1 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
3 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
4 * config/arm/arm-tables.opt: Regenerate.
5 * config/arm/arm-tune.md: Regenerate.
6 * doc/invoke.texi: Update docs.
8 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
12 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
15 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
17 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
18 omp_{c,re}alloc, fix omp_alloc/omp_free.
20 2021-09-30 Martin Liska <mliska@suse.cz>
22 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
23 asm_out_file and stream directly to MYFILE.
25 2021-09-30 Richard Biener <rguenther@suse.de>
27 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
28 Restore and fix condition under which we apply npeel to
29 the DRs misalignment value.
31 2021-09-30 Richard Biener <rguenther@suse.de>
33 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
34 Fix npeel check for variable amount of peeling.
36 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
38 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
40 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
42 * gimple-range.cc (gimple_ranger::debug): New.
43 * gimple-range.h (class gimple_ranger): Add debug.
45 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
48 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
50 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
53 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
54 (btf_finalize): Empty the hash_map btf_var_ids.
56 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
58 * tree-vrp.c (thread_through_all_blocks): Return bool.
59 (execute_vrp_threader): Return TODO_* flags.
60 (pass_data_vrp_threader): Set todo_flags_finish to 0.
62 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
64 * timevar.def (TV_TREE_VRP_THREADER): New.
65 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
67 2021-09-29 David Faust <david.faust@oracle.com>
69 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
71 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
73 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
74 new type qualifier enum.
75 (TYPES_TERNOP_SSSU): Likewise.
76 (TYPES_TERNOP_PPPU): Likewise.
77 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
78 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
80 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
81 builtin and remove casts.
82 (vqtbl1_s8): Likewise.
83 (vqtbl1q_p8): Likewise.
84 (vqtbl1q_s8): Likewise.
85 (vqtbx1_s8): Likewise.
86 (vqtbx1_p8): Likewise.
87 (vqtbx1q_s8): Likewise.
88 (vqtbx1q_p8): Likewise.
93 2021-09-29 Richard Biener <rguenther@suse.de>
95 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
97 (vect_update_misalignment_for_peel): Use it to update
98 misaligned to the value necessary for an aligned access.
99 (vect_get_peeling_costs_all_drs): Likewise.
100 (vect_enhance_data_refs_alignment): Likewise.
102 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
104 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
105 emitted operations and adjust heuristic for code size.
107 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
109 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
110 emitted operations and adjust heuristic for code size.
112 2021-09-29 Jakub Jelinek <jakub@redhat.com>
115 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
116 in OMP_SCOPE clauses, not just on worksharing construct clauses.
118 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
120 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
121 (<u>mulditi3): Call <su>muldi3_highpart.
122 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
123 (<u>mulsidi3): Call <su>mulsi3_highpart.
124 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
126 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
128 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
130 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
132 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
133 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
136 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
138 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
139 (AARCH64_FL_V8_7): Likewise.
140 (AARCH64_FL_FOR_ARCH8_7): Likewise.
141 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
142 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
143 * doc/invoke.texi: Document the above.
145 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
147 * dbgcnt.c (dbg_cnt_counter): New.
148 * dbgcnt.h (dbg_cnt_counter): New.
149 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
150 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
151 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
152 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
155 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
157 * cfgcleanup.c (pass_jump::execute): Check
158 flag_expensive_optimizations.
159 (pass_jump_after_combine::gate): Same.
160 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
161 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
162 * tree-ssa-threadupdate.c
163 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
166 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
168 * tree-ssa-reassoc.c (biased_names): New global.
169 (propagate_bias_p): New function.
170 (loop_carried_phi): Remove.
171 (propagate_rank): Propagate bias along single uses.
172 (get_rank): Update biased_names when needed.
174 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
176 * passes.def (pass_reassoc): Rename parameter to early_p.
177 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
179 (phi_rank): Don't bias loop-carried phi ranks
180 before vectorization pass.
181 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
182 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
184 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
186 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
188 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
190 2021-09-28 Jakub Jelinek <jakub@redhat.com>
193 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
194 special 80387 instruction XFmode constants if flag_rounding_math.
196 2021-09-28 Richard Biener <rguenther@suse.de>
198 PR tree-optimization/100112
199 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
200 referece into the hashtable twice in case last_vuse is
201 different from the original vuse on the stmt.
203 2021-09-28 Jakub Jelinek <jakub@redhat.com>
206 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
207 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
208 clauses on SIMD constructs.
210 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
212 PR tree-optimization/102511
213 * gimple-range-path.cc (path_range_query::range_on_path_entry):
214 Return VARYING when nothing found.
216 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
219 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
221 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
222 (VALID_MMX_REG_MODE): Add V4HFmode.
223 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
224 vector mode condition.
225 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
227 (function_arg_32): Add V4HFmode.
228 (function_arg_advance_32): Likewise.
229 * config/i386/i386.md (mode): Add V4HF/V2HF.
230 (MODE_SIZE): Likewise.
231 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
232 (V_32): Add V2HF mode.
233 (VHF_32_64): New mode iterator.
234 (*mov<mode>_internal): Adjust sse alternatives to support
236 (*mov<mode>_internal): Adjust sse alternatives to support
238 (<insn><mode>3): New define_insn for add/sub/mul/div.
240 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
242 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
244 (pass_early_thread_jumps::gate): Same.
245 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
246 Return if !flag_thread_jumps.
247 * tree-ssa-threadupdate.c
248 (jt_path_registry::register_jump_thread): Assert that
249 flag_thread_jumps is true.
251 2021-09-28 liuhongt <hongtao.liu@intel.com>
254 (simplify_context::simplify_binary_operation_1): Relax
255 condition of simplifying (vec_concat:M (vec_select op0
256 index0)(vec_select op1 index1)) to allow different modes
257 between op0 and M, but have same inner mode.
259 2021-09-28 liuhongt <hongtao.liu@intel.com>
261 * config/i386/i386-expand.c (emit_reduc_half): Handle
262 V8HF/V16HF/V32HFmode.
263 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
264 (REDUC_SSE_SMINMAX_MODE): Ditto.
265 (REDUC_PLUS_MODE): Add V16HF and V32HF.
266 (REDUC_SMINMAX_MODE): Ditto.
268 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
270 * gimple-range-path.cc
271 (path_range_query::precompute_ranges_in_block): Rename to...
272 (path_range_query::compute_ranges_in_block): ...this.
273 (path_range_query::precompute_ranges): Rename to...
274 (path_range_query::compute_ranges): ...this.
275 (path_range_query::precompute_relations): Rename to...
276 (path_range_query::compute_relations): ...this.
277 (path_range_query::precompute_phi_relations): Rename to...
278 (path_range_query::compute_phi_relations): ...this.
279 * gimple-range-path.h: Rename precompute* to compute*.
280 * tree-ssa-threadbackward.c
281 (back_threader::find_taken_edge_switch): Same.
282 (back_threader::find_taken_edge_cond): Same.
283 * tree-ssa-threadedge.c
284 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
285 (hybrid_jt_state::register_equivs_stmt): Inline...
286 * tree-ssa-threadedge.h: ...here.
288 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
290 * tree-vrp.c (lhs_of_dominating_assert): Remove.
291 (class vrp_jt_state): Remove.
292 (class vrp_jt_simplifier): Remove.
293 (vrp_jt_simplifier::simplify): Remove.
294 (class vrp_jump_threader): Remove.
295 (vrp_jump_threader::vrp_jump_threader): Remove.
296 (vrp_jump_threader::~vrp_jump_threader): Remove.
297 (vrp_jump_threader::before_dom_children): Remove.
298 (vrp_jump_threader::after_dom_children): Remove.
300 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
302 * passes.def (pass_vrp_threader): New.
303 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
304 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
305 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
306 (hybrid_jt_simplifier::simplify): New.
307 (hybrid_jt_simplifier::compute_ranges_from_state): New.
308 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
309 (class hybrid_jt_simplifier): New.
310 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
312 (class hybrid_threader): New.
313 (hybrid_threader::hybrid_threader): New.
314 (hybrid_threader::~hybrid_threader): New.
315 (hybrid_threader::before_dom_children): New.
316 (hybrid_threader::after_dom_children): New.
317 (execute_vrp_threader): New.
318 (class pass_vrp_threader): New.
319 (make_pass_vrp_threader): New.
321 2021-09-27 Martin Liska <mliska@suse.cz>
323 * output.h (enum section_flag): New.
324 (SECTION_FORGET): Remove.
325 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
326 (SECTION_STYLE_MASK): Define it based on other enum
328 * varasm.c (switch_to_section): Remove unused handling of
331 2021-09-27 Martin Liska <mliska@suse.cz>
333 * common.opt: Add new variable flag_default_complex_method.
334 * opts.c (finish_options): Handle flags related to
335 x_flag_complex_method.
336 * toplev.c (process_options): Remove option handling related
337 to flag_complex_method.
339 2021-09-27 Richard Biener <rguenther@suse.de>
342 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
343 type_for_size, instead use int_mode_for_size.
345 2021-09-27 Andrew Pinski <apinski@marvell.com>
348 * gimplify.c (gimplify_save_expr): Return early
349 if the type of val is error_mark_node.
351 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
353 * tree-ssanames.c (ssa_name_has_boolean_range): Use
356 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
358 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
360 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
361 (cprop_into_stmt): Same.
362 (dom_opt_dom_walker::optimize_stmt): Same.
364 2021-09-27 Richard Biener <rguenther@suse.de>
366 PR tree-optimization/97351
367 PR tree-optimization/97352
368 PR tree-optimization/82426
369 * tree-vectorizer.h (dr_misalignment): Add vector type
371 (aligned_access_p): Likewise.
372 (known_alignment_for_access_p): Likewise.
373 (vect_supportable_dr_alignment): Likewise.
374 (vect_known_alignment_in_bytes): Likewise. Refactor.
375 (DR_MISALIGNMENT): Remove.
376 (vect_update_shared_vectype): Likewise.
377 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
378 a vector type with larger alignment requirement and apply
379 the negative step adjustment here.
380 (vect_calculate_target_alignment): Remove.
381 (vect_compute_data_ref_alignment): Get explicit vector type
382 argument, do not apply a negative step alignment adjustment
384 (vect_slp_analyze_node_alignment): Re-analyze alignment
385 when we re-visit the DR with a bigger desired alignment but
386 keep more precise results from smaller alignments.
387 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
388 (vect_slp_analyze_node_operations_1): Do not update the
389 shared vector type on stmts.
390 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
391 vector type of an SLP node to the representative stmt-info.
392 (vect_transform_stmt): Likewise.
394 2021-09-27 liuhongt <hongtao.liu@intel.com>
397 2021-09-09 liuhongt <hongtao.liu@intel.com>
400 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
401 (reduc_plus_scal_v4sf): .. this, New define_expand.
402 (reduc_plus_scal_v2df): .. and this, New define_expand.
404 2021-09-26 liuhongt <hongtao.liu@intel.com>
406 * doc/extend.texi (Half-Precision): Remove storage only
407 description for _Float16 w/o avx512fp16.
409 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
411 * config/pru/constraints.md (Rrio): New constraint.
412 * config/pru/predicates.md (regio_operand): New predicate.
413 * config/pru/pru-pragma.c (pru_register_pragmas): Register
414 the __regio_symbol address space.
415 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
416 * config/pru/pru.c (pru_symref2ioregno): New helper function.
417 (pru_legitimate_address_p): Remove.
418 (pru_addr_space_legitimate_address_p): Use the address space
420 (pru_nongeneric_pointer_addrspace): New helper function.
421 (pru_insert_attributes): New function to validate __regio_symbol
423 (TARGET_INSERT_ATTRIBUTES): New macro.
424 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
425 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
426 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
427 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
429 (*regio_nozext_writesi): New pattern to write to I/O registers.
430 (*regio_zext_write_r30<EQS0:mode>): Ditto.
431 * doc/extend.texi: Document the new PRU Named Address Space.
433 2021-09-24 Patrick Palka <ppalka@redhat.com>
437 * real.c (encode_ieee_double): Avoid unwanted sign extension.
438 (encode_ieee_quad): Likewise.
440 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
442 PR rtl-optimization/102147
443 * ira-build.c (ira_conflict_vector_profitable_p): Make
444 profitability calculation independent of host compiler pointer and
447 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
449 * gimple-range-path.cc (path_range_query::path_range_query):
450 Move debugging header...
451 (path_range_query::precompute_ranges): ...here.
452 (path_range_query::internal_range_of_expr): Do not call
453 range_on_path_entry if NAME is defined in the current block.
455 2021-09-24 Richard Biener <rguenther@suse.de>
457 * cfghooks.c (verify_flow_info): Verify unallocated BB and
458 edge flags are not set.
460 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
462 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
464 (jt_path_registry::register_jump_thread): Call
465 cancel_invalid_paths.
466 * tree-ssa-threadupdate.h (class jt_path_registry): Add
467 cancel_invalid_paths.
469 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
471 PR tree-optimization/102400
472 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
473 result_vdef to zero value.
475 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
477 PR tree-optimization/102451
478 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
481 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
483 * config/i386/sse.md (cond_<insn><mode>): Extend to support
485 (cond_mul<mode>): Likewise.
486 (cond_div<mode>): Likewise.
487 (cond_<code><mode>): Likewise.
488 (cond_fma<mode>): Likewise.
489 (cond_fms<mode>): Likewise.
490 (cond_fnma<mode>): Likewise.
491 (cond_fnms<mode>): Likewise.
493 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
495 PR tree-optimization/102463
496 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
497 there is no range-ops handler, don't look for a relation.
499 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
501 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
502 non-executable_edge flag as parameter.
503 * gimple-range-cache.h (ranger_cache): Adjust prototype.
504 * gimple-range-gori.cc (gori_compute::gori_compute): Take
505 non-executable_edge flag as parameter.
506 (gori_compute::outgoing_edge_range_p): Check new flag.
507 * gimple-range-gori.h (gori_compute): Adjust prototype.
508 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
509 (gimple_ranger::range_on_edge): Check new flag.
510 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
511 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
512 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
513 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
514 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
515 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
516 (simplify_using_ranges::simplify_switch_using_ranges): Clear
517 EDGE_EXECUTABLE like it originally did.
518 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
519 NON_EXECUTABLE flags.
520 (simplify_using_ranges::simplify_using_ranges): Adjust.
521 * vr-values.h (class simplify_using_ranges): Adjust.
522 (simplify_using_ranges::set_range_query): Add non-executable flag param.
524 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
527 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
528 zero-width bit fields and return indicator.
529 (rs6000_discover_homogeneous_aggregate): Diagnose when the
530 presence of a zero-width bit field changes parameter passing in
533 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
535 * gimple-range-fold.cc (fold_using_range::range_of_phi):
536 Remove dominator check.
538 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
540 * gimple-range-path.cc (path_range_query::precompute_relations):
541 Hoist edge calculations before using EDGE_SUCC.
543 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
545 * configure.ac: Fix --with-multilib-list description.
546 * configure: Regenerate.
548 2021-09-23 Richard Biener <rguenther@suse.de>
550 PR tree-optimization/102448
551 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
552 Clear alignment info copied from DR_PTR_INFO.
554 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
556 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
558 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
559 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
560 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
561 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
562 (vcond<mode><mode>): New expander.
563 (vcond<mode><sseintvecmodelower>): Likewise.
564 (vcond<sseintvecmodelower><mode>): Likewise.
565 (vcondu<mode><sseintvecmodelower>): Likewise.
567 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
569 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
571 (extendv4hf<mode>2): Likewise.
572 (extendv2hfv2df2): Likewise.
573 (trunc<mode><ssePHmodelower>2): Likewise.
574 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
575 (trunc<mode>v4hf2): ... this, and drop constraints.
576 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
577 (truncv2dfv2hf2): ... this, and likewise.
579 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
581 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
583 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
585 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
586 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
587 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
589 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
591 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
592 (fixuns_trunchfhi2): Likewise.
593 (*fixuns_trunchfsi2zext): New define_insn.
594 * config/i386/sse.md (ssePHmodelower): New mode_attr.
595 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
596 New expander for same element vector fix_truncate.
597 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
598 Likewise for V4HF to V4SI/V4DI fix_truncate.
599 (fix<fixunssuffix>_truncv2hfv2di2):
600 Likeise for V2HF to V2DI fix_truncate.
602 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
604 * config/i386/i386.md (<code>hf3): New expander.
606 2021-09-23 liuhongt <hongtao.liu@intel.com>
608 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
609 (VFH_SF_AVX512VL): Extend to handle HFmode.
610 (VF_SF_AVX512VL): Deleted.
612 2021-09-23 liuhongt <hongtao.liu@intel.com>
614 * config/i386/i386.md (rinthf2): New expander.
615 (nearbyinthf2): New expander.
617 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
619 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
620 (class dom_jt_state): ...this and provide virtual overrides.
621 (dom_jt_state::register_equiv): New.
622 (class dom_jt_simplifier): Rename from
623 dom_jump_threader_simplifier.
624 (dom_jump_threader_simplifier::simplify): Rename...
625 (dom_jt_simplifier::simplify): ...to this.
626 (pass_dominator::execute): Use dom_jt_simplifier and
628 * tree-ssa-threadedge.c (jump_threader::jump_threader):
630 (jt_state::register_equivs_stmt): Abstract out...
631 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
633 (jump_threader::thread_around_empty_blocks): Update state.
634 (jump_threader::thread_through_normal_block): Same.
635 (jt_state::jt_state): Remove.
636 (jt_state::push): Remove pass specific bits. Keep block vector
638 (jt_state::append_path): New.
639 (jt_state::pop): Remove pass specific bits.
640 (jt_state::register_equiv): Same.
641 (jt_state::record_ranges_from_stmt): Same.
642 (jt_state::register_equivs_on_edge): Same. Rename...
643 (jt_state::register_equivs_edge): ...to this.
644 (jt_state::dump): New.
645 (jt_state::debug): New.
