that some tests need pytest module.
23316 * doc/sourcebuild.texi: Likewise.
23318 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23321 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23323 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23325 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23326 (mve_vshrq_n_u<mode>_imm): Likewise.
23327 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23328 * config/arm/vec-common.md: ... here.
23330 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23332 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
23334 * config/arm/neon.md (vashl<mode>3): Delete.
23335 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
23336 (vasl<mode>3): New expander.
23338 2021-01-15 Richard Biener <rguenther@suse.de>
23340 PR tree-optimization/98685
23341 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
23342 of vector extern defs.
23344 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23347 * diagnostic.c (diagnostic_kind_text): Break out this array
23349 (diagnostic_build_prefix): ...here.
23350 (fancy_abort): Detect when diagnostic_initialize has not yet been
23351 called and fall back to a minimal implementation of printing the
23352 ICE, rather than segfaulting in internal_error.
23354 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23356 * diagnostic.c (diagnostic_initialize): Eliminate
23357 parseable_fixits_p in favor of initializing extra_output_kind from
23358 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23359 (convert_column_unit): New function, split out from...
23360 (diagnostic_converted_column): ...this.
23361 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
23362 Use them to call convert_column_unit on the column values.
23363 (diagnostic_report_diagnostic): Eliminate conditional on
23364 parseable_fixits_p in favor of a switch statement on
23365 extra_output_kind, passing the appropriate values to the new
23366 params of print_parseable_fixits.
23367 (selftest::test_print_parseable_fixits_none): Update for new
23368 params of print_parseable_fixits.
23369 (selftest::test_print_parseable_fixits_insert): Likewise.
23370 (selftest::test_print_parseable_fixits_remove): Likewise.
23371 (selftest::test_print_parseable_fixits_replace): Likewise.
23372 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23374 (selftest::diagnostic_c_tests): Call it.
23375 * diagnostic.h (enum diagnostics_extra_output_kind): New.
23376 (diagnostic_context::parseable_fixits_p): Delete field in favor
23378 (diagnostic_context::extra_output_kind): ...this new field.
23379 * doc/invoke.texi (Environment Variables): Add
23380 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23381 * opts.c (common_handle_option): Update handling of
23382 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
23385 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23387 * tree-vect-slp-patterns.c (class complex_operations_pattern,
23388 complex_operations_pattern::matches,
23389 complex_operations_pattern::recognize,
23390 complex_operations_pattern::build): New.
23391 (slp_patterns): Use it.
23393 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23395 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
23396 * optabs.def (cmls_optab, cmls_conj_optab): New.
23397 * doc/md.texi: Document them.
23398 * tree-vect-slp-patterns.c (class complex_fms_pattern,
23399 complex_fms_pattern::matches, complex_fms_pattern::recognize,
23400 complex_fms_pattern::build): New.
23402 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23404 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
23405 * optabs.def (cmla_optab, cmla_conj_optab): New.
23406 * doc/md.texi: Document them.
23407 * tree-vect-slp-patterns.c (vect_match_call_p,
23408 class complex_fma_pattern, vect_slp_reset_pattern,
23409 complex_fma_pattern::matches, complex_fma_pattern::recognize,
23410 complex_fma_pattern::build): New.
23412 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23414 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
23415 * optabs.def (cmul_optab, cmul_conj_optab): New.
23416 * doc/md.texi: Document them.
23417 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
23418 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
23419 vect_build_combine_node, class complex_mul_pattern,
23420 complex_mul_pattern::matches, complex_mul_pattern::recognize,
23421 complex_mul_pattern::build): New.
23423 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23425 * tree-vect-slp.c (optimize_load_redistribution_1): New.
23426 (optimize_load_redistribution, vect_is_slp_load_node): New.
23427 (vect_match_slp_patterns): Use it.
23429 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23431 * tree-vect-slp-patterns.c (complex_add_pattern::build):
23434 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
23436 * config/gcn/mkoffload.c (main): Create an offload image only in
23437 64-bit configurations.
23439 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
23442 * config/i386/i386-options.c (ix86_option_override_internal):
23443 Issue an error for -fcf-protection with CF_BRANCH when compiling
23444 for 32-bit non-TARGET_CMOV targets.
23446 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23449 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
23450 Remove declaration and initialization of shadow variable "ret".
23451 (ix86_option_override_internal): Remove delcaration of
23452 shadow variable "i". Redeclare shadowed variable to unsigned.