646 (jump_threader_simplifier::simplify): Remove.
647 (jt_state::get_path): New.
648 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
649 class. Expose common functionality as virtual methods.
650 (class jump_threader_simplifier): Same. Rename...
651 (class jt_simplifier): ...to this.
652 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
653 (class vrp_jt_simplifier): ...to this. Provide pass specific
655 (class vrp_jt_state): New.
656 (vrp_jump_threader_simplifier::simplify): Rename...
657 (vrp_jt_simplifier::simplify): ...to this. Inline code from
658 what used to be the base class.
659 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
662 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
665 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
668 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
669 Richard Biener <rguenther@suse.de>
671 * match.pd (negation simplifications): Implement some negation
672 folding transformations from fold-const.c's fold_negate_expr.
673 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
674 argument, to control whether the op should be simplified prior
675 to looking up/assigning a value number.
676 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
677 (vn_nary_simplify): Likewise.
678 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
679 now call vn_nary_build_or_lookup_1 disabling simplification.
681 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
683 PR tree-optimization/102087
684 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
685 Update bound/cmp/control for niter.
687 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
689 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
690 Move check for non-empty BB here.
691 (fur_source::register_outgoing_edges): ...from here.
693 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
695 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
696 Remove call to improve_range_with_equivs.
697 (path_range_query::improve_range_with_equivs): Remove
698 * gimple-range-path.h: Remove improve_range_with_equivs.
700 2021-09-22 dianhong xu <dianhong.xu@intel.com>
702 * config/i386/avx512fp16intrin.h:
703 (_mm512_mask_blend_ph): New intrinsic.
704 (_mm512_permutex2var_ph): Ditto.
705 (_mm512_permutexvar_ph): Ditto.
706 * config/i386/avx512fp16vlintrin.h:
707 (_mm256_mask_blend_ph): New intrinsic.
708 (_mm256_permutex2var_ph): Ditto.
709 (_mm256_permutexvar_ph): Ditto.
710 (_mm_mask_blend_ph): Ditto.
711 (_mm_permutex2var_ph): Ditto.
712 (_mm_permutexvar_ph): Ditto.
714 2021-09-22 dianhong xu <dianhong.xu@intel.com>
716 * config/i386/avx512fp16intrin.h: Add new intrinsics.
717 (_mm512_conj_pch): New intrinsic.
718 (_mm512_mask_conj_pch): Ditto.
719 (_mm512_maskz_conj_pch): Ditto.
720 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
721 (_mm256_conj_pch): New intrinsic.
722 (_mm256_mask_conj_pch): Ditto.
723 (_mm256_maskz_conj_pch): Ditto.
724 (_mm_conj_pch): Ditto.
725 (_mm_mask_conj_pch): Ditto.
726 (_mm_maskz_conj_pch): Ditto.
728 2021-09-22 dianhong xu <dianhong.xu@intel.com>
730 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
731 (_mm512_reduce_add_ph): New intrinsic.
732 (_mm512_reduce_mul_ph): Ditto.
733 (_mm512_reduce_min_ph): Ditto.
734 (_mm512_reduce_max_ph): Ditto.
735 * config/i386/avx512fp16vlintrin.h
736 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
737 (_mm256_reduce_add_ph): New intrinsic.
738 (_mm256_reduce_mul_ph): Ditto.
739 (_mm256_reduce_min_ph): Ditto.
740 (_mm256_reduce_max_ph): Ditto.
741 (_mm_reduce_add_ph): Ditto.
742 (_mm_reduce_mul_ph): Ditto.
743 (_mm_reduce_min_ph): Ditto.
744 (_mm_reduce_max_ph): Ditto.
746 2021-09-22 dianhong xu <dianhong.xu@intel.com>
748 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
749 __m128h_u): New typedef.
750 (_mm512_load_ph): New intrinsic.
751 (_mm256_load_ph): Ditto.
752 (_mm_load_ph): Ditto.
753 (_mm512_loadu_ph): Ditto.
754 (_mm256_loadu_ph): Ditto.
755 (_mm_loadu_ph): Ditto.
756 (_mm512_store_ph): Ditto.
757 (_mm256_store_ph): Ditto.
758 (_mm_store_ph): Ditto.
759 (_mm512_storeu_ph): Ditto.
760 (_mm256_storeu_ph): Ditto.
761 (_mm_storeu_ph): Ditto.
762 (_mm512_abs_ph): Ditto.
763 * config/i386/avx512fp16vlintrin.h
765 (_mm256_abs_ph): Ditto.
767 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
769 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
771 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
774 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
775 is actually a full copy of the source operand into the target.
776 Don't emit a strict low part move if source and target mode match.
778 2021-09-22 Jakub Jelinek <jakub@redhat.com>
781 * omp-expand.c (expand_omp_single): If region->exit is NULL,
782 assert region->entry is GIMPLE_OMP_SCOPE region and return.
784 2021-09-22 Jakub Jelinek <jakub@redhat.com>
786 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
787 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
788 arguments from 2 to 3.
789 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
790 allocate clause allocator and print align if present.
791 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
792 for omp_default_mem_alloc if align modifier is present. If align
793 modifier is present, use TREE_LIST to encode both allocator and
795 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
796 Handle align modifier on allocator clause if present.
798 2021-09-22 liuhongt <hongtao.liu@intel.com>
800 * config/i386/i386.md (define_attr "isa"): Add
802 (define_attr "enabled"): Correspond fma_or_avx512vl to
803 TARGET_FMA || TARGET_AVX512VL.
804 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
809 2021-09-22 liuhongt <hongtao.liu@intel.com>
811 * config/i386/i386.md (cstorehf3): New define_expand.
813 2021-09-22 liuhongt <hongtao.liu@intel.com>
815 * config/i386/i386.md (<rounding_insn>hf2): New expander.
816 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
817 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
818 Extend from VF_128 to VFH_128.
820 2021-09-22 liuhongt <hongtao.liu@intel.com>
822 * config/i386/i386-features.c (i386-features.c): Handle
824 * config/i386/i386.md (sqrthf2): New expander.
825 (*sqrthf2): New define_insn.
827 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
830 2021-09-22 liuhongt <hongtao.liu@intel.com>
832 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
834 (_mm_mask3_fcmadd_sch): Likewise.
835 (_mm_maskz_fcmadd_sch): Likewise.
836 (_mm_fcmadd_sch): Likewise.
837 (_mm_mask_fmadd_sch): Likewise.
838 (_mm_mask3_fmadd_sch): Likewise.
839 (_mm_maskz_fmadd_sch): Likewise.
840 (_mm_fmadd_sch): Likewise.
841 (_mm_mask_fcmadd_round_sch): Likewise.
842 (_mm_mask3_fcmadd_round_sch): Likewise.
843 (_mm_maskz_fcmadd_round_sch): Likewise.
844 (_mm_fcmadd_round_sch): Likewise.
845 (_mm_mask_fmadd_round_sch): Likewise.
846 (_mm_mask3_fmadd_round_sch): Likewise.
847 (_mm_maskz_fmadd_round_sch): Likewise.
848 (_mm_fmadd_round_sch): Likewise.
849 (_mm_fcmul_sch): Likewise.
850 (_mm_mask_fcmul_sch): Likewise.
851 (_mm_maskz_fcmul_sch): Likewise.
852 (_mm_fmul_sch): Likewise.
853 (_mm_mask_fmul_sch): Likewise.
854 (_mm_maskz_fmul_sch): Likewise.
855 (_mm_fcmul_round_sch): Likewise.
856 (_mm_mask_fcmul_round_sch): Likewise.
857 (_mm_maskz_fcmul_round_sch): Likewise.
858 (_mm_fmul_round_sch): Likewise.
859 (_mm_mask_fmul_round_sch): Likewise.
860 (_mm_maskz_fmul_round_sch): Likewise.
861 * config/i386/i386-builtin.def: Add corresponding new builtins.
863 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
864 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
865 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
867 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
868 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
870 * config/i386/subst.md (mask_scalarcz_name): New.
871 (mask_scalarc_name): Ditto.
872 (mask_scalarc_operand3): Ditto.
873 (mask_scalarcz_operand4): Ditto.
874 (round_scalarcz_name): Ditto.
875 (round_scalarc_mask_operand3): Ditto.
876 (round_scalarcz_mask_operand4): Ditto.
877 (round_scalarc_mask_op3): Ditto.
878 (round_scalarcz_mask_op4): Ditto.
879 (round_scalarcz_constraint): Ditto.
880 (round_scalarcz_nimm_predicate): Ditto.
881 (mask_scalarcz): Ditto.
882 (mask_scalarc): Ditto.
883 (round_scalarcz): Ditto.
885 2021-09-22 liuhongt <hongtao.liu@intel.com>
887 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
889 (_mm512_mask_fcmadd_pch): Likewise.
890 (_mm512_mask3_fcmadd_pch): Likewise.
891 (_mm512_maskz_fcmadd_pch): Likewise.
892 (_mm512_fmadd_pch): Likewise.
893 (_mm512_mask_fmadd_pch): Likewise.
894 (_mm512_mask3_fmadd_pch): Likewise.
895 (_mm512_maskz_fmadd_pch): Likewise.
896 (_mm512_fcmadd_round_pch): Likewise.
897 (_mm512_mask_fcmadd_round_pch): Likewise.
898 (_mm512_mask3_fcmadd_round_pch): Likewise.
899 (_mm512_maskz_fcmadd_round_pch): Likewise.
900 (_mm512_fmadd_round_pch): Likewise.
901 (_mm512_mask_fmadd_round_pch): Likewise.
902 (_mm512_mask3_fmadd_round_pch): Likewise.
903 (_mm512_maskz_fmadd_round_pch): Likewise.
904 (_mm512_fcmul_pch): Likewise.
905 (_mm512_mask_fcmul_pch): Likewise.
906 (_mm512_maskz_fcmul_pch): Likewise.
907 (_mm512_fmul_pch): Likewise.
908 (_mm512_mask_fmul_pch): Likewise.
909 (_mm512_maskz_fmul_pch): Likewise.
910 (_mm512_fcmul_round_pch): Likewise.
911 (_mm512_mask_fcmul_round_pch): Likewise.
912 (_mm512_maskz_fcmul_round_pch): Likewise.
913 (_mm512_fmul_round_pch): Likewise.
914 (_mm512_mask_fmul_round_pch): Likewise.
915 (_mm512_maskz_fmul_round_pch): Likewise.
916 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
918 (_mm_mask_fmadd_pch): Likewise.
919 (_mm_mask3_fmadd_pch): Likewise.
920 (_mm_maskz_fmadd_pch): Likewise.
921 (_mm256_fmadd_pch): Likewise.
922 (_mm256_mask_fmadd_pch): Likewise.
923 (_mm256_mask3_fmadd_pch): Likewise.
924 (_mm256_maskz_fmadd_pch): Likewise.
925 (_mm_fcmadd_pch): Likewise.
926 (_mm_mask_fcmadd_pch): Likewise.
927 (_mm_mask3_fcmadd_pch): Likewise.
928 (_mm_maskz_fcmadd_pch): Likewise.
929 (_mm256_fcmadd_pch): Likewise.
930 (_mm256_mask_fcmadd_pch): Likewise.
931 (_mm256_mask3_fcmadd_pch): Likewise.
932 (_mm256_maskz_fcmadd_pch): Likewise.
933 (_mm_fmul_pch): Likewise.
934 (_mm_mask_fmul_pch): Likewise.
935 (_mm_maskz_fmul_pch): Likewise.
936 (_mm256_fmul_pch): Likewise.
937 (_mm256_mask_fmul_pch): Likewise.
938 (_mm256_maskz_fmul_pch): Likewise.
939 (_mm_fcmul_pch): Likewise.
940 (_mm_mask_fcmul_pch): Likewise.
941 (_mm_maskz_fcmul_pch): Likewise.
942 (_mm256_fcmul_pch): Likewise.
943 (_mm256_mask_fcmul_pch): Likewise.
944 (_mm256_maskz_fcmul_pch): Likewise.
945 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
946 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
947 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
948 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
949 * config/i386/i386-builtin.def: Add new builtins.
950 * config/i386/i386-expand.c: Handle new builtin types.
951 * config/i386/subst.md (SUBST_CV): New.
953 (maskc_operand3): Ditto.
955 (sdc_maskz_name): Ditto.
956 (sdc_mask_op4): Ditto.
957 (sdc_mask_op5): Ditto.
958 (sdc_mask_mode512bit_condition): Ditto.
960 (round_maskc_operand3): Ditto.
961 (round_sdc_mask_operand4): Ditto.
962 (round_maskc_op3): Ditto.
963 (round_sdc_mask_op4): Ditto.
964 (round_saeonly_sdc_mask_operand5): Ditto.
965 * config/i386/sse.md (unspec): Add complex fma unspecs.
966 (avx512fmaskcmode): New.
967 (UNSPEC_COMPLEX_F_C_MA): Ditto.
968 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
969 (complexopname): Ditto.
970 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
971 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
972 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
974 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
975 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
977 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
979 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
980 rs6000-density-size-threshold, rs6000-density-penalty,
981 rs6000-density-load-pct-threshold,
982 rs6000-density-load-num-threshold): New parameter.
983 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
984 corresponding parameters.
986 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
988 * gimple-range-path.cc (path_range_query::defined_outside_path):
990 (path_range_query::range_on_path_entry): New.
991 (path_range_query::internal_range_of_expr): Resolve unknowns
993 (path_range_query::improve_range_with_equivs): New.
994 (path_range_query::ssa_range_in_phi): Resolve unknowns with
996 * gimple-range-path.h (class path_range_query): Add
997 defined_outside_path, range_on_path_entry, and
998 improve_range_with_equivs.
1000 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1002 * gimple-range-path.cc (path_range_query::add_to_imports): New.
1003 (path_range_query::add_copies_to_imports): New.
1004 (path_range_query::precompute_ranges): Call
1005 add_copies_to_imports.
1006 * gimple-range-path.h (class path_range_query): Add prototypes
1007 for add_copies_to_imports and add_to_imports.
1009 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1011 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1012 Remove useless code.
1014 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1016 * gimple-range-fold.h (class fur_source): Make oracle protected.
1017 * gimple-range-path.cc (path_range_query::path_range_query): Add
1018 resolve argument. Initialize oracle.
1019 (path_range_query::~path_range_query): Delete oracle.
1020 (path_range_query::range_of_stmt): Adapt to use relations.
1021 (path_range_query::precompute_ranges): Pre-compute relations.
1022 (class jt_fur_source): New
1023 (jt_fur_source::jt_fur_source): New.
1024 (jt_fur_source::register_relation): New.
1025 (jt_fur_source::query_relation): New.
1026 (path_range_query::precompute_relations): New.
1027 (path_range_query::precompute_phi_relations): New.
1028 * gimple-range-path.h (path_range_query): Add resolve argument.
1029 Add oracle, precompute_relations, precompute_phi_relations.
1030 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
1031 resolve argument to solver.
1033 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1035 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1036 Rename postfold_gcond_edges to register_outgoing_edges and
1038 (fold_using_range::postfold_gcond_edges): Rename...
1039 (fur_source::register_outgoing_edges): ...to this.
1040 * gimple-range-fold.h (postfold_gcond_edges): Rename to
1041 register_outgoing_edges and move to fur_source.
1043 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1045 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
1046 dom_info_available_p.
1048 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1050 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
1051 and quick_grow_cleared instead of safe_grow_cleared.
1053 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
1056 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
1059 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
1061 * configure.ac: Detect when the assembler supports new-style
1062 architecture extensions.
1063 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
1064 the full CPU string if the assembler can grok it.
1065 (arm_rewrite_march): Likewise but for the architecture.
1066 * config.in: Regenerate.
1067 * configure: Regenerate.
1069 2021-09-21 Richard Biener <rguenther@suse.de>
1071 PR tree-optimization/102421
1072 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
1073 adjust alignment info.
1075 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
1077 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
1078 bitfield stream out.
1080 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1082 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
1083 undefined edges, apply an equivalence if appropriate.
1084 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
1085 UNDEFINED if EDGE_EXECUTABLE is not set.
1086 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
1087 as EXECUTABLE upon startup.
1088 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
1089 EDGE_EXECUTABLE set.
1090 * vr-values.c (set_and_propagate_unexecutable): New.
1091 (simplify_using_ranges::fold_cond): Call set_and_propagate.
1092 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
1093 * vr-values.h: Add prototype.
1095 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1097 * value-relation.cc (equiv_oracle::register_initial_def): New.
1098 (equiv_oracle::register_relation): Call register_initial_def.
1099 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
1100 (relation_oracle::register_stmt): Check def block of PHI arguments.
1101 * value-relation.h (equiv_oracle): Add new prototypes.
1103 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
1105 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1106 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
1107 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
1108 __ROUNDING_MATH__ according to their corresponding flags.
1109 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
1110 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
1111 and __ROUNDING_MATH__.
1113 2021-09-20 Richard Biener <rguenther@suse.de>
1115 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1118 2021-09-20 Richard Biener <rguenther@suse.de>
1120 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1121 Do not compute alignment of the vectorized access here.
1123 2021-09-20 Richard Biener <rguenther@suse.de>
1125 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1126 Store -1 for runtime alias peeling iterations.
1128 2021-09-20 Richard Biener <rguenther@suse.de>
1130 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1132 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1134 * input.c (string_concat_db::record_string_concatenation)
1135 (string_concat_db::get_string_concatenation): Skip for
1136 'RESERVED_LOCATION_P'.
1138 2021-09-20 Richard Biener <rguenther@suse.de>
1140 PR tree-optimization/65206
1141 * tree-data-ref.h (struct data_reference): Add alt_indices,
1143 * tree-data-ref.c (free_data_ref): Release alt_indices.
1144 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1145 (create_data_ref): Adjust.
1146 (initialize_data_dependence_relation): Split into head
1147 and tail. When the base objects fail to match up try
1148 again with pointer-based analysis of indices.
1149 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1150 not compare the lazily computed alternate set of indices.
1152 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1154 * gcc.c: Test for execute OK when we find the
1155 programs for assembler linker and dsymutil and those
1156 were specified at configure-time.
1158 2021-09-19 Martin Sebor <msebor@redhat.com>
1160 PR middle-end/102403
1161 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1162 Correct a function pre/postcondition.
1164 2021-09-19 Martin Sebor <msebor@redhat.com>
1166 PR middle-end/102243
1167 * tree-ssa-strlen.c (get_range): Handle null cfun.
1169 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1171 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1172 unwinder shim as a convenience library.
1174 2021-09-19 Andrew Pinski <apinski@marvell.com>
1176 * doc/install.texi: Add note about
1177 binutils 2.35 is required for LTO usage.
1179 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1181 * tree-ssa-threadbackward.c
1182 (back_threader_registry::register_path): Use push_edge.
1183 * tree-ssa-threadedge.c
1184 (jump_threader::thread_around_empty_blocks): Same.
1185 (jump_threader::thread_through_normal_block): Same.
1186 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1188 * tree-ssa-threadupdate.c
1189 (jt_path_registry::allocate_thread_edge): Remove.
1190 (jt_path_registry::push_edge): New.
1191 (dump_jump_thread_path): Make static.
1192 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1195 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1197 * gimple-range-path.cc (path_range_query::path_range_query): Add
1199 (path_range_query::dump): Remove extern declaration of dump_ranger.
1200 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1201 * gimple-range-trace.h (dump_ranger): Add prototype.
1203 2021-09-19 John Ericson <git@JohnEricson.me>
1205 * gcc.c (find_a_program): New function, factored out of...
1206 (find_a_file): Here.
1207 (execute): Use find_a_program when looking for programs rather
1210 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1212 * config/avr/avr-mcus.def: Add atmega324pb.
1213 * doc/avr-mmcu.texi: Corresponding changes.
1215 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1218 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1219 the same transformations as to @1. For comparisons against NaN,
1220 don't check HONOR_SNANS but confirm that neither operand is a
1223 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1225 * attribs.c (make_unique_name): Delete.
1226 * attribs.h (make_unique_name): Delete.
1228 2021-09-19 Andrew Pinski <apinski@marvell.com>
1230 * lra-constraints.c (check_and_process_move): Assert
1231 that dclass and sclass are greater than or equal to NO_REGS.
1233 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1235 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
1236 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
1237 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
1239 2021-09-18 liuhongt <hongtao.liu@intel.com>
1241 * config/i386/i386-features.c (remove_partial_avx_dependency):
1242 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
1243 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
1245 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1247 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
1248 if file/namespace scope variable doesn't have predetermined sharing,
1249 treat it as if there was default(none).
1251 2021-09-18 liuhongt <hongtao.liu@intel.com>
1253 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
1255 (_mm_mask_fmadd_sh): Likewise.
1256 (_mm_mask3_fmadd_sh): Likewise.
1257 (_mm_maskz_fmadd_sh): Likewise.
1258 (_mm_fmadd_round_sh): Likewise.
1259 (_mm_mask_fmadd_round_sh): Likewise.
1260 (_mm_mask3_fmadd_round_sh): Likewise.
1261 (_mm_maskz_fmadd_round_sh): Likewise.
1262 (_mm_fnmadd_sh): Likewise.
1263 (_mm_mask_fnmadd_sh): Likewise.
1264 (_mm_mask3_fnmadd_sh): Likewise.
1265 (_mm_maskz_fnmadd_sh): Likewise.
1266 (_mm_fnmadd_round_sh): Likewise.
1267 (_mm_mask_fnmadd_round_sh): Likewise.
1268 (_mm_mask3_fnmadd_round_sh): Likewise.
1269 (_mm_maskz_fnmadd_round_sh): Likewise.
1270 (_mm_fmsub_sh): Likewise.
1271 (_mm_mask_fmsub_sh): Likewise.
1272 (_mm_mask3_fmsub_sh): Likewise.
1273 (_mm_maskz_fmsub_sh): Likewise.
1274 (_mm_fmsub_round_sh): Likewise.
1275 (_mm_mask_fmsub_round_sh): Likewise.
1276 (_mm_mask3_fmsub_round_sh): Likewise.
1277 (_mm_maskz_fmsub_round_sh): Likewise.
1278 (_mm_fnmsub_sh): Likewise.
1279 (_mm_mask_fnmsub_sh): Likewise.
1280 (_mm_mask3_fnmsub_sh): Likewise.
1281 (_mm_maskz_fnmsub_sh): Likewise.
1282 (_mm_fnmsub_round_sh): Likewise.
1283 (_mm_mask_fnmsub_round_sh): Likewise.
1284 (_mm_mask3_fnmsub_round_sh): Likewise.
1285 (_mm_maskz_fnmsub_round_sh): Likewise.
1286 * config/i386/i386-builtin-types.def
1287 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1288 * config/i386/i386-builtin.def: Add new builtins.
1289 * config/i386/i386-expand.c: Handle new builtin type.
1290 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1291 Ajdust to support FP16.
1292 (fmai_vmfmsub_<mode><round_name>): Ditto.
1293 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1294 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1295 (*fmai_fmadd_<mode>): Ditto.
1296 (*fmai_fmsub_<mode>): Ditto.
1297 (*fmai_fnmadd_<mode><round_name>): Ditto.
1298 (*fmai_fnmsub_<mode><round_name>): Ditto.
1299 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1300 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1301 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1302 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1303 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1304 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1305 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1306 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1307 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1308 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1309 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1310 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1311 adjust to support FP16.
1312 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1313 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1314 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1317 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1319 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1321 (maskload<mode><avx512fmaskmodelower>): Ditto.
1322 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1324 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1326 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1328 (ix86_expand_copysign): Ditto.
1329 (ix86_expand_xorsign): Ditto.
1330 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1332 (ix86_build_signbit_mask): Ditto.
1333 (ix86_can_change_mode_class): Ditto.
1334 * config/i386/i386.md
1335 (SSEMODEF): Add HFmode.
1336 (ssevecmodef): Ditto.
1337 (<code>hf2): New define_expand.
1338 (*<code>hf2_1): New define_insn_and_split.
1339 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1340 (xorsign<mode>): Ditto.
1341 * config/i386/sse.md (VFB): New mode iterator.
1342 (VFB_128_256): Ditto.
1344 (sseintvecmode2): Support HF vector mode.
1345 (<code><mode>2): Use new mode iterator.
1346 (*<code><mode>2): Ditto.
1347 (copysign<mode>3): Ditto.
1348 (xorsign<mode>3): Ditto.
1349 (<code><mode>3<mask_name>): Ditto.
1350 (<code><mode>3<mask_name>): Ditto.
1351 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1352 (<sse>_andnot<mode>3<mask_name>): Ditto.
1353 (*<code><mode>3<mask_name>): Ditto.
1354 (*<code><mode>3<mask_name>): Ditto.
1356 2021-09-18 liuhongt <hongtao.liu@intel.com>
1358 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1360 (_mm512_mask3_fmadd_ph): Likewise.
1361 (_mm512_maskz_fmadd_ph): Likewise.
1362 (_mm512_fmadd_round_ph): Likewise.
1363 (_mm512_mask_fmadd_round_ph): Likewise.
1364 (_mm512_mask3_fmadd_round_ph): Likewise.
1365 (_mm512_maskz_fmadd_round_ph): Likewise.
1366 (_mm512_fnmadd_ph): Likewise.
1367 (_mm512_mask_fnmadd_ph): Likewise.
1368 (_mm512_mask3_fnmadd_ph): Likewise.
1369 (_mm512_maskz_fnmadd_ph): Likewise.
1370 (_mm512_fnmadd_round_ph): Likewise.
1371 (_mm512_mask_fnmadd_round_ph): Likewise.
1372 (_mm512_mask3_fnmadd_round_ph): Likewise.
1373 (_mm512_maskz_fnmadd_round_ph): Likewise.
1374 (_mm512_fmsub_ph): Likewise.
1375 (_mm512_mask_fmsub_ph): Likewise.
1376 (_mm512_mask3_fmsub_ph): Likewise.
1377 (_mm512_maskz_fmsub_ph): Likewise.
1378 (_mm512_fmsub_round_ph): Likewise.
1379 (_mm512_mask_fmsub_round_ph): Likewise.
1380 (_mm512_mask3_fmsub_round_ph): Likewise.
1381 (_mm512_maskz_fmsub_round_ph): Likewise.
1382 (_mm512_fnmsub_ph): Likewise.
1383 (_mm512_mask_fnmsub_ph): Likewise.
1384 (_mm512_mask3_fnmsub_ph): Likewise.
1385 (_mm512_maskz_fnmsub_ph): Likewise.
1386 (_mm512_fnmsub_round_ph): Likewise.
1387 (_mm512_mask_fnmsub_round_ph): Likewise.
1388 (_mm512_mask3_fnmsub_round_ph): Likewise.
1389 (_mm512_maskz_fnmsub_round_ph): Likewise.
1390 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1392 (_mm256_mask_fmadd_ph): Likewise.
1393 (_mm256_mask3_fmadd_ph): Likewise.
1394 (_mm256_maskz_fmadd_ph): Likewise.
1395 (_mm_fmadd_ph): Likewise.
1396 (_mm_mask_fmadd_ph): Likewise.
1397 (_mm_mask3_fmadd_ph): Likewise.
1398 (_mm_maskz_fmadd_ph): Likewise.
1399 (_mm256_fnmadd_ph): Likewise.
1400 (_mm256_mask_fnmadd_ph): Likewise.
1401 (_mm256_mask3_fnmadd_ph): Likewise.
1402 (_mm256_maskz_fnmadd_ph): Likewise.
1403 (_mm_fnmadd_ph): Likewise.
1404 (_mm_mask_fnmadd_ph): Likewise.
1405 (_mm_mask3_fnmadd_ph): Likewise.
1406 (_mm_maskz_fnmadd_ph): Likewise.
1407 (_mm256_fmsub_ph): Likewise.
1408 (_mm256_mask_fmsub_ph): Likewise.
1409 (_mm256_mask3_fmsub_ph): Likewise.
1410 (_mm256_maskz_fmsub_ph): Likewise.
1411 (_mm_fmsub_ph): Likewise.
1412 (_mm_mask_fmsub_ph): Likewise.
1413 (_mm_mask3_fmsub_ph): Likewise.
1414 (_mm_maskz_fmsub_ph): Likewise.
1415 (_mm256_fnmsub_ph): Likewise.
1416 (_mm256_mask_fnmsub_ph): Likewise.
1417 (_mm256_mask3_fnmsub_ph): Likewise.
1418 (_mm256_maskz_fnmsub_ph): Likewise.
1419 (_mm_fnmsub_ph): Likewise.
1420 (_mm_mask_fnmsub_ph): Likewise.
1421 (_mm_mask3_fnmsub_ph): Likewise.
1422 (_mm_maskz_fnmsub_ph): Likewise.
1423 * config/i386/i386-builtin.def: Add corresponding new builtins.
1424 * config/i386/sse.md
1425 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1426 support HF vector modes.
1427 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1429 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1430 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1431 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1432 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1433 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1434 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1435 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1437 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1438 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1439 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1440 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1441 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1442 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1444 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1445 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1446 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1447 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1448 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1449 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1450 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1452 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1453 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1454 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1455 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1456 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1458 2021-09-18 liuhongt <hongtao.liu@intel.com>
1460 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1462 (_mm512_mask_fmaddsub_ph): Likewise.
1463 (_mm512_mask3_fmaddsub_ph): Likewise.
1464 (_mm512_maskz_fmaddsub_ph): Likewise.
1465 (_mm512_fmaddsub_round_ph): Likewise.
1466 (_mm512_mask_fmaddsub_round_ph): Likewise.
1467 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1468 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1469 (_mm512_mask_fmsubadd_ph): Likewise.
1470 (_mm512_mask3_fmsubadd_ph): Likewise.
1471 (_mm512_maskz_fmsubadd_ph): Likewise.
1472 (_mm512_fmsubadd_round_ph): Likewise.
1473 (_mm512_mask_fmsubadd_round_ph): Likewise.
1474 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1475 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1476 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1478 (_mm256_mask_fmaddsub_ph): Likewise.
1479 (_mm256_mask3_fmaddsub_ph): Likewise.
1480 (_mm256_maskz_fmaddsub_ph): Likewise.
1481 (_mm_fmaddsub_ph): Likewise.
1482 (_mm_mask_fmaddsub_ph): Likewise.
1483 (_mm_mask3_fmaddsub_ph): Likewise.
1484 (_mm_maskz_fmaddsub_ph): Likewise.
1485 (_mm256_fmsubadd_ph): Likewise.
1486 (_mm256_mask_fmsubadd_ph): Likewise.
1487 (_mm256_mask3_fmsubadd_ph): Likewise.
1488 (_mm256_maskz_fmsubadd_ph): Likewise.
1489 (_mm_fmsubadd_ph): Likewise.
1490 (_mm_mask_fmsubadd_ph): Likewise.
1491 (_mm_mask3_fmsubadd_ph): Likewise.
1492 (_mm_maskz_fmsubadd_ph): Likewise.
1493 * config/i386/i386-builtin.def: Add corresponding new builtins.
1494 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1495 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1496 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1498 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1500 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1501 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1502 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1504 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1505 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1507 2021-09-18 liuhongt <hongtao.liu@intel.com>
1510 * config/i386/i386.c (ix86_print_operand): Handle
1511 V8HF/V16HF/V32HFmode.
1512 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1513 * config/i386/sse.md (avx512bcst): Remove.
1515 2021-09-17 Martin Sebor <msebor@redhat.com>
1517 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1518 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1519 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1520 (check_defs): Add comment.
1521 (can_skip_redundant_opnd): Update comment.
1522 (compute_uninit_opnds_pos): Adjust to namespace change.
1523 (find_pdom): Move to gimple-predicate-analysis.cc.
1525 (struct uninit_undef_val_t): New.
1526 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1527 (find_control_equiv_block): Same.
1528 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1529 (MAX_SWITCH_CASES): Same.
1530 (compute_control_dep_chain): Same.
1531 (find_uninit_use): Use predicate analyzer.
1532 (struct pred_info): Move to gimple-predicate-analysis.
1533 (convert_control_dep_chain_into_preds): Same.
1534 (find_predicates): Same.
1535 (collect_phi_def_edges): Same.
1536 (warn_uninitialized_phi): Use predicate analyzer.
1537 (find_def_preds): Move to gimple-predicate-analysis.
1538 (dump_pred_info): Same.
1539 (dump_pred_chain): Same.
1540 (dump_predicates): Same.
1541 (destroy_predicate_vecs): Remove.
1542 (execute_late_warn_uninitialized): New.
1543 (get_cmp_code): Move to gimple-predicate-analysis.
1544 (is_value_included_in): Same.
1545 (value_sat_pred_p): Same.
1546 (find_matching_predicate_in_rest_chains): Same.
1547 (is_use_properly_guarded): Same.
1548 (prune_uninit_phi_opnds): Same.
1549 (find_var_cmp_const): Same.
1550 (use_pred_not_overlap_with_undef_path_pred): Same.
1551 (pred_equal_p): Same.
1552 (is_neq_relop_p): Same.
1553 (is_neq_zero_form_p): Same.
1554 (pred_expr_equal_p): Same.
1555 (is_pred_expr_subset_of): Same.
1556 (is_pred_chain_subset_of): Same.
1557 (is_included_in): Same.
1558 (is_superset_of): Same.
1560 (simplify_pred): Same.
1561 (simplify_preds_2): Same.
1562 (simplify_preds_3): Same.
1563 (simplify_preds_4): Same.
1564 (simplify_preds): Same.
1566 (push_to_worklist): Same.
1567 (get_pred_info_from_cmp): Same.
1568 (is_degenerated_phi): Same.
1569 (normalize_one_pred_1): Same.
1570 (normalize_one_pred): Same.
1571 (normalize_one_pred_chain): Same.
1572 (normalize_preds): Same.
1573 (can_one_predicate_be_invalidated_p): Same.
1574 (can_chain_union_be_invalidated_p): Same.
1575 (uninit_uses_cannot_happen): Same.
1576 (pass_late_warn_uninitialized::execute): Define.
1577 * gimple-predicate-analysis.cc: New file.
1578 * gimple-predicate-analysis.h: New file.
1580 2021-09-17 Julian Brown <julian@codesourcery.com>
1582 * config/gcn/gcn.c (gimple.h): Include.
1583 (gcn_fork_join): Emit barrier for worker-level joins.
1584 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1585 writes_gang_private bitmap parameter. Set bit for blocks
1586 containing gang-private variable writes.
1587 (worker_single_simple): Don't emit barrier after predicated block.
1588 (worker_single_copy): Don't emit barrier if we're not broadcasting
1589 anything and the block contains no gang-private writes.
1590 (neuter_worker_single): Don't predicate blocks that only contain
1591 NOPs or internal marker functions. Pass has_gang_private_write
1592 argument to worker_single_copy.
1593 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1595 2021-09-17 Julian Brown <julian@codesourcery.com>
1597 * config/gcn/gcn-protos.h
1598 (gcn_goacc_create_worker_broadcast_record): Update prototype.
1599 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
1600 preallocated block of LDS memory. Do not cache/share decls for
1601 reduction temporaries between invocations.
1602 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
1603 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
1604 and return temporary LDS space at that offset. Return pointer in
1606 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
1608 (ACC_LDS_SIZE): Define as acc_lds_size.
1609 (gcn_init_machine_status): Don't initialise lds_allocated,
1610 lds_allocs, reduc_decls fields of machine function struct.
1611 (gcn_option_override): Handle default size for gang-private
1612 variables and -mgang-private-size option.
1613 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
1614 initialising M0_REG.
1615 (gcn_shared_mem_layout): New function.
1616 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
1617 gang_private_hwm variable.