23453 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
23454 * config/i386/i386-builtins.c (get_builtin_code_for_version):
23455 Update for redeclaration.
23456 * config/i386/i386.h (pta_size): Ditto.
23458 2021-01-14 Richard Biener <rguenther@suse.de>
23460 PR tree-optimization/98674
23461 * tree-data-ref.c (base_supports_access_fn_components_p): New.
23462 (initialize_data_dependence_relation): For two bases without
23463 possible access fns resort to type size equality when determining
23464 shape compatibility.
23466 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23469 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
23470 <=, >= operators in vcle and vcge intrinsics respectively.
23471 * config/arm/arm_neon_builtins.def: Remove entry for
23474 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23477 * config/i386/i386-options.c (ix86_function_specific_save):
23478 Remove redundant assignment to opts->x_ix86_branch_cost.
23479 * config/i386/i386.c (ix86_prefetch_sse):
23480 Rename from x86_prefetch_sse. Update all uses.
23481 * config/i386/i386.h: Update for rename.
23482 * config/i386/i386-options.h: Ditto.
23484 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23487 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
23488 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
23489 Use Bm instead of m for non-avx. Add isa attribute.
23491 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23493 PR tree-optimization/96688
23494 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
23495 ~X can be simplified.
23497 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
23499 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
23500 IFN_LOAD_LANES results.
23502 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23504 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
23506 (aarch64_xtn<mode>): Likewise.
23507 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
23510 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
23512 (vmovl_s16): Likewise.
23513 (vmovl_s32): Likewise.
23514 (vmovl_u8): Likewise.
23515 (vmovl_u16): Likewise.
23516 (vmovl_u32): Likewise.
23517 (vmovn_s16): Likewise.
23518 (vmovn_s32): Likewise.
23519 (vmovn_s64): Likewise.
23520 (vmovn_u16): Likewise.
23521 (vmovn_u32): Likewise.
23522 (vmovn_u64): Likewise.
23524 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23526 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
23528 (aarch64_<su>qxtn2<mode>_be): Likewise.
23529 (aarch64_<su>qxtn2<mode>): Likewise.
23530 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
23532 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
23533 (su): Handle ss_truncate and us_truncate.
23534 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
23536 (vqmovn_high_s32): Likewise.
23537 (vqmovn_high_s64): Likewise.
23538 (vqmovn_high_u16): Likewise.
23539 (vqmovn_high_u32): Likewise.
23540 (vqmovn_high_u64): Likewise.
23542 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23544 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
23546 (aarch64_xtn2<mode>_be): Likewise.
23547 (aarch64_xtn2<mode>): Likewise.
23548 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
23550 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
23552 (vmovn_high_s32): Likewise.
23553 (vmovn_high_s64): Likewise.
23554 (vmovn_high_u16): Likewise.
23555 (vmovn_high_u32): Likewise.
23556 (vmovn_high_u64): Likewise.
23558 2021-01-13 Stafford Horne <shorne@gmail.com>
23560 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
23562 2021-01-13 Stafford Horne <shorne@gmail.com>
23564 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
23566 2021-01-13 Stafford Horne <shorne@gmail.com>
23568 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
23569 define for __or1k_hard_float__.
23571 2021-01-13 Stafford Horne <shorne@gmail.com>
23573 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
23574 (PROFILE_HOOK): Define to call _mcount.
23575 (FUNCTION_PROFILER): Change from abort to no-op.
23577 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23579 PR tree-optimization/96691
23580 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
23581 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
23582 (~D ^ C) or (D ^ C) can be simplified.
23584 2021-01-13 Richard Biener <rguenther@suse.de>
23586 PR tree-optimization/92645
23587 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
23588 until after vector lowering.
23590 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23592 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
23594 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
23595 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
23597 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23599 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
23601 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
23602 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
23604 2021-01-13 Richard Biener <rguenther@suse.de>
23606 PR tree-optimization/92645
23607 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
23608 BIT_FIELD_REF argument.
23609 (vect_build_slp_tree_2): Record the desired vector type
23610 on the external vector def.
23611 (vectorizable_slp_permutation): Handle required punning
23612 of existing vector defs.
23614 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23616 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
23618 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23620 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
23622 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
23624 * config.gcc [$target == *-*-gnu*]: Enable
23625 'default_gnu_indirect_function'.