1618 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
1619 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
1620 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
1621 * config/gcn/gcn.opt (gang_private_size_opt): New global.
1622 (mgang-private-size=): New option.
1623 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
1625 * doc/tm.texi: Regenerate.
1626 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
1628 (build_sender_ref): Handle sender_decl being pointer.
1629 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
1630 parameters. Pass placement argument to
1631 create_worker_broadcast_record hook invocations. Handle
1632 sender_decl being pointer and isolate_broadcasts inserting extra
1634 (blk_offset_map_t): Add typedef.
1635 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
1636 preallocated range to worker_single_copy call.
1637 (dfs_broadcast_reachable_1): New function.
1638 (idx_decl_pair_t, used_range_vec_t): New typedefs.
1639 (sort_size_descending): New function.
1640 (addr_range): New class.
1641 (splay_tree_compare_addr_range, splay_tree_free_key)
1642 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
1643 (execute_omp_oacc_neuter_broadcast): Rename to...
1644 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
1645 parameters. Arrange layout of shared memory for broadcast
1647 (execute_omp_oacc_neuter_broadcast): New function.
1648 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
1649 handling from here. Enable pass for all OpenACC routines in order
1650 to call shared memory-layout hook.
1651 * target.def (create_worker_broadcast_record): Add OFFSET
1653 (shared_mem_layout): New hook.
1655 2021-09-17 Julian Brown <julian@codesourcery.com>
1656 Thomas Schwinge <thomas@codesourcery.com>
1658 * omp-oacc-neuter-broadcast.cc
1659 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
1661 (execute_omp_oacc_neuter_broadcast): Adjust.
1663 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1665 * value-relation.cc (class equiv_chain): Move to header file.
1666 (path_oracle::path_oracle): New.
1667 (path_oracle::~path_oracle): New.
1668 (path_oracle::register_relation): New.
1669 (path_oracle::query_relation): New.
1670 (path_oracle::reset_path): New.
1671 (path_oracle::dump): New.
1672 * value-relation.h (class equiv_chain): Move to here.
1673 (class path_oracle): New.
1675 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1677 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
1679 * gimple-range-fold.cc (fur_depend::register_relation): Use
1680 register_stmt/edge routines.
1681 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
1682 (equiv_oracle::equiv_oracle): Create self equivalence cache.
1683 (equiv_oracle::~equiv_oracle): Release same.
1684 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
1685 are no equivalences.
1686 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
1687 (equiv_oracle::register_relation): Rename from register_equiv.
1688 (relation_chain_head::find_relation): Relocate from dom_oracle.
1689 (relation_oracle::register_stmt): New.
1690 (relation_oracle::register_edge): New.
1691 (dom_oracle::*): Rename from relation_oracle.
1692 (dom_oracle::register_relation): Adjust to call equiv_oracle.
1693 (dom_oracle::set_one_relation): Split from register_relation.
1694 (dom_oracle::register_transitives): Consolidate 2 methods.
1695 (dom_oracle::find_relation_block): Move core to relation_chain.
1696 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
1697 * value-relation.h (class relation_oracle): New pure virtual base.
1698 (class equiv_oracle): Inherit from relation_oracle and adjust.
1699 (class dom_oracle): Rename from old relation_oracle and adjust.
1701 2021-09-17 Martin Sebor <msebor@redhat.com>
1703 PR middle-end/102200
1704 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
1705 (handle_min_max_size): Change argument. Store original SSA_NAME for
1706 operands to potentially distinct (sub)objects.
1707 (compute_objsize_r): Adjust call to the above.
1709 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1711 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
1712 (rs6000_new_builtin_vectorized_function): New function.
1713 (rs6000_new_builtin_md_vectorized_function): Likewise.
1714 (rs6000_builtin_vectorized_function): Call
1715 rs6000_new_builtin_vectorized_function.
1716 (rs6000_builtin_md_vectorized_function): Call
1717 rs6000_new_builtin_md_vectorized_function.
1719 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1721 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
1722 (ASSEMBLE_PAIR): Likewise.
1723 (BUILD_ACC): Likewise.
1724 (DISASSEMBLE_ACC): Likewise.
1725 (DISASSEMBLE_PAIR): Likewise.
1726 (PMXVBF16GER2): Likewise.
1727 (PMXVBF16GER2NN): Likewise.
1728 (PMXVBF16GER2NP): Likewise.
1729 (PMXVBF16GER2PN): Likewise.
1730 (PMXVBF16GER2PP): Likewise.
1731 (PMXVF16GER2): Likewise.
1732 (PMXVF16GER2NN): Likewise.
1733 (PMXVF16GER2NP): Likewise.
1734 (PMXVF16GER2PN): Likewise.
1735 (PMXVF16GER2PP): Likewise.
1736 (PMXVF32GER): Likewise.
1737 (PMXVF32GERNN): Likewise.
1738 (PMXVF32GERNP): Likewise.
1739 (PMXVF32GERPN): Likewise.
1740 (PMXVF32GERPP): Likewise.
1741 (PMXVF64GER): Likewise.
1742 (PMXVF64GERNN): Likewise.
1743 (PMXVF64GERNP): Likewise.
1744 (PMXVF64GERPN): Likewise.
1745 (PMXVF64GERPP): Likewise.
1746 (PMXVI16GER2): Likewise.
1747 (PMXVI16GER2PP): Likewise.
1748 (PMXVI16GER2S): Likewise.
1749 (PMXVI16GER2SPP): Likewise.
1750 (PMXVI4GER8): Likewise.
1751 (PMXVI4GER8PP): Likewise.
1752 (PMXVI8GER4): Likewise.
1753 (PMXVI8GER4PP): Likewise.
1754 (PMXVI8GER4SPP): Likewise.
1755 (XVBF16GER2): Likewise.
1756 (XVBF16GER2NN): Likewise.
1757 (XVBF16GER2NP): Likewise.
1758 (XVBF16GER2PN): Likewise.
1759 (XVBF16GER2PP): Likewise.
1760 (XVF16GER2): Likewise.
1761 (XVF16GER2NN): Likewise.
1762 (XVF16GER2NP): Likewise.
1763 (XVF16GER2PN): Likewise.
1764 (XVF16GER2PP): Likewise.
1765 (XVF32GER): Likewise.
1766 (XVF32GERNN): Likewise.
1767 (XVF32GERNP): Likewise.
1768 (XVF32GERPN): Likewise.
1769 (XVF32GERPP): Likewise.
1770 (XVF64GER): Likewise.
1771 (XVF64GERNN): Likewise.
1772 (XVF64GERNP): Likewise.
1773 (XVF64GERPN): Likewise.
1774 (XVF64GERPP): Likewise.
1775 (XVI16GER2): Likewise.
1776 (XVI16GER2PP): Likewise.
1777 (XVI16GER2S): Likewise.
1778 (XVI16GER2SPP): Likewise.
1779 (XVI4GER8): Likewise.
1780 (XVI4GER8PP): Likewise.
1781 (XVI8GER4): Likewise.
1782 (XVI8GER4PP): Likewise.
1783 (XVI8GER4SPP): Likewise.
1784 (XXMFACC): Likewise.
1785 (XXMTACC): Likewise.
1786 (XXSETACCZ): Likewise.
1787 (ASSEMBLE_PAIR_V): Likewise.
1788 (BUILD_PAIR): Likewise.
1789 (DISASSEMBLE_PAIR_V): Likewise.
1792 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
1793 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
1794 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
1795 (parse_bif_attrs): Handle ismmaint.
1796 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
1797 (write_bif_static_init): Handle ismmaint.
1799 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1801 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
1803 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
1804 (rs6000_new_builtin_valid_without_lhs): New function.
1805 (rs6000_gimple_fold_new_mma_builtin): Likewise.
1806 (rs6000_gimple_fold_new_builtin): Likewise.
1808 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
1810 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
1811 Destruct stale Value objects.
1812 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
1815 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
1818 * match.pd (shift optimizations): Disable recent sign-changing
1819 optimization for shifts by zero, these will be folded later.
1821 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1823 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
1824 [power9] to [always].
1826 2021-09-17 Richard Biener <rguenther@suse.de>
1828 * tree-vect-stmts.c (vectorizable_load): Do not frob
1831 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1833 * config/i386/i386-features.c (remove_partial_avx_dependency):
1834 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
1835 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
1837 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
1839 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1840 * config/i386/i386.md (SSE FP to FP splitters): Replace
1841 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
1842 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
1843 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
1844 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
1845 * config/i386/x86-tune.def
1846 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
1847 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1849 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1852 * config/i386/i386-features.c (remove_partial_avx_dependency):
1853 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
1854 before generating vxorps.
1856 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1858 * config/i386/i386-options.c (processor_cost_table): Use
1859 tremont_cost for Tremont.
1860 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
1861 (tremont_memset): Likewise.
1862 (tremont_cost): Likewise.
1863 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
1866 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1868 * common/config/i386/i386-common.c: Use Haswell scheduling model
1870 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
1872 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
1874 (ix86_adjust_cost): Handle Tremont.
1875 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
1877 (X86_TUNE_USE_LEAVE): Likewise.
1878 (X86_TUNE_PUSH_MEMORY): Likewise.
1879 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
1880 (X86_TUNE_USE_CLTD): Likewise.
1881 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
1882 (X86_TUNE_AVOID_MFENCE): Likewise.
1883 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
1884 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
1885 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
1886 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
1887 (X86_TUNE_OPT_AGU): Likewise.
1888 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
1889 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
1890 (X86_TUNE_EXPAND_ABS): Likewise.
1891 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
1892 (X86_TUNE_SLOW_PSHUFB): Likewise.
1894 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
1896 PR rtl-optimization/102306
1897 * combine.c (try_combine): Abort the combination if we are about to
1898 duplicate volatile references.
1900 2021-09-17 liuhongt <hongtao.liu@intel.com>
1902 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
1904 (_mm256_undefined_ph): Likewise.
1905 (_mm512_undefined_ph): Likewise.
1906 (_mm_cvtsh_h): Likewise.
1907 (_mm256_cvtsh_h): Likewise.
1908 (_mm512_cvtsh_h): Likewise.
1909 (_mm512_castph_ps): Likewise.
1910 (_mm512_castph_pd): Likewise.
1911 (_mm512_castph_si512): Likewise.
1912 (_mm512_castph512_ph128): Likewise.
1913 (_mm512_castph512_ph256): Likewise.
1914 (_mm512_castph128_ph512): Likewise.
1915 (_mm512_castph256_ph512): Likewise.
1916 (_mm512_zextph128_ph512): Likewise.
1917 (_mm512_zextph256_ph512): Likewise.
1918 (_mm512_castps_ph): Likewise.
1919 (_mm512_castpd_ph): Likewise.
1920 (_mm512_castsi512_ph): Likewise.
1921 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
1923 (_mm256_castph_ps): Likewise.
1924 (_mm_castph_pd): Likewise.
1925 (_mm256_castph_pd): Likewise.
1926 (_mm_castph_si128): Likewise.
1927 (_mm256_castph_si256): Likewise.
1928 (_mm_castps_ph): Likewise.
1929 (_mm256_castps_ph): Likewise.
1930 (_mm_castpd_ph): Likewise.
1931 (_mm256_castpd_ph): Likewise.
1932 (_mm_castsi128_ph): Likewise.
1933 (_mm256_castsi256_ph): Likewise.
1934 (_mm256_castph256_ph128): Likewise.
1935 (_mm256_castph128_ph256): Likewise.
1936 (_mm256_zextph128_ph256): Likewise.
1938 2021-09-17 liuhongt <hongtao.liu@intel.com>
1940 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
1942 (_mm_mask_cvtsh_ss): Likewise.
1943 (_mm_maskz_cvtsh_ss): Likewise.
1944 (_mm_cvtsh_sd): Likewise.
1945 (_mm_mask_cvtsh_sd): Likewise.
1946 (_mm_maskz_cvtsh_sd): Likewise.
1947 (_mm_cvt_roundsh_ss): Likewise.
1948 (_mm_mask_cvt_roundsh_ss): Likewise.
1949 (_mm_maskz_cvt_roundsh_ss): Likewise.
1950 (_mm_cvt_roundsh_sd): Likewise.
1951 (_mm_mask_cvt_roundsh_sd): Likewise.
1952 (_mm_maskz_cvt_roundsh_sd): Likewise.
1953 (_mm_cvtss_sh): Likewise.
1954 (_mm_mask_cvtss_sh): Likewise.
1955 (_mm_maskz_cvtss_sh): Likewise.
1956 (_mm_cvtsd_sh): Likewise.
1957 (_mm_mask_cvtsd_sh): Likewise.
1958 (_mm_maskz_cvtsd_sh): Likewise.
1959 (_mm_cvt_roundss_sh): Likewise.
1960 (_mm_mask_cvt_roundss_sh): Likewise.
1961 (_mm_maskz_cvt_roundss_sh): Likewise.
1962 (_mm_cvt_roundsd_sh): Likewise.
1963 (_mm_mask_cvt_roundsd_sh): Likewise.
1964 (_mm_maskz_cvt_roundsd_sh): Likewise.
1965 * config/i386/i386-builtin-types.def
1966 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
1967 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
1968 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
1969 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
1970 * config/i386/i386-builtin.def: Add corrresponding new builtins.
1971 * config/i386/i386-expand.c: Handle new builtin types.
1972 * config/i386/sse.md (VF48_128): New mode iterator.
1973 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
1975 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
1978 2021-09-17 liuhongt <hongtao.liu@intel.com>
1980 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
1982 (_mm512_mask_cvtph_pd): Likewise.
1983 (_mm512_maskz_cvtph_pd): Likewise.
1984 (_mm512_cvt_roundph_pd): Likewise.
1985 (_mm512_mask_cvt_roundph_pd): Likewise.
1986 (_mm512_maskz_cvt_roundph_pd): Likewise.
1987 (_mm512_cvtxph_ps): Likewise.
1988 (_mm512_mask_cvtxph_ps): Likewise.
1989 (_mm512_maskz_cvtxph_ps): Likewise.
1990 (_mm512_cvtx_roundph_ps): Likewise.
1991 (_mm512_mask_cvtx_roundph_ps): Likewise.
1992 (_mm512_maskz_cvtx_roundph_ps): Likewise.
1993 (_mm512_cvtxps_ph): Likewise.
1994 (_mm512_mask_cvtxps_ph): Likewise.
1995 (_mm512_maskz_cvtxps_ph): Likewise.
1996 (_mm512_cvtx_roundps_ph): Likewise.
1997 (_mm512_mask_cvtx_roundps_ph): Likewise.
1998 (_mm512_maskz_cvtx_roundps_ph): Likewise.
1999 (_mm512_cvtpd_ph): Likewise.
2000 (_mm512_mask_cvtpd_ph): Likewise.
2001 (_mm512_maskz_cvtpd_ph): Likewise.
2002 (_mm512_cvt_roundpd_ph): Likewise.
2003 (_mm512_mask_cvt_roundpd_ph): Likewise.
2004 (_mm512_maskz_cvt_roundpd_ph): Likewise.
2005 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
2007 (_mm_mask_cvtph_pd): Likewise.
2008 (_mm_maskz_cvtph_pd): Likewise.
2009 (_mm256_cvtph_pd): Likewise.
2010 (_mm256_mask_cvtph_pd): Likewise.
2011 (_mm256_maskz_cvtph_pd): Likewise.
2012 (_mm_cvtxph_ps): Likewise.
2013 (_mm_mask_cvtxph_ps): Likewise.
2014 (_mm_maskz_cvtxph_ps): Likewise.
2015 (_mm256_cvtxph_ps): Likewise.
2016 (_mm256_mask_cvtxph_ps): Likewise.
2017 (_mm256_maskz_cvtxph_ps): Likewise.
2018 (_mm_cvtxps_ph): Likewise.
2019 (_mm_mask_cvtxps_ph): Likewise.
2020 (_mm_maskz_cvtxps_ph): Likewise.
2021 (_mm256_cvtxps_ph): Likewise.
2022 (_mm256_mask_cvtxps_ph): Likewise.
2023 (_mm256_maskz_cvtxps_ph): Likewise.
2024 (_mm_cvtpd_ph): Likewise.
2025 (_mm_mask_cvtpd_ph): Likewise.
2026 (_mm_maskz_cvtpd_ph): Likewise.
2027 (_mm256_cvtpd_ph): Likewise.
2028 (_mm256_mask_cvtpd_ph): Likewise.
2029 (_mm256_maskz_cvtpd_ph): Likewise.
2030 * config/i386/i386-builtin.def: Add corresponding new builtins.
2031 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2032 * config/i386/i386-expand.c: Handle new builtin types.
2033 * config/i386/sse.md
2034 (VF4_128_8_256): New.
2035 (VF48H_AVX512VL): Ditto.
2036 (ssePHmode): Add HF vector modes.
2037 (castmode): Add new convertable modes.
2040 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
2041 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2042 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2043 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2044 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2045 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
2046 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
2048 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
2049 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
2050 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
2051 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
2053 2021-09-17 liuhongt <hongtao.liu@intel.com>
2055 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
2057 (_mm_cvttsh_u32): Likewise.
2058 (_mm_cvtt_roundsh_i32): Likewise.
2059 (_mm_cvtt_roundsh_u32): Likewise.
2060 (_mm_cvttsh_i64): Likewise.
2061 (_mm_cvttsh_u64): Likewise.
2062 (_mm_cvtt_roundsh_i64): Likewise.
2063 (_mm_cvtt_roundsh_u64): Likewise.
2064 * config/i386/i386-builtin.def: Add corresponding new builtins.
2065 * config/i386/sse.md
2066 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
2069 2021-09-17 liuhongt <hongtao.liu@intel.com>
2071 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
2073 (_mm512_mask_cvttph_epi32): Likewise.
2074 (_mm512_maskz_cvttph_epi32): Likewise.
2075 (_mm512_cvtt_roundph_epi32): Likewise.