23627 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23630 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
23631 registers before calling targetm.vectorize.vec_perm_const, only after
23633 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
23634 two argument permutation when one operand is zero vector and only
23635 after that force operands into registers.
23636 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
23637 define_insn_and_split pattern.
23638 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
23639 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
23640 (*avx2_zero_extendv8hiv8si2_1): Likewise.
23641 (*avx512f_zero_extendv8siv8di2_1): Likewise.
23642 (*avx2_zero_extendv4siv4di2_1): Likewise.
23643 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
23645 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
23646 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
23647 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
23648 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
23649 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
23650 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
23652 2021-01-13 Martin Liska <mliska@suse.cz>
23654 PR tree-optimization/98455
23655 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
23656 Record also virtual PHIs.
23657 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
23660 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
23662 * doc/invoke.texi (C++ Modules): Fix typos.
23664 2021-01-13 Richard Biener <rguenther@suse.de>
23666 PR tree-optimization/98640
23667 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
23668 handle plus or minus from a truncated operand to be
23671 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23674 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
23675 define_insn_and_split patterns.
23676 (splitter after *btr<mode>_2): New splitter.
23678 2021-01-13 Martin Liska <mliska@suse.cz>
23681 * cgraphunit.c (analyze_functions): Remove dead code.
23683 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
23685 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
23686 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
23687 (a64fx_regmove_cost, a64fx_vector_cost): New.
23688 (a64fx_tunings): Use the new added cost tables.
23690 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23693 * config/i386/predicates.md (pmovzx_parallel): New predicate.
23694 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
23695 define_insn_and_split pattern.
23696 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
23697 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
23699 2021-01-13 Julian Brown <julian@codesourcery.com>
23701 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
23702 to fix v0 register.
23704 2021-01-13 Julian Brown <julian@codesourcery.com>
23706 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
23709 2021-01-13 Julian Brown <julian@codesourcery.com>
23711 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
23712 for reciprocal-approximation instructions.
23713 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
23714 refinement and division result.
23715 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
23717 2021-01-13 Julian Brown <julian@codesourcery.com>
23719 * config/gcn/gcn-valu.md (subdf): Rename to...
23722 2021-01-12 Martin Liska <mliska@suse.cz>
23724 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
23726 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
23728 * function-abi.h: Fix typo.
23730 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23734 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
23735 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
23736 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
23737 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
23738 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
23739 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
23740 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
23741 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
23742 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
23743 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
23744 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
23745 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
23746 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
23747 (ARM_HAVE_V2DI_LDST): Likewise.
23748 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
23749 (*movmisalign<mode>_mve_load): New pattern.
23750 * config/arm/neon.md (movmisalign<mode>): Move to ...
23751 * config/arm/vec-common.md: ... here.
23753 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
23756 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
23757 of pattern 'plus (plus (hard reg, const), pseudo)'.
23759 2021-01-12 Richard Biener <rguenther@suse.de>
23761 PR tree-optimization/98550
23762 * tree-vect-slp.c (vect_record_max_nunits): Check whether
23763 the group size is a multiple of the vector element count.
23764 (vect_build_slp_tree_1): When we need to fail because
23765 the vector type choosen causes unrolling do so lazily
23766 without affecting matches only at the end to guide group splitting.
23768 2021-01-12 Martin Liska <mliska@suse.cz>
23771 * optc-save-gen.awk: Compare also n_target_save vars with
23774 2021-01-12 Martin Liska <mliska@suse.cz>
23776 * gcov.c (source_info::debug): New.
23777 (print_usage): Add --debug (-D) option.
23778 (process_args): Likewise.
23779 (generate_results): Call src->debug after
23780 accumulate_line_counts.
23781 (read_graph_file): Properly assign id for EXIT_BLOCK.
23782 * profile.c (branch_prob): Dump function body before it is
23785 2021-01-12 Jakub Jelinek <jakub@redhat.com>
23787 PR tree-optimization/98629
23788 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
23789 unless returning non-zero.
23791 2021-01-12 Jakub Jelinek <jakub@redhat.com>
23793 PR tree-optimization/95731
23794 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
23795 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
23796 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
23797 only after optimize_range_tests_var_bound.
23799 2021-01-12 Jakub Jelinek <jakub@redhat.com>
23801 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
23802 * configure: Regenerated.
23804 2021-01-12 liuhongt <hongtao.liu@intel.com>
23807 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
23809 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
23812 2021-01-12 Alexandre Oliva <oliva@adacore.com>
23814 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
23816 (auto_end_imm_use_stmt_traverse): New struct.
23817 (FOR_EACH_IMM_USE_STMT): Use it.
23818 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
23820 * gimple-ssa-strength-reduction.c: ... here, ...
23821 * graphite-scop-detection.c: ... here, ...
23822 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
23823 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
23824 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
23825 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
23826 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
23827 * tree-vect-slp.c: ... and here, ...
23828 * doc/tree-ssa.texi: ... and the example here.
23830 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23832 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
23833 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
23834 (*sdiv_pow2<mode>3): New pattern.
23835 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
23836 Wrap the ASRD in an UNSPEC_PRED_X.
23837 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
23838 predicate with a constant PTRUE, if it isn't already.
23839 (*cond_<sve_int_op><mode>_z): Replace with...
23840 (*cond_<sve_int_op><mode>_any): ...this new pattern.
23842 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23844 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
23845 SVE_FULL_I to SVE_I.
23846 (*cond_bic<mode>_any): Likewise.
23848 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23850 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
23851 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
23854 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23856 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
23857 SVE_FULL_I to SVE_I.
23858 (*aarch64_cond_<su>abd<mode>_2): Likewise.
23859 (*aarch64_cond_<su>abd<mode>_any): Likewise.
23860 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
23861 for the max and min but not for the minus.
23862 (*aarch64_cond_<su>abd<mode>_3): New pattern.
23864 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23866 * config/aarch64/iterators.md (SVE_24I): New iterator.
23867 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
23868 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
23870 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23872 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
23873 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
23875 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
23876 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
23877 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
23878 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
23880 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23882 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
23883 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
23884 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
23887 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23889 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
23890 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
23891 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
23893 2021-01-11 Martin Liska <mliska@suse.cz>
23896 * symtab-clones.h (clone_info::release): Release
23897 symtab::m_clones with ggc_delete as it's a GGC memory.
23899 2021-01-11 Matthias Klose <doko@ubuntu.com>
23901 * Makefile.in (LINK_PROGRESS): Show the link target.
23903 2021-01-11 Richard Biener <rguenther@suse.de>
23905 PR tree-optimization/91403
23906 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
23907 single-element interleaving group size at 4096 elements.
23909 2021-01-11 Richard Biener <rguenther@suse.de>
23911 PR tree-optimization/98526
23912 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
23913 of the actual reduction op for the regular case.
23914 (vectorizable_reduction): Cost the stmts
23915 vect_transform_reduction produces here.
23917 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
23919 * tree-ssa-forwprop.c (simplify_vector_constructor): For
23920 big-endian, use UNPACK[_FLOAT]_HI.
23922 2021-01-11 Tamar Christina <tamar.christina@arm.com>
23924 * tree-vect-slp-patterns.c (class complex_pattern,
23925 class complex_add_pattern): Add parameters to matches.
23926 (complex_add_pattern::build): Free memory.
23927 (complex_add_pattern::matches): Move validation end of match.
23928 (complex_add_pattern::recognize): Likewise.
23930 2021-01-11 Tamar Christina <tamar.christina@arm.com>
23932 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
23934 2021-01-11 Tamar Christina <tamar.christina@arm.com>
23936 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
23938 2021-01-11 Jakub Jelinek <jakub@redhat.com>
23940 PR tree-optimization/95867
23941 * tree-ssa-math-opts.h: New header.
23942 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
23943 (powi_as_mults): No longer static. Use build_one_cst instead of
23944 build_real. Formatting fix.
23945 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
23946 (attempt_builtin_powi): Handle multiplication reassociation without
23947 powi_fndecl using powi_as_mults.
23948 (reassociate_bb): For integral types don't require
23949 -funsafe-math-optimizations to call attempt_builtin_powi.
23951 2021-01-11 Jakub Jelinek <jakub@redhat.com>
23953 PR tree-optimization/95852
23954 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
23955 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
23956 allow in the bb any of the stmts in that vector, div_stmt and
23957 up to 3 cast stmts.
23958 (arith_cast_equal_p): New function.
23959 (arith_overflow_check_p): Add cast_stmt argument, handle signed
23960 multiply overflow checks.