2076 (_mm512_mask_cvtt_roundph_epi32): Likewise.
2077 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
2078 (_mm512_cvttph_epu32): Likewise.
2079 (_mm512_mask_cvttph_epu32): Likewise.
2080 (_mm512_maskz_cvttph_epu32): Likewise.
2081 (_mm512_cvtt_roundph_epu32): Likewise.
2082 (_mm512_mask_cvtt_roundph_epu32): Likewise.
2083 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
2084 (_mm512_cvttph_epi64): Likewise.
2085 (_mm512_mask_cvttph_epi64): Likewise.
2086 (_mm512_maskz_cvttph_epi64): Likewise.
2087 (_mm512_cvtt_roundph_epi64): Likewise.
2088 (_mm512_mask_cvtt_roundph_epi64): Likewise.
2089 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
2090 (_mm512_cvttph_epu64): Likewise.
2091 (_mm512_mask_cvttph_epu64): Likewise.
2092 (_mm512_maskz_cvttph_epu64): Likewise.
2093 (_mm512_cvtt_roundph_epu64): Likewise.
2094 (_mm512_mask_cvtt_roundph_epu64): Likewise.
2095 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
2096 (_mm512_cvttph_epi16): Likewise.
2097 (_mm512_mask_cvttph_epi16): Likewise.
2098 (_mm512_maskz_cvttph_epi16): Likewise.
2099 (_mm512_cvtt_roundph_epi16): Likewise.
2100 (_mm512_mask_cvtt_roundph_epi16): Likewise.
2101 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
2102 (_mm512_cvttph_epu16): Likewise.
2103 (_mm512_mask_cvttph_epu16): Likewise.
2104 (_mm512_maskz_cvttph_epu16): Likewise.
2105 (_mm512_cvtt_roundph_epu16): Likewise.
2106 (_mm512_mask_cvtt_roundph_epu16): Likewise.
2107 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
2108 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
2110 (_mm_mask_cvttph_epi32): Likewise.
2111 (_mm_maskz_cvttph_epi32): Likewise.
2112 (_mm256_cvttph_epi32): Likewise.
2113 (_mm256_mask_cvttph_epi32): Likewise.
2114 (_mm256_maskz_cvttph_epi32): Likewise.
2115 (_mm_cvttph_epu32): Likewise.
2116 (_mm_mask_cvttph_epu32): Likewise.
2117 (_mm_maskz_cvttph_epu32): Likewise.
2118 (_mm256_cvttph_epu32): Likewise.
2119 (_mm256_mask_cvttph_epu32): Likewise.
2120 (_mm256_maskz_cvttph_epu32): Likewise.
2121 (_mm_cvttph_epi64): Likewise.
2122 (_mm_mask_cvttph_epi64): Likewise.
2123 (_mm_maskz_cvttph_epi64): Likewise.
2124 (_mm256_cvttph_epi64): Likewise.
2125 (_mm256_mask_cvttph_epi64): Likewise.
2126 (_mm256_maskz_cvttph_epi64): Likewise.
2127 (_mm_cvttph_epu64): Likewise.
2128 (_mm_mask_cvttph_epu64): Likewise.
2129 (_mm_maskz_cvttph_epu64): Likewise.
2130 (_mm256_cvttph_epu64): Likewise.
2131 (_mm256_mask_cvttph_epu64): Likewise.
2132 (_mm256_maskz_cvttph_epu64): Likewise.
2133 (_mm_cvttph_epi16): Likewise.
2134 (_mm_mask_cvttph_epi16): Likewise.
2135 (_mm_maskz_cvttph_epi16): Likewise.
2136 (_mm256_cvttph_epi16): Likewise.
2137 (_mm256_mask_cvttph_epi16): Likewise.
2138 (_mm256_maskz_cvttph_epi16): Likewise.
2139 (_mm_cvttph_epu16): Likewise.
2140 (_mm_mask_cvttph_epu16): Likewise.
2141 (_mm_maskz_cvttph_epu16): Likewise.
2142 (_mm256_cvttph_epu16): Likewise.
2143 (_mm256_mask_cvttph_epu16): Likewise.
2144 (_mm256_maskz_cvttph_epu16): Likewise.
2145 * config/i386/i386-builtin.def: Add new builtins.
2146 * config/i386/sse.md
2147 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2149 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2150 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2151 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2152 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2154 2021-09-17 liuhongt <hongtao.liu@intel.com>
2156 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2157 (_mm_cvtsh_u32): Likewise.
2158 (_mm_cvt_roundsh_i32): Likewise.
2159 (_mm_cvt_roundsh_u32): Likewise.
2160 (_mm_cvtsh_i64): Likewise.
2161 (_mm_cvtsh_u64): Likewise.
2162 (_mm_cvt_roundsh_i64): Likewise.
2163 (_mm_cvt_roundsh_u64): Likewise.
2164 (_mm_cvti32_sh): Likewise.
2165 (_mm_cvtu32_sh): Likewise.
2166 (_mm_cvt_roundi32_sh): Likewise.
2167 (_mm_cvt_roundu32_sh): Likewise.
2168 (_mm_cvti64_sh): Likewise.
2169 (_mm_cvtu64_sh): Likewise.
2170 (_mm_cvt_roundi64_sh): Likewise.
2171 (_mm_cvt_roundu64_sh): Likewise.
2172 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2173 * config/i386/i386-builtin.def: Add corresponding new builtins.
2174 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2175 Handle new builtin types.
2176 * config/i386/sse.md
2177 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2179 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2180 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2182 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2184 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2185 (altivec_resolve_new_overloaded_builtin): New forward decl.
2186 (rs6000_new_builtin_type_compatible): New function.
2187 (altivec_resolve_overloaded_builtin): Call
2188 altivec_resolve_new_overloaded_builtin.
2189 (altivec_build_new_resolved_builtin): New function.
2190 (altivec_resolve_new_overloaded_builtin): Likewise.
2191 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2193 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2194 name of rs6000_new_builtin_is_supported.
2196 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2198 * config/i386/i386-protos.h (ix86_decompose_address):
2199 Change return type to bool.
2200 * config/i386/i386.c (ix86_decompose_address): Ditto.
2202 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2205 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2206 Added 'build/' to target, use build/%.o rule.
2207 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2208 '$(build_exeext)' to target and 'build/' for the *.o files.
2209 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2212 2021-09-16 Martin Jambor <mjambor@suse.cz>
2214 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2215 the body has been removed. Remove trailing whitespace.
2217 2021-09-16 Richard Biener <rguenther@suse.de>
2219 PR middle-end/102360
2220 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2221 of non-memory more robust.
2223 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2225 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2226 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2227 (leon5_adjust_cost): Increase cost of store with data dependency
2228 on ALU instruction and FPU anti-dependencies.
2229 (sparc_option_override): Add LEON5 costs
2230 (sparc_adjust_cost): Add LEON5 cost adjustments
2231 * config/sparc/sparc.h: Add LEON5
2232 * config/sparc/sparc.md: Include LEON5 scheduling information
2233 * config/sparc/sparc.opt: Add LEON5
2234 * doc/invoke.texi: Add LEON5
2235 * config/sparc/leon5.md: New file.
2237 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2239 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
2240 sensitive sequence for B2BST errata workaround.
2242 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2244 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
2245 functions with atomic instruction in the UT700 errata workaround.
2247 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2249 * config/sparc/sparc.c (next_active_non_empty_insn): New function
2250 that returns next active non empty assembly instruction.
2251 (sparc_do_work_around_errata): Use new function.
2253 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2255 * config/sparc/sparc.c (store_insn_p): Add predicate for store
2257 (load_insn_p): Add predicate for load attributes.
2258 (sparc_do_work_around_errata): Use new predicates.
2260 2021-09-16 Andreas Larsson <andreas@gaisler.com>
2262 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
2265 2021-09-16 Martin Liska <mliska@suse.cz>
2267 * config/mips/netbsd.h: Fix typo in name of a macro.
2269 2021-09-16 liuhongt <hongtao.liu@intel.com>
2271 PR middle-end/102080
2272 * match.pd: Check mask type when doing cond_op related gimple
2274 * tree.c (is_truth_type_for): New function.
2275 * tree.h (is_truth_type_for): New declaration.
2277 2021-09-16 liuhongt <hongtao.liu@intel.com>
2279 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2281 (_mm512_mask_cvtepi32_ph): Likewise.
2282 (_mm512_maskz_cvtepi32_ph): Likewise.
2283 (_mm512_cvt_roundepi32_ph): Likewise.
2284 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2285 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2286 (_mm512_cvtepu32_ph): Likewise.
2287 (_mm512_mask_cvtepu32_ph): Likewise.
2288 (_mm512_maskz_cvtepu32_ph): Likewise.
2289 (_mm512_cvt_roundepu32_ph): Likewise.
2290 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2291 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2292 (_mm512_cvtepi64_ph): Likewise.
2293 (_mm512_mask_cvtepi64_ph): Likewise.
2294 (_mm512_maskz_cvtepi64_ph): Likewise.
2295 (_mm512_cvt_roundepi64_ph): Likewise.
2296 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2297 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2298 (_mm512_cvtepu64_ph): Likewise.
2299 (_mm512_mask_cvtepu64_ph): Likewise.
2300 (_mm512_maskz_cvtepu64_ph): Likewise.
2301 (_mm512_cvt_roundepu64_ph): Likewise.
2302 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2303 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2304 (_mm512_cvtepi16_ph): Likewise.
2305 (_mm512_mask_cvtepi16_ph): Likewise.
2306 (_mm512_maskz_cvtepi16_ph): Likewise.
2307 (_mm512_cvt_roundepi16_ph): Likewise.
2308 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2309 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2310 (_mm512_cvtepu16_ph): Likewise.
2311 (_mm512_mask_cvtepu16_ph): Likewise.
2312 (_mm512_maskz_cvtepu16_ph): Likewise.
2313 (_mm512_cvt_roundepu16_ph): Likewise.
2314 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2315 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2316 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2318 (_mm_mask_cvtepi32_ph): Likewise.
2319 (_mm_maskz_cvtepi32_ph): Likewise.
2320 (_mm256_cvtepi32_ph): Likewise.
2321 (_mm256_mask_cvtepi32_ph): Likewise.
2322 (_mm256_maskz_cvtepi32_ph): Likewise.
2323 (_mm_cvtepu32_ph): Likewise.
2324 (_mm_mask_cvtepu32_ph): Likewise.
2325 (_mm_maskz_cvtepu32_ph): Likewise.
2326 (_mm256_cvtepu32_ph): Likewise.
2327 (_mm256_mask_cvtepu32_ph): Likewise.
2328 (_mm256_maskz_cvtepu32_ph): Likewise.
2329 (_mm_cvtepi64_ph): Likewise.
2330 (_mm_mask_cvtepi64_ph): Likewise.
2331 (_mm_maskz_cvtepi64_ph): Likewise.
2332 (_mm256_cvtepi64_ph): Likewise.
2333 (_mm256_mask_cvtepi64_ph): Likewise.
2334 (_mm256_maskz_cvtepi64_ph): Likewise.
2335 (_mm_cvtepu64_ph): Likewise.
2336 (_mm_mask_cvtepu64_ph): Likewise.
2337 (_mm_maskz_cvtepu64_ph): Likewise.
2338 (_mm256_cvtepu64_ph): Likewise.
2339 (_mm256_mask_cvtepu64_ph): Likewise.
2340 (_mm256_maskz_cvtepu64_ph): Likewise.
2341 (_mm_cvtepi16_ph): Likewise.
2342 (_mm_mask_cvtepi16_ph): Likewise.
2343 (_mm_maskz_cvtepi16_ph): Likewise.
2344 (_mm256_cvtepi16_ph): Likewise.
2345 (_mm256_mask_cvtepi16_ph): Likewise.
2346 (_mm256_maskz_cvtepi16_ph): Likewise.
2347 (_mm_cvtepu16_ph): Likewise.
2348 (_mm_mask_cvtepu16_ph): Likewise.
2349 (_mm_maskz_cvtepu16_ph): Likewise.
2350 (_mm256_cvtepu16_ph): Likewise.
2351 (_mm256_mask_cvtepu16_ph): Likewise.
2352 (_mm256_maskz_cvtepu16_ph): Likewise.
2353 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2354 * config/i386/i386-builtin.def: Add corresponding new builtins.
2355 * config/i386/i386-expand.c
2356 (ix86_expand_args_builtin): Handle new builtin types.
2357 (ix86_expand_round_builtin): Ditto.
2358 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2359 * config/i386/sse.md (VI2H_AVX512VL): New.
2361 (sseintvecmode): Add HF vector modes.
2362 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2364 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2365 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2366 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2367 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2368 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2369 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2370 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2371 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2372 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2373 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2374 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2376 2021-09-16 liuhongt <hongtao.liu@intel.com>
2378 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2380 (_mm512_mask_cvtph_epi32): Likewise.
2381 (_mm512_maskz_cvtph_epi32): Likewise.
2382 (_mm512_cvt_roundph_epi32): Likewise.
2383 (_mm512_mask_cvt_roundph_epi32): Likewise.
2384 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2385 (_mm512_cvtph_epu32): Likewise.
2386 (_mm512_mask_cvtph_epu32): Likewise.
2387 (_mm512_maskz_cvtph_epu32): Likewise.
2388 (_mm512_cvt_roundph_epu32): Likewise.
2389 (_mm512_mask_cvt_roundph_epu32): Likewise.
2390 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2391 (_mm512_cvtph_epi64): Likewise.
2392 (_mm512_mask_cvtph_epi64): Likewise.
2393 (_mm512_maskz_cvtph_epi64): Likewise.
2394 (_mm512_cvt_roundph_epi64): Likewise.
2395 (_mm512_mask_cvt_roundph_epi64): Likewise.
2396 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2397 (_mm512_cvtph_epu64): Likewise.
2398 (_mm512_mask_cvtph_epu64): Likewise.
2399 (_mm512_maskz_cvtph_epu64): Likewise.
2400 (_mm512_cvt_roundph_epu64): Likewise.
2401 (_mm512_mask_cvt_roundph_epu64): Likewise.
2402 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2403 (_mm512_cvtph_epi16): Likewise.
2404 (_mm512_mask_cvtph_epi16): Likewise.
2405 (_mm512_maskz_cvtph_epi16): Likewise.
2406 (_mm512_cvt_roundph_epi16): Likewise.
2407 (_mm512_mask_cvt_roundph_epi16): Likewise.
2408 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2409 (_mm512_cvtph_epu16): Likewise.
2410 (_mm512_mask_cvtph_epu16): Likewise.
2411 (_mm512_maskz_cvtph_epu16): Likewise.
2412 (_mm512_cvt_roundph_epu16): Likewise.
2413 (_mm512_mask_cvt_roundph_epu16): Likewise.
2414 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2415 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2417 (_mm_mask_cvtph_epi32): Likewise.
2418 (_mm_maskz_cvtph_epi32): Likewise.
2419 (_mm256_cvtph_epi32): Likewise.
2420 (_mm256_mask_cvtph_epi32): Likewise.
2421 (_mm256_maskz_cvtph_epi32): Likewise.
2422 (_mm_cvtph_epu32): Likewise.
2423 (_mm_mask_cvtph_epu32): Likewise.
2424 (_mm_maskz_cvtph_epu32): Likewise.
2425 (_mm256_cvtph_epu32): Likewise.
2426 (_mm256_mask_cvtph_epu32): Likewise.
2427 (_mm256_maskz_cvtph_epu32): Likewise.
2428 (_mm_cvtph_epi64): Likewise.
2429 (_mm_mask_cvtph_epi64): Likewise.
2430 (_mm_maskz_cvtph_epi64): Likewise.
2431 (_mm256_cvtph_epi64): Likewise.
2432 (_mm256_mask_cvtph_epi64): Likewise.
2433 (_mm256_maskz_cvtph_epi64): Likewise.
2434 (_mm_cvtph_epu64): Likewise.
2435 (_mm_mask_cvtph_epu64): Likewise.
2436 (_mm_maskz_cvtph_epu64): Likewise.
2437 (_mm256_cvtph_epu64): Likewise.
2438 (_mm256_mask_cvtph_epu64): Likewise.
2439 (_mm256_maskz_cvtph_epu64): Likewise.
2440 (_mm_cvtph_epi16): Likewise.
2441 (_mm_mask_cvtph_epi16): Likewise.
2442 (_mm_maskz_cvtph_epi16): Likewise.
2443 (_mm256_cvtph_epi16): Likewise.
2444 (_mm256_mask_cvtph_epi16): Likewise.
2445 (_mm256_maskz_cvtph_epi16): Likewise.
2446 (_mm_cvtph_epu16): Likewise.
2447 (_mm_mask_cvtph_epu16): Likewise.
2448 (_mm_maskz_cvtph_epu16): Likewise.
2449 (_mm256_cvtph_epu16): Likewise.
2450 (_mm256_mask_cvtph_epu16): Likewise.
2451 (_mm256_maskz_cvtph_epu16): Likewise.
2452 * config/i386/i386-builtin-types.def: Add new builtin types.
2453 * config/i386/i386-builtin.def: Add new builtins.
2454 * config/i386/i386-expand.c
2455 (ix86_expand_args_builtin): Handle new builtin types.
2456 (ix86_expand_round_builtin): Ditto.
2457 * config/i386/sse.md (sseintconvert): New.
2459 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2460 (sseintconvertsignprefix): Ditto.
2461 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2464 2021-09-16 liuhongt <hongtao.liu@intel.com>
2466 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2468 (_mm_cvtsi128_si16): Likewise.
2469 (_mm_mask_load_sh): Likewise.
2470 (_mm_maskz_load_sh): Likewise.
2471 (_mm_mask_store_sh): Likewise.
2472 (_mm_move_sh): Likewise.
2473 (_mm_mask_move_sh): Likewise.
2474 (_mm_maskz_move_sh): Likewise.
2475 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2476 * config/i386/i386-builtin.def: Add corresponding new builtins.