23961 (match_arith_overflow): Adjust caller. Handle signed multiply
23964 2021-01-11 Jakub Jelinek <jakub@redhat.com>
23966 PR tree-optimization/95852
23967 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
23968 (uaddsub_overflow_check_p): Renamed to ...
23969 (arith_overflow_check_p): ... this. Handle also multiplication
23970 with overflow check.
23971 (match_uaddsub_overflow): Renamed to ...
23972 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
23973 also multiplication with overflow check. Adjust function comment.
23974 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
23975 match_arith_overflow also for MULT_EXPR.
23977 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23979 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
23980 __builtin_convertvector.
23981 (vmovl_s16): Likewise.
23982 (vmovl_s32): Likewise.
23983 (vmovl_u8): Likewise.
23984 (vmovl_u16): Likewise.
23985 (vmovl_u32): Likewise.
23986 (vmovn_s16): Likewise.
23987 (vmovn_s32): Likewise.
23988 (vmovn_s64): Likewise.
23989 (vmovn_u16): Likewise.
23990 (vmovn_u32): Likewise.
23991 (vmovn_u64): Likewise.
23993 2021-01-11 Martin Liska <mliska@suse.cz>
23995 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
23996 (if_chain::is_beneficial): Delete clusters
23997 (find_conditions): Make second argument of conditions_in_bbs a
23998 pointer so that we control over it's lifetime.
23999 (pass_if_to_switch::execute): Delete them.
24001 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24003 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24006 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24008 * config/vax/vax.md (cc): Remove mode attribute.
24009 (subst_<cc>, subst_f<cc>): Rename to...
24010 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24011 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24012 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24013 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24015 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24017 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24018 `const_double_zero'.
24020 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24022 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24023 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24026 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24028 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24030 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24031 with `const_double_zero'.
24032 * doc/rtl.texi (Constant Expression Types): Document it.
24034 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24037 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24038 POINTER_DIFF_EXPR to be any integral type.
24040 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24042 PR rtl-optimization/98603
24043 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24044 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24045 if any, set ASM_OPERANDS mode to VOIDmode and change
24046 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24048 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24051 * final.c (notice_source_line): Narrow down the condition to
24052 skip a line-0 marker.
24054 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24056 * ipa-modref.c (merge_call_side_effects): Fix
24057 linebreak split by reordering two print calls.
24059 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24061 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24062 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24063 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24064 consistency, fix constraint.
24066 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24068 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24069 callables instead of mask values.
24070 (struct target_flag_set_p): New predicate.
24071 (s390_cpu_cpp_builtins_internal): Define or undefine
24072 __LONG_DOUBLE_VX__ macro.
24074 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24077 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24078 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24081 2021-01-08 Richard Biener <rguenther@suse.de>
24083 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24085 2021-01-08 Richard Biener <rguenther@suse.de>
24087 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24088 (vect_build_slp_tree): On cache hit release the matched
24089 scalar stmts vector.
24090 * tree-vect-stmts.c (vectorizable_store): Properly free
24091 vec_oprnds before possibly gathering them again.
24093 2021-01-08 Richard Biener <rguenther@suse.de>
24095 PR tree-optimization/98544
24096 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24097 permutes at a permute node.
24099 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24102 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24103 mcount in large model. Sorry for large model with PIC.
24105 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24108 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24109 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24110 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24111 TargetSave and initialize for variables with enum types.
24112 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24113 mstack-protector-guard-symbol=): Add Save.
24114 * config/i386/i386-options.c (ix86_function_specific_save,
24115 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24116 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24117 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24118 x_ix86_veclibabi_type.
24120 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24122 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24123 SVE_FULL_I to SVE_I.
24124 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24126 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24128 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24129 SVE_FULL_I to SVE_I.
24130 (*cond_uxt<mode>_any): Likewise.
24132 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24134 * config/aarch64/iterators.md (Vwhalf): New iterator.
24135 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24137 (aarch64_<sur>adalp<mode>): ... This. Make more
24139 (<sur>sadv16qi): Adjust callsite of the above.
24140 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24142 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24144 (vpadal_s16): Likewise.
24145 (vpadal_u8): Likewise.
24146 (vpadal_u16): Likewise.
24147 (vpadalq_s8): Likewise.
24148 (vpadalq_s16): Likewise.
24149 (vpadalq_s32): Likewise.
24150 (vpadalq_u8): Likewise.
24151 (vpadalq_u16): Likewise.
24152 (vpadalq_u32): Likewise.
24154 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24156 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24158 (aarch64_<su>abd<mode>): ... This.
24159 (<sur>sadv16qi): Adjust callsite of the above.
24160 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24162 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24164 (vabd_s16): Likewise.
24165 (vabd_s32): Likewise.
24166 (vabd_u8): Likewise.
24167 (vabd_u16): Likewise.
24168 (vabd_u32): Likewise.
24169 (vabdq_s8): Likewise.
24170 (vabdq_s16): Likewise.
24171 (vabdq_s32): Likewise.
24172 (vabdq_u8): Likewise.
24173 (vabdq_u16): Likewise.
24174 (vabdq_u32): Likewise.
24176 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24178 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24180 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24181 (vaba_s16): Likewise.
24182 (vaba_s32): Likewise.
24183 (vaba_u8): Likewise.
24184 (vaba_u16): Likewise.
24185 (vaba_u32): Likewise.
24186 (vabaq_s8): Likewise.
24187 (vabaq_s16): Likewise.
24188 (vabaq_s32): Likewise.
24189 (vabaq_u8): Likewise.
24190 (vabaq_u16): Likewise.
24191 (vabaq_u32): Likewise.
24193 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24195 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24196 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24197 Change RTL pattern to match.
24199 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24201 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24202 * config/riscv/riscv-c.c (riscv-subset.h): New.
24203 (INCLUDE_STRING): Define.
24204 (riscv_cpu_cpp_builtins): Add new style architecture extension
24206 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24207 (riscv_subset_list::end): New.
24208 (riscv_current_subset_list): New.
24210 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24212 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24213 Move to riscv-subset.h.
24214 (struct riscv_subset_t): Ditto.
24215 (class riscv_subset_list): Ditto.
24216 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24217 from riscv-common.c.
24218 (struct riscv_subset_t): Ditto.
24219 (class riscv_subset_list): Ditto.
24220 * config/riscv/t-riscv ($(common_out_file)): Add file
24223 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24226 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24227 New define_insn patterns.
24229 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24231 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24232 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24233 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24235 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24237 PR tree-optimization/98560
24238 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24239 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24241 (vec_cond_direct): Likewise argument 2.
24242 (vec_condu_direct, vec_condeq_direct): Delete.
24243 (expand_vect_cond_optab_fn): Rename to...
24244 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24245 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24246 (expand_vect_cond_mask_optab_fn): Rename to...
24247 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24248 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24250 (direct_vec_cond_optab_supported_p): Likewise.
24251 (direct_vec_condu_optab_supported_p): Delete.
24252 (direct_vec_condeq_optab_supported_p): Delete.
24253 * gimple-isel.cc: Include internal-fn.h.
24254 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24257 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24259 PR tree-optimization/98560
24260 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24261 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24263 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24265 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24266 rotate_insn and optab code attributes.
24267 Update all uses to merged code attribute.
24268 * config/i386/sse.md: Update all uses to merged code attribute.
24269 * config/i386/mmx.md: Update all uses to merged code attribute.
24271 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24273 PR tree-optimization/98568
24274 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24275 (bswap_replace): Use it.
24277 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24279 PR rtl-optimization/97978
24280 * lra-int.h (lra_hard_reg_split_p): New external.
24281 * lra.c (lra_hard_reg_split_p): New global.
24282 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24283 * lra-assigns.c (lra_assign): Don't check allocation correctness
24284 after hard reg splitting.
24286 2021-01-06 Martin Sebor <msebor@redhat.com>
24289 * builtins.c (new_delete_mismatch_p): New overload.
24290 (new_delete_mismatch_p (tree, tree)): Call it.
24292 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24294 * Makefile.in (T_GLIMITS_H): New.
24295 (stmp-int-hdrs): Depend on it, use it.
24296 * config/t-vxworks (T_GLIMITS_H): Override it.
24297 (vxw-glimits.h): New.
24299 2021-01-06 Richard Biener <rguenther@suse.de>
24301 PR tree-optimization/98513
24302 * value-range.cc (intersect_ranges): Compare the upper bounds
24303 for the expected relation.
24305 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24308 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24310 * doc/standards.texi (HSAIL): Remove section.
24312 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24314 * configure: Re-generate.
24316 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24318 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24319 ISO 14882:2020 standard.