2477 * config/i386/i386-expand.c
2478 (ix86_expand_special_args_builtin): Handle new builtin types.
2479 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2480 * config/i386/sse.md (VI2F): New mode iterator.
2481 (vec_set<mode>_0): Use new mode iterator.
2482 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2483 (avx512f_store<mode>_mask): Ditto.
2485 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2487 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2489 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2491 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2492 Proceed if no symbol summary or the symbol alias flag is false.
2494 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2498 * varasm.c (output_constructor_regular_field): Instead of assertion
2499 that array_size_for_constructor result is equal to size of
2500 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2501 or equal to array_size_for_constructor result and use type size as
2504 2021-09-15 Martin Liska <mliska@suse.cz>
2507 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2509 2021-09-15 Martin Liska <mliska@suse.cz>
2512 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2513 Check that we have a symbol summary for a symbol.
2515 2021-09-15 Richard Biener <rguenther@suse.de>
2518 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2519 to inherit from elfos.h
2521 2021-09-15 liuhongt <hongtao.liu@intel.com>
2524 * config/i386/i386-expand.c
2525 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2527 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2528 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2529 AVX512FP16 which supports 16bit vector load.
2530 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2532 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2533 extend to V32HFmode.
2534 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2535 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2537 (vec_interleave_highv8hi<mask_name>): Rename to ..
2538 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2539 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2541 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2542 this, and extend to V32HFmode.
2543 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2544 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2545 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2546 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2547 (sse4_1_pblendw): Rename to ..
2548 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2549 (avx2_pblendph): New define_expand.
2550 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2551 sseintmodesuffix instead of ssemodesuffix.
2552 (blendsuf): New mode attr.
2554 2021-09-15 Richard Biener <rguenther@suse.de>
2556 * tree-vectorizer.h (dr_misalignment): Move out of line.
2557 (dr_target_alignment): New.
2558 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2559 (set_dr_target_alignment): New.
2560 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2561 * tree-vect-data-refs.c (dr_misalignment): Compute and
2562 return the group members misalignment.
2563 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2564 (vect_analyze_data_refs_alignment): Compute alignment only
2565 for the first element of a DR group.
2566 (vect_slp_analyze_node_alignment): Likewise.
2568 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2570 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2571 * config/i386/avx512fp16vlintrin.h: Likewise.
2572 * config/i386/i386-builtin.def: Adjust builtin name and
2573 enumeration to match AVX512F style.
2575 2021-09-15 Richard Biener <rguenther@suse.de>
2577 PR tree-optimization/102318
2578 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2579 previous change and do the mode conversion separately from
2580 the sign conversion.
2582 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2583 Peter Cordes <peter@cordes.ca>
2586 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2587 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2588 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2591 2021-09-15 Richard Biener <rguenther@suse.de>
2593 * config.gcc: Remove vax-*-openbsd* configuration.
2595 2021-09-15 Richard Biener <rguenther@suse.de>
2597 * config.gcc: Remove m68k-openbsd.
2599 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
2602 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
2604 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
2606 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
2607 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
2608 (*mma_xxsetaccz): Delete.
2609 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
2610 Use UNSPECV_MMA_XXSETACCZ. Update comment.
2611 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
2613 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
2615 * Makefile.in: Remove variables related to applying no-PIE
2616 to the exes on $build.
2617 * configure: Regenerate.
2618 * configure.ac: Remove configuration related to applying
2619 no-PIE to the exes on $build.
2621 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
2623 * config/arc/arc.md (doloop_end): Add missing mode.
2624 (loop_end): Likewise.
2626 2021-09-14 Jakub Jelinek <jakub@redhat.com>
2628 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
2629 it to recursive calls, for depth above 7 just gimplify or return.
2630 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
2631 __builtin_clear_padding and TARGET_EXPR.
2632 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
2634 2021-09-14 liuhongt <hongtao.liu@intel.com>
2636 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
2638 (_mm_mask_fpclass_sh_mask): Likewise.
2639 (_mm512_mask_fpclass_ph_mask): Likewise.
2640 (_mm512_fpclass_ph_mask): Likewise.
2641 (_mm_getexp_sh): Likewise.
2642 (_mm_mask_getexp_sh): Likewise.
2643 (_mm_maskz_getexp_sh): Likewise.
2644 (_mm512_getexp_ph): Likewise.
2645 (_mm512_mask_getexp_ph): Likewise.
2646 (_mm512_maskz_getexp_ph): Likewise.
2647 (_mm_getexp_round_sh): Likewise.
2648 (_mm_mask_getexp_round_sh): Likewise.
2649 (_mm_maskz_getexp_round_sh): Likewise.
2650 (_mm512_getexp_round_ph): Likewise.
2651 (_mm512_mask_getexp_round_ph): Likewise.
2652 (_mm512_maskz_getexp_round_ph): Likewise.
2653 (_mm_getmant_sh): Likewise.
2654 (_mm_mask_getmant_sh): Likewise.
2655 (_mm_maskz_getmant_sh): Likewise.
2656 (_mm512_getmant_ph): Likewise.
2657 (_mm512_mask_getmant_ph): Likewise.
2658 (_mm512_maskz_getmant_ph): Likewise.
2659 (_mm_getmant_round_sh): Likewise.
2660 (_mm_mask_getmant_round_sh): Likewise.
2661 (_mm_maskz_getmant_round_sh): Likewise.
2662 (_mm512_getmant_round_ph): Likewise.
2663 (_mm512_mask_getmant_round_ph): Likewise.
2664 (_mm512_maskz_getmant_round_ph): Likewise.
2665 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
2667 (_mm_fpclass_ph_mask): Likewise.
2668 (_mm256_mask_fpclass_ph_mask): Likewise.
2669 (_mm256_fpclass_ph_mask): Likewise.
2670 (_mm256_getexp_ph): Likewise.
2671 (_mm256_mask_getexp_ph): Likewise.
2672 (_mm256_maskz_getexp_ph): Likewise.
2673 (_mm_getexp_ph): Likewise.
2674 (_mm_mask_getexp_ph): Likewise.
2675 (_mm_maskz_getexp_ph): Likewise.
2676 (_mm256_getmant_ph): Likewise.
2677 (_mm256_mask_getmant_ph): Likewise.
2678 (_mm256_maskz_getmant_ph): Likewise.
2679 (_mm_getmant_ph): Likewise.
2680 (_mm_mask_getmant_ph): Likewise.
2681 (_mm_maskz_getmant_ph): Likewise.
2682 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2683 * config/i386/i386-builtin.def: Add corresponding new builtins.
2684 * config/i386/i386-expand.c
2685 (ix86_expand_args_builtin): Handle new builtin types.
2686 (ix86_expand_round_builtin): Ditto.
2687 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
2688 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
2689 to support HF vector modes.
2690 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
2692 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
2693 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
2694 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2695 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2698 2021-09-14 liuhongt <hongtao.liu@intel.com>
2700 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
2702 (_mm512_mask_reduce_ph): Likewise.
2703 (_mm512_maskz_reduce_ph): Likewise.
2704 (_mm512_reduce_round_ph): Likewise.
2705 (_mm512_mask_reduce_round_ph): Likewise.
2706 (_mm512_maskz_reduce_round_ph): Likewise.
2707 (_mm_reduce_sh): Likewise.
2708 (_mm_mask_reduce_sh): Likewise.
2709 (_mm_maskz_reduce_sh): Likewise.
2710 (_mm_reduce_round_sh): Likewise.
2711 (_mm_mask_reduce_round_sh): Likewise.
2712 (_mm_maskz_reduce_round_sh): Likewise.
2713 (_mm512_roundscale_ph): Likewise.
2714 (_mm512_mask_roundscale_ph): Likewise.
2715 (_mm512_maskz_roundscale_ph): Likewise.
2716 (_mm512_roundscale_round_ph): Likewise.
2717 (_mm512_mask_roundscale_round_ph): Likewise.
2718 (_mm512_maskz_roundscale_round_ph): Likewise.
2719 (_mm_roundscale_sh): Likewise.
2720 (_mm_mask_roundscale_sh): Likewise.
2721 (_mm_maskz_roundscale_sh): Likewise.
2722 (_mm_roundscale_round_sh): Likewise.
2723 (_mm_mask_roundscale_round_sh): Likewise.
2724 (_mm_maskz_roundscale_round_sh): Likewise.
2725 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
2727 (_mm_mask_reduce_ph): Likewise.
2728 (_mm_maskz_reduce_ph): Likewise.
2729 (_mm256_reduce_ph): Likewise.
2730 (_mm256_mask_reduce_ph): Likewise.
2731 (_mm256_maskz_reduce_ph): Likewise.
2732 (_mm_roundscale_ph): Likewise.
2733 (_mm_mask_roundscale_ph): Likewise.
2734 (_mm_maskz_roundscale_ph): Likewise.
2735 (_mm256_roundscale_ph): Likewise.
2736 (_mm256_mask_roundscale_ph): Likewise.
2737 (_mm256_maskz_roundscale_ph): Likewise.
2738 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2739 * config/i386/i386-builtin.def: Add corresponding new builtins.
2740 * config/i386/i386-expand.c
2741 (ix86_expand_args_builtin): Handle new builtin types.
2742 (ix86_expand_round_builtin): Ditto.
2743 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
2745 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
2746 ... this, and adjust for round operands.
2747 (reduces<mode><mask_scalar_name>): Likewise, with ...
2748 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
2750 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
2751 Adjust for HF vector modes.
2752 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
2754 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2756 2021-09-14 liuhongt <hongtao.liu@intel.com>
2758 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
2760 (_mm512_mask_rcp_ph): Likewise.
2761 (_mm512_maskz_rcp_ph): Likewise.
2762 (_mm_rcp_sh): Likewise.
2763 (_mm_mask_rcp_sh): Likewise.
2764 (_mm_maskz_rcp_sh): Likewise.
2765 (_mm512_scalef_ph): Likewise.
2766 (_mm512_mask_scalef_ph): Likewise.
2767 (_mm512_maskz_scalef_ph): Likewise.
2768 (_mm512_scalef_round_ph): Likewise.
2769 (_mm512_mask_scalef_round_ph): Likewise.
2770 (_mm512_maskz_scalef_round_ph): Likewise.
2771 (_mm_scalef_sh): Likewise.
2772 (_mm_mask_scalef_sh): Likewise.
2773 (_mm_maskz_scalef_sh): Likewise.
2774 (_mm_scalef_round_sh): Likewise.
2775 (_mm_mask_scalef_round_sh): Likewise.
2776 (_mm_maskz_scalef_round_sh): Likewise.
2777 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
2779 (_mm256_rcp_ph): Likewise.
2780 (_mm_mask_rcp_ph): Likewise.
2781 (_mm256_mask_rcp_ph): Likewise.
2782 (_mm_maskz_rcp_ph): Likewise.
2783 (_mm256_maskz_rcp_ph): Likewise.
2784 (_mm_scalef_ph): Likewise.
2785 (_mm256_scalef_ph): Likewise.
2786 (_mm_mask_scalef_ph): Likewise.
2787 (_mm256_mask_scalef_ph): Likewise.
2788 (_mm_maskz_scalef_ph): Likewise.
2789 (_mm256_maskz_scalef_ph): Likewise.
2790 * config/i386/i386-builtin.def: Add new builtins.
2791 * config/i386/sse.md (VFH_AVX512VL): New.
2792 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
2793 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
2794 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
2795 Adjust to support HF vector modes.
2796 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
2798 2021-09-14 liuhongt <hongtao.liu@intel.com>
2800 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
2802 (_mm512_mask_sqrt_ph): Likewise.
2803 (_mm512_maskz_sqrt_ph): Likewise.
2804 (_mm512_sqrt_round_ph): Likewise.
2805 (_mm512_mask_sqrt_round_ph): Likewise.
2806 (_mm512_maskz_sqrt_round_ph): Likewise.
2807 (_mm512_rsqrt_ph): Likewise.
2808 (_mm512_mask_rsqrt_ph): Likewise.
2809 (_mm512_maskz_rsqrt_ph): Likewise.
2810 (_mm_rsqrt_sh): Likewise.
2811 (_mm_mask_rsqrt_sh): Likewise.
2812 (_mm_maskz_rsqrt_sh): Likewise.
2813 (_mm_sqrt_sh): Likewise.
2814 (_mm_mask_sqrt_sh): Likewise.
2815 (_mm_maskz_sqrt_sh): Likewise.
2816 (_mm_sqrt_round_sh): Likewise.
2817 (_mm_mask_sqrt_round_sh): Likewise.
2818 (_mm_maskz_sqrt_round_sh): Likewise.
2819 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
2820 (_mm256_sqrt_ph): Likewise.
2821 (_mm_mask_sqrt_ph): Likewise.
2822 (_mm256_mask_sqrt_ph): Likewise.
2823 (_mm_maskz_sqrt_ph): Likewise.
2824 (_mm256_maskz_sqrt_ph): Likewise.
2825 (_mm_rsqrt_ph): Likewise.
2826 (_mm256_rsqrt_ph): Likewise.
2827 (_mm_mask_rsqrt_ph): Likewise.
2828 (_mm256_mask_rsqrt_ph): Likewise.
2829 (_mm_maskz_rsqrt_ph): Likewise.
2830 (_mm256_maskz_rsqrt_ph): Likewise.
2831 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2832 * config/i386/i386-builtin.def: Add corresponding new builtins.
2833 * config/i386/i386-expand.c
2834 (ix86_expand_args_builtin): Handle new builtins.
2835 (ix86_expand_round_builtin): Ditto.
2836 * config/i386/sse.md (VF_AVX512FP16VL): New.
2837 (sqrt<mode>2): Adjust for HF vector modes.
2838 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
2839 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2841 (<sse>_rsqrt<mode>2<mask_name>): New.
2842 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
2844 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2847 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
2848 'RESERVED_LOCATION_P' locations.
2849 * warning-control.cc (get_nowarn_spec, suppress_warning)
2850 (copy_warning): Likewise.
2852 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2854 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
2855 (typedef key_type_t): ... this. Remove.
2856 (nowarn_map): Document.
2857 * diagnostic-spec.c (nowarn_map): Likewise.
2858 * warning-control.cc (convert_to_key): Evolve functions into...
2859 (get_location): ... these. Adjust all users.
2861 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2863 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
2865 2021-09-13 Jason Merrill <jason@redhat.com>
2867 * params.opt: Add destructive-interference-size and
2868 constructive-interference-size.
2869 * doc/invoke.texi: Document them.
2870 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2872 * config/arm/arm.c (arm_option_override): Set them.
2873 * config/i386/i386-options.c (ix86_option_override_internal):
2876 2021-09-13 Martin Liska <mliska@suse.cz>
2877 H.J. Lu <hjl.tools@gmail.com>
2880 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
2881 for x86-64 micro levels for __builtin_cpu_supports.
2882 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
2883 Add priorities for the micro-arch levels.
2884 (enum processor_features): Add new features.
2885 * common/config/i386/i386-isas.h: Add micro-arch features.
2886 * config/i386/i386-builtins.c (get_builtin_code_for_version):
2887 Support the micro-arch levels by callsing
2888 __builtin_cpu_supports.
2889 * doc/extend.texi: Document that the levels are support by
2890 __builtin_cpu_supports.
2892 2021-09-13 Andrew Pinski <apinski@marvell.com>
2895 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
2897 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
2898 (aarch64_general_gimple_fold_builtin): Likewise.
2900 2021-09-13 Andrew Pinski <apinski@marvell.com>
2902 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
2903 to the Unsupported targets list.
2904 Remove support for m32r-*-linux* and m32rle-*-linux*.
2905 * config/m32r/linux.h: Removed.
2906 * config/m32r/t-linux: Removed.
2908 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2911 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
2912 register index for SVE predicate modes.
2914 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
2916 * tree-ssa-threadbackward.c
2917 (back_threader_profitability::profitable_path_p): Remove FSM
2919 (back_threader_registry::register_path): Same.
2920 * tree-ssa-threadedge.c
2921 (jump_threader::simplify_control_stmt_condition): Same.
2922 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
2923 Add backedge_threads argument.
2924 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
2925 backedge_threads argument.
2926 (back_jt_path_registry::back_jt_path_registry): Same.
2927 (dump_jump_thread_path): Adjust for FSM removal.
2928 (back_jt_path_registry::rewire_first_differing_edge): Same.
2929 (back_jt_path_registry::adjust_paths_after_duplication): Same.
2930 (back_jt_path_registry::update_cfg): Same.
2931 (jt_path_registry::register_jump_thread): Same.
2932 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
2934 (class back_jt_path_registry): Add backedge_threads to
2937 2021-09-13 Martin Liska <mliska@suse.cz>
2940 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
2942 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
2945 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
2946 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
2947 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
2948 TARGET_AVX256_STORE_BY_PIECES instead of
2949 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
2950 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
2951 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
2952 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
2953 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
2954 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
2956 2021-09-13 liuhongt <hongtao.liu@intel.com>
2959 * expmed.c (extract_bit_field_using_extv): Use
2960 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
2962 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
2964 * Makefile.in (OBJS): Add value-pointer-equiv.o.
2965 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
2966 value-pointer-equiv.*.
2967 (ssa_equiv_stack::ssa_equiv_stack): Same.
2968 (ssa_equiv_stack::enter): Same.
2969 (ssa_equiv_stack::leave): Same.
2970 (ssa_equiv_stack::push_replacement): Same.
2971 (ssa_equiv_stack::get_replacement): Same.
2972 (is_pointer_ssa): Same.
2973 (class pointer_equiv_analyzer): Same.
2974 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
2975 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
2976 (pointer_equiv_analyzer::set_global_equiv): Same.
2977 (pointer_equiv_analyzer::set_cond_equiv): Same.
2978 (pointer_equiv_analyzer::get_equiv): Same.
2979 (pointer_equiv_analyzer::enter): Same.
2980 (pointer_equiv_analyzer::leave): Same.