24320 * doc/standards.texi: Likewise.
24322 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24324 PR tree-optimization/94802
24325 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24326 * expr.c: Include tree-pretty-print.h and flags.h.
24327 (maybe_optimize_sub_cmp_0): New function.
24328 (do_store_flag): Use it.
24329 * cfgexpand.c (expand_gimple_cond): Likewise.
24331 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24333 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
24334 * rtlanal.c (simple_regno_set): Tweak description to clarify the
24337 2021-01-05 Richard Biener <rguenther@suse.de>
24339 PR tree-optimization/98516
24340 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
24341 lanes when materializing on a VEC_PERM node.
24342 (vectorizable_slp_permutation): Dump the permute properly.
24344 2021-01-05 Richard Biener <rguenther@suse.de>
24346 * tree-vect-slp.c (vect_slp_region): Move debug counter
24347 to cover individual subgraphs.
24349 2021-01-05 Richard Biener <rguenther@suse.de>
24351 PR tree-optimization/98428
24352 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
24353 vector lane extracts for loop vectorization.
24355 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24357 PR tree-optimization/98514
24358 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
24360 (operand_rank): Change type from hash_map<tree, long> to
24361 hash_map<tree, int64_t>.
24362 (phi_rank): Change return type from long to int64_t.
24363 (loop_carried_phi): Change block_rank variable type from long to
24365 (propagate_rank): Change return type, rank parameter type and
24366 op_rank variable type from long to int64_t.
24367 (find_operand_rank): Change return type from long to int64_t
24368 and change slot variable type from long * to int64_t *.
24369 (insert_operand_rank): Change rank parameter type from long to
24371 (get_rank): Change return type and rank variable type from long to
24372 int64_t. Use PRId64 instead of ld to print the rank.
24373 (init_reassoc): Change rank variable type from long to int64_t
24374 and adjust correspondingly bb_rank and operand_rank initialization.
24376 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24378 PR tree-optimization/96928
24379 * tree-ssa-phiopt.c (xor_replacement): New function.
24380 (tree_ssa_phiopt_worker): Call it.
24382 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24384 PR tree-optimization/96930
24385 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
24386 from narrower value which has the same type as 1 << B, perform
24387 the right shift on the narrower value followed by extension.
24389 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24391 PR tree-optimization/96239
24392 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
24394 (get_status_for_store_merging): Don't return BB_INVALID for blocks
24395 with potential bswap optimizable CONSTRUCTORs.
24396 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
24399 2021-01-05 Richard Biener <rguenther@suse.de>
24401 PR tree-optimization/98381
24402 * tree.c (vector_element_bits): Properly compute bool vector
24404 * tree-vect-loop.c (vectorizable_live_operation): Properly
24405 compute the last lane bit offset.
24407 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24410 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
24411 Clear the top 64 bytes of the input XMM register.
24412 (sse_cvttps2pi): Ditto.
24414 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24417 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
24419 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
24422 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
24425 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
24427 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
24428 (umaddsidi4_split): Likewise.
24430 2021-01-05 liuhongt <hongtao.liu@intel.com>
24433 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
24434 define_insn_and_split for zero_extend of subreg HI of pmovskb
24436 (*sse2_pmovskb_zexthisi): Add new combine splitters for
24437 zero_extend of not of subreg HI of pmovskb result.
24439 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24442 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
24444 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
24445 convert_memory_address to convert symbolic immediates to ptr_mode
24446 before forcing them to memory.
24448 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24450 PR rtl-optimization/97144
24451 * recog.c (constrain_operands): Initialize matching_operand
24452 for each alternative, rather than only doing it once.
24454 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24456 PR rtl-optimization/98403
24457 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
24458 why we don't remove call clobbers.
24459 (function_info::apply_changes_to_insn): Don't attempt to add
24460 call clobbers here.
24462 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24464 PR tree-optimization/98371
24465 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
24466 (vect_analyze_loop): If an epilogue loop appears to be cheaper
24467 than the main loop, re-analyze it as a main loop before adopting
24470 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24473 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
24474 * aclocal.m4, configure: Regenerate.
24475 * Makefile.in (NETLIBS): Define.
24476 (BACKEND): Remove $(CODYLIB).
24478 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24480 PR rtl-optimization/98334
24481 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
24482 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
24484 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
24486 * tree-inline.c (expand_call_inline): Restore input_location.
24487 Return result from recursive call.
24489 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24491 PR tree-optimization/95401
24492 * config/aarch64/aarch64-sve-builtins.cc
24493 (gimple_folder::load_store_cookie): Use bits rather than bytes
24494 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
24495 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
24496 * tree-vect-stmts.c (vectorizable_store): Likewise.
24497 (vectorizable_load): Likewise.
24499 2021-01-04 Richard Biener <rguenther@suse.de>
24501 PR tree-optimization/98308
24502 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
24505 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24507 PR tree-optimization/95771
24508 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
24509 with precision smaller than int's precision and types with precision
24510 twice as large as long long. Formatting fixes.
24512 2021-01-04 Richard Biener <rguenther@suse.de>
24514 PR tree-optimization/98464
24515 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
24516 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
24517 (process_bb): Adjust.
24519 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
24522 * doc/invoke.texi (-fsanitize=address): Fix wording describing
24523 clash with -fsanitize=hwaddress.
24525 2021-01-04 Richard Biener <rguenther@suse.de>
24527 PR tree-optimization/98282
24528 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
24529 invariants as VN_NARY.
24531 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24534 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
24535 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
24536 to handle zero operands.
24538 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24540 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
24541 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
24542 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
24545 2021-01-04 Richard Biener <rguenther@suse.de>
24547 PR tree-optimization/98393
24548 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
24549 when hitting the limit.
24551 2021-01-04 Richard Biener <rguenther@suse.de>
24553 PR tree-optimization/98291
24554 * tree-vect-loop.c (vectorizable_reduction): Bypass
24555 associativity check for SLP reductions with VF 1.
24557 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24559 PR tree-optimization/96782
24560 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
24562 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
24564 * collect-utils.c (collect_execute): Check dumppfx.
24565 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
24566 to collect_execute.
24567 (do_link): Add new parameter atsuffix.
24568 (main): Handle -dumpdir option. Skip one argument for
24569 -o, -isystem and -B options.
24570 * gcc.c (make_at_file): New helper function.
24571 (close_at_file): Use it.
24573 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24575 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
24576 Amend handling for LD64_VERSION fallback defaults.
24578 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24580 * config.gcc: Compute default version information
24581 from the configured target. Likewise defaults for
24583 * config/darwin10.h: Removed.
24584 * config/darwin12.h: Removed.
24585 * config/darwin9.h: Removed.
24586 * config/rs6000/darwin8.h: Removed.
24588 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24590 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
24592 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24594 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
24595 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
24597 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24599 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
24601 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
24603 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24605 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
24606 for the Darwin10 unwinder stub from here ...
24607 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
24609 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24611 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
24612 (ASM_DEBUG_SPEC):Only define if the assembler supports
24614 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
24615 (DARWIN_PREFER_DWARF): Define.
24616 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
24617 (DARWIN_PREFER_DWARF): Likewise
24618 (DSYMUTIL_SPEC): Likewise.
24619 (COLLECT_RUN_DSYMUTIL): Likewise.
24620 (ASM_DEBUG_SPEC): Likewise.
24621 (ASM_DEBUG_OPTION_SPEC): Likewise.
24623 2021-01-02 Jan Hubicka <jh@suse.cz>
24625 * cfg.c (free_block): ggc_free bb.
24627 2021-01-01 Jakub Jelinek <jakub@redhat.com>
24629 * gcc.c (process_command): Update copyright notice dates.
24630 * gcov-dump.c (print_version): Ditto.
24631 * gcov.c (print_version): Ditto.
24632 * gcov-tool.c (print_version): Ditto.
24633 * gengtype.c (create_file): Ditto.
24634 * doc/cpp.texi: Bump @copying's copyright year.
24635 * doc/cppinternals.texi: Ditto.
24636 * doc/gcc.texi: Ditto.
24637 * doc/gccint.texi: Ditto.
24638 * doc/gcov.texi: Ditto.
24639 * doc/install.texi: Ditto.
24640 * doc/invoke.texi: Ditto.
24642 2021-01-01 Jakub Jelinek <jakub@redhat.com>
24644 * ChangeLog-2020: Rotate ChangeLog. New file.
24647 Copyright (C) 2021 Free Software Foundation, Inc.
24649 Copying and distribution of this file, with or without modification,
24650 are permitted in any medium without royalty provided the copyright
24651 notice and this notice are preserved.