2981 (pointer_equiv_analyzer::get_equiv_expr): Same.
2982 (pta_valueize): Same.
2983 (pointer_equiv_analyzer::visit_stmt): Same.
2984 (pointer_equiv_analyzer::visit_edge): Same.
2985 (hybrid_folder::value_of_expr): Same.
2986 (hybrid_folder::value_on_edge): Same.
2987 * value-pointer-equiv.cc: New file.
2988 * value-pointer-equiv.h: New file.
2990 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2993 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
2994 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
2996 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2999 * config/arm/arm.md (movmisaligndi): New define_expand.
3000 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
3002 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3005 * emit-rtl.c (gen_highpart): Use adjust_address to handle
3006 MEM rather than calling simplify_gen_subreg.
3008 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
3010 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
3011 define into a block.
3013 2021-09-13 Richard Biener <rguenther@suse.de>
3015 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
3016 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
3017 change based on DARWIN_PREFER_DWARF not being defined.
3019 2021-09-13 Richard Biener <rguenther@suse.de>
3021 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3022 to inherit from elfos.h
3024 2021-09-13 Richard Biener <rguenther@suse.de>
3026 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
3028 2021-09-13 Richard Biener <rguenther@suse.de>
3030 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
3031 override, pick up DWARF2_DEBUG define from elfos.h
3033 2021-09-13 Richard Biener <rguenther@suse.de>
3035 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
3038 2021-09-13 Richard Biener <rguenther@suse.de>
3040 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
3043 2021-09-13 Richard Biener <rguenther@suse.de>
3045 * config/i386/cygming.h: Always default to DWARF2 debugging.
3046 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
3048 * doc/install.texi: Document binutils 2.16 as minimum
3049 requirement for mingw.
3051 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3053 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
3054 nstmts, nloads and extra_ctor_cost.
3055 (rs6000_density_test): Add load density related heuristics. Do
3056 extra costing on vector construction statements if need.
3057 (rs6000_init_cost): Init new members.
3058 (rs6000_update_target_cost_per_stmt): New function.
3059 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
3060 rs6000_update_target_cost_per_stmt and call it.
3062 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3064 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
3065 (rs6000_init_cost): Adjust.
3067 2021-09-13 liuhongt <hongtao.liu@intel.com>
3069 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
3070 (UNSPEC_XORSIGN): Ditto.
3072 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
3074 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
3075 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
3078 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
3080 * tree-ssa-threadbackward.c (class back_threader_registry): Use
3081 back_jt_path_registry.
3082 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
3083 fwd_jt_path_registry.
3084 * tree-ssa-threadedge.h (class jump_threader): Same..
3085 * tree-ssa-threadupdate.c
3086 (jump_thread_path_registry::jump_thread_path_registry): Rename...
3087 (jt_path_registry::jt_path_registry): ...to this.
3088 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
3089 (jt_path_registry::~jt_path_registry): ...this.
3090 (fwd_jt_path_registry::fwd_jt_path_registry): New.
3091 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
3092 (jump_thread_path_registry::allocate_thread_edge): Rename...
3093 (jt_path_registry::allocate_thread_edge): ...to this.
3094 (jump_thread_path_registry::allocate_thread_path): Rename...
3095 (jt_path_registry::allocate_thread_path): ...to this.
3096 (jump_thread_path_registry::lookup_redirection_data): Rename...
3097 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
3098 (jump_thread_path_registry::thread_block_1): Rename...
3099 (fwd_jt_path_registry::thread_block_1): ...to this.
3100 (jump_thread_path_registry::thread_block): Rename...
3101 (fwd_jt_path_registry::thread_block): ...to this.
3102 (jt_path_registry::thread_through_loop_header): Rename...
3103 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
3104 (jump_thread_path_registry::mark_threaded_blocks): Rename...
3105 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
3106 (jump_thread_path_registry::debug_path): Rename...
3107 (jt_path_registry::debug_path): ...to this.
3108 (jump_thread_path_registry::dump): Rename...
3109 (jt_path_registry::debug): ...to this.
3110 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
3111 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
3112 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
3113 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
3114 (jump_thread_path_registry::duplicate_thread_path): Rename...
3115 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
3116 drop ill-formed candidates.
3117 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3118 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3119 (jt_path_registry::thread_through_all_blocks): New.
3120 (back_jt_path_registry::update_cfg): New.
3121 (fwd_jt_path_registry::update_cfg): New.
3122 (jump_thread_path_registry::register_jump_thread): Rename...
3123 (jt_path_registry::register_jump_thread): ...to this.
3124 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3126 (class jt_path_registry): ...here.
3127 (class fwd_jt_path_registry): New.
3128 (class back_jt_path_registry): New.
3130 2021-09-10 liuhongt <hongtao.liu@intel.com>
3133 2021-09-01 liuhongt <hongtao.liu@intel.com>
3135 * emit-rtl.c (validate_subreg): Get rid of all float-int
3138 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3140 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3141 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3142 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3143 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3144 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3145 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3146 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3147 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3149 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3150 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3152 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3154 * tree.h (OMP_ATOMIC_WEAK): Define.
3155 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3156 fail memory order being encoded in the same enum and also print
3157 fail clause if present.
3158 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3159 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3160 handle pre_p == NULL case as a test mode that only returns value
3161 but doesn't change gimplify nor change anything otherwise, adjust
3162 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3163 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3164 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3166 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3167 COND_EXPR rhs. Set weak flag on gimple load/store for
3169 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3170 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3172 (expand_omp_atomic_cas): New function.
3173 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3175 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3176 using expand_omp_atomic_cas.
3178 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3179 Michael Matz <matz@suse.de>
3181 * tree-pass.h (PROP_loop_opts_done): New.
3182 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3183 Intersect with global range.
3184 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3185 * tree-ssa-threadbackward.c
3186 (back_threader_profitability::profitable_path_p): Disable
3187 threading through latches until after loop optimizations have run.
3189 2021-09-10 David Faust <david.faust@oracle.com>
3191 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3194 2021-09-10 David Faust <david.faust@oracle.com>
3196 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3197 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3198 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3199 (bpf_expand_cbranch): New function.
3200 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3202 (zero_extendsidi2): Only use mov32 instruction if it is available.
3203 (SIM mode iterator): Conditionalize support for SI mode.
3204 (JM mode iterator): New.
3205 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3206 (*branch_on_di): Update name, use new JM iterator.
3207 * config/bpf/bpf.opt: (mjmpext): New option.
3211 (bpf_isa): New enum.
3213 2021-09-10 David Faust <david.faust@oracle.com>
3215 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3216 for register-to-register extensions.
3217 (zero_extendqidi2): Likewise.
3219 2021-09-10 Richard Biener <rguenther@suse.de>
3221 PR middle-end/102273
3222 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3224 2021-09-10 Richard Biener <rguenther@suse.de>
3226 PR middle-end/102269
3227 * gimplify.c (is_var_need_auto_init): Empty types do not need
3230 2021-09-10 Richard Biener <rguenther@suse.de>
3232 * configure.ac (--with-stabs): Remove.
3233 * configure: Regenerate.
3234 * doc/install.texi: Remove --with-stabs documentation.
3236 2021-09-10 liuhongt <hongtao.liu@intel.com>
3238 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
3240 (_mm512_mask_cmp_ph_mask): Likewise.
3241 (_mm512_cmp_round_ph_mask): Likewise.
3242 (_mm512_mask_cmp_round_ph_mask): Likewise.
3243 (_mm_cmp_sh_mask): Likewise.
3244 (_mm_mask_cmp_sh_mask): Likewise.
3245 (_mm_cmp_round_sh_mask): Likewise.
3246 (_mm_mask_cmp_round_sh_mask): Likewise.
3247 (_mm_comieq_sh): Likewise.
3248 (_mm_comilt_sh): Likewise.
3249 (_mm_comile_sh): Likewise.
3250 (_mm_comigt_sh): Likewise.
3251 (_mm_comige_sh): Likewise.
3252 (_mm_comineq_sh): Likewise.
3253 (_mm_ucomieq_sh): Likewise.
3254 (_mm_ucomilt_sh): Likewise.
3255 (_mm_ucomile_sh): Likewise.
3256 (_mm_ucomigt_sh): Likewise.
3257 (_mm_ucomige_sh): Likewise.
3258 (_mm_ucomineq_sh): Likewise.
3259 (_mm_comi_round_sh): Likewise.
3260 (_mm_comi_sh): Likewise.
3261 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
3262 (_mm_mask_cmp_ph_mask): Likewise.
3263 (_mm256_cmp_ph_mask): Likewise.
3264 (_mm256_mask_cmp_ph_mask): Likewise.
3265 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3266 * config/i386/i386-builtin.def: Add corresponding new builtins.
3267 * config/i386/i386-expand.c
3268 (ix86_expand_args_builtin): Handle new builtin types.
3269 (ix86_expand_round_builtin): Ditto.
3270 * config/i386/i386.md (ssevecmode): Add HF mode.
3271 (MODEFH): New mode iterator.
3272 * config/i386/sse.md
3273 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3274 Ajdust corresponding description.
3275 (ssecmpintprefix): New.
3276 (VI12_AVX512VL): Adjust to support HF vector modes.
3277 (cmp_imm_predicate): Likewise.
3278 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3280 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3281 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3282 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3284 2021-09-10 liuhongt <hongtao.liu@intel.com>
3286 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3287 (_mm512_mask_max_ph): Likewise.
3288 (_mm512_maskz_max_ph): Likewise.
3289 (_mm512_min_ph): Likewise.
3290 (_mm512_mask_min_ph): Likewise.
3291 (_mm512_maskz_min_ph): Likewise.
3292 (_mm512_max_round_ph): Likewise.
3293 (_mm512_mask_max_round_ph): Likewise.
3294 (_mm512_maskz_max_round_ph): Likewise.
3295 (_mm512_min_round_ph): Likewise.
3296 (_mm512_mask_min_round_ph): Likewise.
3297 (_mm512_maskz_min_round_ph): Likewise.
3298 (_mm_max_sh): Likewise.
3299 (_mm_mask_max_sh): Likewise.
3300 (_mm_maskz_max_sh): Likewise.
3301 (_mm_min_sh): Likewise.
3302 (_mm_mask_min_sh): Likewise.
3303 (_mm_maskz_min_sh): Likewise.
3304 (_mm_max_round_sh): Likewise.
3305 (_mm_mask_max_round_sh): Likewise.
3306 (_mm_maskz_max_round_sh): Likewise.
3307 (_mm_min_round_sh): Likewise.
3308 (_mm_mask_min_round_sh): Likewise.
3309 (_mm_maskz_min_round_sh): Likewise.
3310 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3311 (_mm256_max_ph): Likewise.
3312 (_mm_mask_max_ph): Likewise.
3313 (_mm256_mask_max_ph): Likewise.
3314 (_mm_maskz_max_ph): Likewise.
3315 (_mm256_maskz_max_ph): Likewise.
3316 (_mm_min_ph): Likewise.
3317 (_mm256_min_ph): Likewise.
3318 (_mm_mask_min_ph): Likewise.
3319 (_mm256_mask_min_ph): Likewise.
3320 (_mm_maskz_min_ph): Likewise.
3321 (_mm256_maskz_min_ph): Likewise.
3322 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3323 * config/i386/i386-builtin.def: Add corresponding new builtins.
3324 * config/i386/i386-expand.c
3325 (ix86_expand_args_builtin): Handle new builtin types.
3326 * config/i386/sse.md
3327 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3328 support HF vector modes.
3329 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3330 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3332 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3334 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3335 Adjust for HF vector modes.
3337 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3339 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3340 (_mm_mask_add_sh): Likewise.
3341 (_mm_maskz_add_sh): Likewise.
3342 (_mm_sub_sh): Likewise.
3343 (_mm_mask_sub_sh): Likewise.
3344 (_mm_maskz_sub_sh): Likewise.
3345 (_mm_mul_sh): Likewise.
3346 (_mm_mask_mul_sh): Likewise.
3347 (_mm_maskz_mul_sh): Likewise.
3348 (_mm_div_sh): Likewise.
3349 (_mm_mask_div_sh): Likewise.
3350 (_mm_maskz_div_sh): Likewise.
3351 (_mm_add_round_sh): Likewise.
3352 (_mm_mask_add_round_sh): Likewise.
3353 (_mm_maskz_add_round_sh): Likewise.
3354 (_mm_sub_round_sh): Likewise.
3355 (_mm_mask_sub_round_sh): Likewise.
3356 (_mm_maskz_sub_round_sh): Likewise.
3357 (_mm_mul_round_sh): Likewise.
3358 (_mm_mask_mul_round_sh): Likewise.
3359 (_mm_maskz_mul_round_sh): Likewise.
3360 (_mm_div_round_sh): Likewise.
3361 (_mm_mask_div_round_sh): Likewise.
3362 (_mm_maskz_div_round_sh): Likewise.
3363 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3364 * config/i386/i386-builtin.def: Add corresponding new builtins.
3365 * config/i386/i386-expand.c
3366 (ix86_expand_round_builtin): Handle new builtins.
3367 * config/i386/sse.md (VF_128): Change description.
3368 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3369 Adjust to support HF vector modes.
3370 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3373 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3375 * config/i386/i386-expand.c
3376 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3377 * config/i386/i386.c
3378 (ix86_preferred_simd_mode): Handle HF mode.
3379 * config/i386/sse.md (V_256H): New mode iterator.
3380 (avx_vextractf128<mode>): Use it.
3381 (VEC_INIT_MODE): Align vector HFmode condition to vector
3382 HImodes since there're no real HF instruction used.
3383 (VEC_INIT_HALF_MODE): Ditto.
3385 (VIHF_AVX512BW): Ditto.
3386 (*vec_extracthf): Ditto.
3387 (VEC_EXTRACT_MODE): Ditto.
3389 2021-09-10 Richard Biener <rguenther@suse.de>
3392 * config/dbx.h: Remove.
3393 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3394 * config/lynx.h: Likewise.
3396 2021-09-10 liuhongt <hongtao.liu@intel.com>
3398 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3399 right into ANDNOT + AND + IOR, using paradoxical subregs.
3400 (ix86_split_copysign_const): Remove.
3401 (ix86_split_copysign_var): Ditto.
3402 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3403 (ix86_split_copysign_var): Ditto.
3404 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3405 (@copysign<mode>3_var): Ditto.
3407 2021-09-09 qing zhao <qing.zhao@oracle.com>
3409 * builtins.c (expand_builtin_memset): Make external visible.
3410 * builtins.h (expand_builtin_memset): Declare extern.
3411 * common.opt (ftrivial-auto-var-init=): New option.
3412 * doc/extend.texi: Document the uninitialized attribute.
3413 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3414 * flag-types.h (enum auto_init_type): New enumerated type
3416 * gimple-fold.c (clear_padding_type): Add one new parameter.
3417 (clear_padding_union): Likewise.
3418 (clear_padding_emit_loop): Likewise.
3419 (clear_type_padding_in_mask): Likewise.
3420 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3421 * gimplify.c (gimple_add_init_for_auto_var): New function.
3422 (gimple_add_padding_init_for_auto_var): New function.
3423 (is_var_need_auto_init): New function.
3424 (gimplify_decl_expr): Add initialization to automatic variables per
3426 (gimplify_call_expr): Add one new parameter for call to
3427 __builtin_clear_padding.
3428 (gimplify_init_constructor): Add padding initialization in the end.
3429 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3430 (expand_DEFERRED_INIT): New function.
3431 * internal-fn.def (DEFERRED_INIT): New internal function.
3432 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3433 * tree-sra.c (generate_subtree_deferred_init): New function.
3434 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3435 calls to .DEFERRED_INIT.
3436 (sra_modify_deferred_init): New function.
3437 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3438 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3439 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3441 (check_defs): Likewise.
3442 (warn_uninitialized_vars): Likewise.
3443 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3444 * tree.c (build_common_builtin_nodes): Build tree node for
3445 BUILT_IN_CLEAR_PADDING when needed.
3447 2021-09-09 Richard Biener <rguenther@suse.de>
3449 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3452 2021-09-09 Richard Biener <rguenther@suse.de>
3454 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3455 DOM walk from get_loop_body_in_dom_order using a worklist
3458 2021-09-09 liuhongt <hongtao.liu@intel.com>
3460 * config.gcc: Add avx512fp16vlintrin.h.
3461 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3462 (_mm512_mask_add_ph): Likewise.
3463 (_mm512_maskz_add_ph): Likewise.
3464 (_mm512_sub_ph): Likewise.
3465 (_mm512_mask_sub_ph): Likewise.
3466 (_mm512_maskz_sub_ph): Likewise.
3467 (_mm512_mul_ph): Likewise.
3468 (_mm512_mask_mul_ph): Likewise.
3469 (_mm512_maskz_mul_ph): Likewise.
3470 (_mm512_div_ph): Likewise.
3471 (_mm512_mask_div_ph): Likewise.
3472 (_mm512_maskz_div_ph): Likewise.
3473 (_mm512_add_round_ph): Likewise.
3474 (_mm512_mask_add_round_ph): Likewise.
3475 (_mm512_maskz_add_round_ph): Likewise.
3476 (_mm512_sub_round_ph): Likewise.
3477 (_mm512_mask_sub_round_ph): Likewise.
3478 (_mm512_maskz_sub_round_ph): Likewise.
3479 (_mm512_mul_round_ph): Likewise.
3480 (_mm512_mask_mul_round_ph): Likewise.
3481 (_mm512_maskz_mul_round_ph): Likewise.
3482 (_mm512_div_round_ph): Likewise.
3483 (_mm512_mask_div_round_ph): Likewise.
3484 (_mm512_maskz_div_round_ph): Likewise.
3485 * config/i386/avx512fp16vlintrin.h: New header.
3486 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3487 Add new builtin types.
3488 * config/i386/i386-builtin.def: Add corresponding builtins.
3489 * config/i386/i386-expand.c
3490 (ix86_expand_args_builtin): Handle new builtin types.
3491 (ix86_expand_round_builtin): Likewise.
3492 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3493 * config/i386/sse.md (VFH): New mode_iterator.
3495 (avx512fmaskmode): Add HF vector modes.
3496 (avx512fmaskhalfmode): Likewise.
3497 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3499 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3500 (mul<mode>3<mask_name><round_name>): Likewise.
3501 (*mul<mode>3<mask_name><round_name>): Likewise.
3502 (div<mode>3): Likewise.
3503 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3504 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3505 (SUBST_A): Likewise.
3506 (round_mode512bit_condition): Adjust for V32HFmode.
3508 2021-09-09 liuhongt <hongtao.liu@intel.com>
3511 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3512 (reduc_plus_scal_v4sf): .. this, New define_expand.
3513 (reduc_plus_scal_v2df): .. and this, New define_expand.
3515 2021-09-09 liuhongt <hongtao.liu@intel.com>
3518 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3521 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3524 * doc/trouble.texi (Copy Assignment): Fix description of
3525 behaviour and fix code in example.
3527 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3530 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3531 r11 instead of r12 for restoring CR.
3533 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3534 liuhongt <hongtao.liu@intel.com>
3537 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3538 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3539 into AND with mask and XOR, using paradoxical subregs.
3540 (ix86_split_xorsign): Remove.
3541 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3543 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3545 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3547 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3550 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3551 operands[2], emit abs<mode>2 instead.
3552 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3553 first alternative even for avx, add another alternative with
3554 =&Yv <- 0, Yv, Yvm constraints.
3555 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3556 to op1, emit vpandn instead.
3558 2021-09-08 liuhongt <hongtao.liu@intel.com>
3560 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3561 (_mm256_set_ph): Likewise.
3562 (_mm512_set_ph): Likewise.
3563 (_mm_setr_ph): Likewise.
3564 (_mm256_setr_ph): Likewise.
3565 (_mm512_setr_ph): Likewise.
3566 (_mm_set1_ph): Likewise.
3567 (_mm256_set1_ph): Likewise.
3568 (_mm512_set1_ph): Likewise.
3569 (_mm_setzero_ph): Likewise.
3570 (_mm256_setzero_ph): Likewise.
3571 (_mm512_setzero_ph): Likewise.
3572 (_mm_set_sh): Likewise.
3573 (_mm_load_sh): Likewise.
3574 (_mm_store_sh): Likewise.
3575 * config/i386/i386-builtin-types.def (V8HF): New type.
3576 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3577 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3578 Support vector HFmodes.
3579 (ix86_expand_vector_init_one_nonzero): Likewise.
3580 (ix86_expand_vector_init_one_var): Likewise.
3581 (ix86_expand_vector_init_interleave): Likewise.
3582 (ix86_expand_vector_init_general): Likewise.
3583 (ix86_expand_vector_set): Likewise.
3584 (ix86_expand_vector_extract): Likewise.
3585 (ix86_expand_vector_init_concat): Likewise.
3586 (ix86_expand_sse_movcc): Handle vector HFmodes.
3587 (ix86_expand_vector_set_var): Ditto.
3588 * config/i386/i386-modes.def: Add HF vector modes in comment.
3589 * config/i386/i386.c (classify_argument): Add HF vector modes.
3590 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3591 (ix86_vector_mode_supported_p): Likewise.
3592 (ix86_set_reg_reg_cost): Handle vector HFmode.
3593 (ix86_get_ssemov): Handle vector HFmode.
3594 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3596 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3598 (function_arg_advance_32): Ditto.
3599 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
3600 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
3601 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
3602 (VALID_SSE2_REG_VHF_MODE): New.
3603 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
3604 (SSE_REG_MODE_P): Add vector HFmode.
3605 * config/i386/i386.md (mode): Add HF vector modes.
3606 (MODE_SIZE): Likewise.
3607 (ssemodesuffix): Add ph suffix for HF vector modes.
3608 * config/i386/sse.md (VFH_128): New mode iterator.
3609 (VMOVE): Adjust for HF vector modes.
3611 (V_256_512): Likewise.
3613 (avx512fmaskmode): Likewise.
3614 (shuffletype): Likewise.
3615 (sseinsnmode): Likewise.
3616 (ssedoublevecmode): Likewise.
3617 (ssehalfvecmode): Likewise.
3618 (ssehalfvecmodelower): Likewise.
3619 (ssePScmode): Likewise.
3620 (ssescalarmode): Likewise.
3621 (ssescalarmodelower): Likewise.
3622 (sseintprefix): Likewise.
3624 (bcstscalarsuff): Likewise.
3625 (xtg_mode): Likewise.
3626 (VI12HF_AVX512VL): New mode_iterator.
3627 (VF_AVX512FP16): Likewise.
3629 (VIHF_256): Likewise.
3630 (VIHF_AVX512BW): Likewise.
3631 (V16_256): Likewise.
3632 (V32_512): Likewise.
3633 (sseintmodesuffix): New mode_attr.
3634 (sse): Add scalar and vector HFmodes.
3635 (ssescalarmode): Add vector HFmode mapping.
3636 (ssescalarmodesuffix): Add sh suffix for HFmode.
3637 (*<sse>_vm<insn><mode>3): Use VFH_128.
3638 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
3639 (*ieee_<ieee_maxmin><mode>3): Likewise.
3640 (<avx512>_blendm<mode>): New define_insn.
3641 (vec_setv8hf): New define_expand.
3642 (vec_set<mode>_0): New define_insn for HF vector set.
3643 (*avx512fp16_movsh): Likewise.
3644 (avx512fp16_movsh): Likewise.
3645 (vec_extract_lo_v32hi): Rename to ...
3646 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
3648 (vec_extract_hi_v32hi): Likewise.
3649 (vec_extract_hi_<mode>): Likewise.
3650 (vec_extract_lo_v16hi): Likewise.
3651 (vec_extract_lo_<mode>): Likewise.
3652 (vec_extract_hi_v16hi): Likewise.
3653 (vec_extract_hi_<mode>): Likewise.
3654 (vec_set_hi_v16hi): Likewise.
3655 (vec_set_hi_<mode>): Likewise.
3656 (vec_set_lo_v16hi): Likewise.
3657 (vec_set_lo_<mode>): Likewise.
3658 (*vec_extract<mode>_0): New define_insn_and_split for HF
3660 (*vec_extracthf): New define_insn.
3661 (VEC_EXTRACT_MODE): Add HF vector modes.
3662 (PINSR_MODE): Add V8HF.
3663 (sse2p4_1): Likewise.
3664 (pinsr_evex_isa): Likewise.
3665 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
3666 insert for V8HFmode.
3667 (pbroadcast_evex_isa): Add HF vector modes.
3668 (AVX2_VEC_DUP_MODE): Likewise.
3669 (VEC_INIT_MODE): Likewise.
3670 (VEC_INIT_HALF_MODE): Likewise.
3671 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
3673 (avx2_pbroadcast<mode>_1): Likewise.
3674 (<avx512>_vec_dup<mode>_1): Likewise.
3675 (<avx512>_vec_dup<mode><mask_name>): Likewise.
3676 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
3679 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
3680 H.J. Lu <hongjiu.lu@intel.com>
3681 Liu Hongtao <hongtao.liu@intel.com>
3682 Wang Hongyu <hongyu.wang@intel.com>
3683 Xu Dianhong <dianhong.xu@intel.com>
3685 * common/config/i386/cpuinfo.h (get_available_features):
3686 Detect FEATURE_AVX512FP16.
3687 * common/config/i386/i386-common.c
3688 (OPTION_MASK_ISA_AVX512FP16_SET,
3689 OPTION_MASK_ISA_AVX512FP16_UNSET,
3690 OPTION_MASK_ISA2_AVX512FP16_SET,
3691 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
3692 (OPTION_MASK_ISA2_AVX512BW_UNSET,
3693 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
3694 (ix86_handle_option): Handle -mavx512fp16.
3695 * common/config/i386/i386-cpuinfo.h (enum processor_features):
3696 Add FEATURE_AVX512FP16.
3697 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
3698 * config.gcc: Add avx512fp16intrin.h.
3699 * config/i386/avx512fp16intrin.h: New intrinsic header.
3700 * config/i386/cpuid.h: Add bit_AVX512FP16.
3701 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
3702 * config/i386/i386-builtins.c: Support _Float16 type for i386
3704 (ix86_register_float16_builtin_type): New function.
3705 (ix86_float16_type_node): New.
3706 * config/i386/i386-c.c (ix86_target_macros_internal): Define
3708 * config/i386/i386-expand.c (ix86_expand_branch): Support
3710 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
3711 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3712 (ix86_expand_fp_movcc): Ditto.
3713 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
3714 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
3715 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
3716 * config/i386/i386.c (ix86_get_ssemov): Use
3717 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
3718 (ix86_get_excess_precision): Use
3719 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
3721 (sse_store_index): Use SFmode cost for HFmode cost.
3722 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
3723 GPR cost for HFmode.
3724 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
3725 (ix86_mangle_type): Add manlging for _Float16 type.
3726 (inline_secondary_memory_needed): No memory is needed for
3727 16bit movement between gpr and sse reg under
3729 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
3730 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3731 (ix86_division_cost): Ditto.
3732 (ix86_rtx_costs): Ditto.
3733 (ix86_add_stmt_cost): Ditto.
3734 (ix86_optab_supported_p): Ditto.
3735 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
3736 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
3737 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
3738 * config/i386/i386.md (mode): Add HFmode.
3739 (MODE_SIZE): Add HFmode.
3740 (isa): Add avx512fp16.
3741 (enabled): Handle avx512fp16.
3742 (ssemodesuffix): Add sh suffix for HFmode.
3743 (comm): Add mult, div.
3744 (plusminusmultdiv): New code iterator.
3745 (insn): Add mult, div.
3746 (*movhf_internal): Adjust for avx512fp16 instruction.
3747 (*movhi_internal): Ditto.
3748 (*cmpi<unord>hf): New define_insn for HFmode.
3749 (*ieee_s<ieee_maxmin>hf3): Likewise.
3750 (extendhf<mode>2): Likewise.
3751 (trunc<mode>hf2): Likewise.
3752 (float<floatunssuffix><mode>hf2): Likewise.
3753 (*<insn>hf): Likewise.
3754 (cbranchhf4): New expander.
3755 (movhfcc): Likewise.
3756 (<insn>hf3): Likewise.
3759 * config/i386/i386.opt: Add mavx512fp16.
3760 * config/i386/immintrin.h: Include avx512fp16intrin.h.
3761 * doc/invoke.texi: Add mavx512fp16.
3762 * doc/extend.texi: Add avx512fp16 Usage Notes.
3764 2021-09-08 liuhongt <hongtao.liu@intel.com>
3766 * common.opt: Support -fexcess-precision=16.
3767 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
3768 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
3769 EXCESS_PRECISION_TYPE_FLOAT16.
3770 * config/arm/arm.c (arm_excess_precision): Ditto.
3771 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
3772 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
3773 when EXCESS_PRECISION_TYPE_FLOAT16.
3774 * config/s390/s390.c (s390_excess_precision): Ditto.
3775 * coretypes.h (enum excess_precision_type): Add
3776 EXCESS_PRECISION_TYPE_FLOAT16.
3777 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
3778 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
3779 * doc/extend.texi (Half-Precision): Document
3780 -fexcess-precision=16.
3781 * flag-types.h (enum excess_precision): Add
3782 EXCESS_PRECISION_FLOAT16.
3783 * target.def (excess_precision): Update document.
3784 * tree.c (excess_precision_type): Set excess_precision_type to
3785 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
3787 2021-09-08 liuhongt <hongtao.liu@intel.com>
3789 * doc/extend.texi: (@node Floating Types): Adjust the wording.
3790 (@node Half-Precision): Ditto.
3792 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
3795 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
3796 'CONST_INT_P (src)' to the condition of the block that tries to
3797 eliminate literal when loading integer contant.
3799 2021-09-07 David Faust <david.faust@oracle.com>
3801 * doc/extend.texi (BPF Type Attributes) New node.
3802 Document new preserve_access_index attribute.
3803 Document new preserve_access_index builtin.
3804 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
3806 2021-09-07 David Faust <david.faust@oracle.com>
3808 * config/bpf/bpf.c: Adjust includes.
3809 (bpf_handle_preserve_access_index_attribute): New function.
3810 (bpf_attribute_table): Use it here.
3811 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
3812 (bpf_option_override): Handle "-mco-re" option.
3813 (bpf_asm_init_sections): New.
3814 (TARGET_ASM_INIT_SECTIONS): Redefine.
3815 (bpf_file_end): New.
3816 (TARGET_ASM_FILE_END): Redefine.
3817 (bpf_init_builtins): Add "__builtin_preserve_access_index".
3818 (bpf_core_compute, bpf_core_get_index): New.
3819 (is_attr_preserve_access): New.
3820 (bpf_expand_builtin): Handle new builtins.
3821 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
3822 (bpf_core_walk): New.
3823 (bpf_resolve_overloaded_builtin): New.
3824 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
3826 (pass_bpf_core_attr): New RTL pass.
3827 * config/bpf/bpf-passes.def: New file.
3828 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
3829 * config/bpf/coreout.c: New file.
3830 * config/bpf/coreout.h: Likewise.
3831 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
3832 (coreout.o): New rule.
3833 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
3834 * config.gcc (bpf): Add coreout.h to extra_headers.
3835 Add coreout.o to extra_objs.
3836 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
3838 2021-09-07 David Faust <david.faust@oracle.com>
3840 * btfout.c (get_btf_id): Function is no longer static.
3841 * ctfc.h: Expose it here.
3843 2021-09-07 David Faust <david.faust@oracle.com>
3845 * ctfc.c (ctf_lookup_tree_type): New function.
3848 2021-09-07 David Faust <david.faust@oracle.com>
3850 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
3851 * ctfc.h: Analogous change.
3853 2021-09-07 David Faust <david.faust@oracle.com>
3855 * dwarf2out.c (lookup_type_die): Function is no longer static.
3856 * dwarf2out.h: Expose it here.
3858 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3860 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
3861 (ctf_debug_early_finish): New definition.
3862 (ctf_debug_finish): Likewise.
3863 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
3864 (ctf_debug_early_finish): New declaration.
3865 (ctf_debug_finish): Likewise.
3866 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
3867 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
3869 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3871 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
3872 support when compiling for CO-RE.
3873 * config/bpf/bpf.opt: Add new command line option -mco-re.
3875 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3877 * flag-types.h (enum debug_info_type): Add new enum
3878 DINFO_TYPE_BTF_WITH_CORE.
3879 (BTF_WITH_CORE_DEBUG): New bitmask.
3880 * flags.h (btf_with_core_debuginfo_p): New declaration.
3881 * opts.c (btf_with_core_debuginfo_p): New definition.
3883 2021-09-07 Jason Merrill <jason@redhat.com>
3885 * tree.h (error_operand_p): Change to inline function.
3887 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
3889 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
3890 (empty_block_with_phis_p): ...this.
3891 (potentially_threadable_block): Same.
3892 (jump_threader::thread_through_normal_block): Same.
3894 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
3897 * dwarf2out.c (mark_base_types): New overloaded function.
3898 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
3899 as the compilation unit, and call move_marked_base_types afterward.
3901 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
3904 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
3906 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
3908 2021-09-07 Richard Biener <rguenther@suse.de>
3910 PR tree-optimization/102226
3911 * tree-vect-loop.c (vect_transform_cycle_phi): Record
3912 the converted value for the epilogue PHI use.
3914 2021-09-07 Martin Liska <mliska@suse.cz>
3916 PR gcov-profile/80223
3917 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
3918 options, do not inline when no_profile_instrument_function
3919 attributes are different in early inliner. It's fine to inline
3920 it after PGO instrumentation.
3922 2021-09-07 Richard Biener <rguenther@suse.de>
3924 PR tree-optimization/101555
3925 * tree-ssa-pre.c (translate_vuse_through_block): Do not
3926 perform an alias walk to determine the validity of the
3927 mem at the start of the block which is already guaranteed
3928 by means of prune_clobbered_mems.
3929 (phi_translate_1): Pass edge to translate_vuse_through_block.
3931 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
3934 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
3935 (remainder<mode>3): Likewise.
3937 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
3939 * config/mips/mips.c (mips_file_start): add .module for
3942 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
3944 * wide-int.cc (wi::clz): Reorder tests to ensure the result
3945 is zero for all negative values.
3947 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
3949 * doc/invoke.texi (-foffload-options): Fix @opindex.
3951 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
3954 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
3955 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
3956 source alternative for AVX.
3958 2021-09-06 liuhongt <hongtao.liu@intel.com>
3960 PR middle-end/102182
3961 * optabs.c (expand_fix): Add from1 to avoid from being
3964 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
3966 * dwarf2out.c (modified_type_die): Deal with all array types earlier
3967 and use local variable consistently throughout the function.
3969 2021-09-06 Jakub Jelinek <jakub@redhat.com>
3971 PR tree-optimization/102207
3972 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
3973 were promoted from signed to wider unsigned type.
3975 2021-09-06 Andrew Pinski <apinski@marvell.com>
3977 PR tree-optimization/63184
3978 * match.pd: Add simplification of pointer_diff of two pointer_plus
3979 with addr_expr in the first operand of each pointer_plus.
3980 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
3981 in the first operand of each pointer_plus.
3983 2021-09-06 Richard Biener <rguenther@suse.de>
3985 PR tree-optimization/102176
3986 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
3988 (vect_bb_slp_scalar_cost): Use the computed set of
3989 vectorized scalar stmts instead of relying on the out-of-date
3990 and not accurate PURE_SLP_STMT.
3991 (vect_bb_vectorization_profitable_p): Compute the set
3992 of vectorized scalar stmts.
3994 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
3996 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
3997 GIMPLE_COND special casing.
3998 (path_range_query::range_defined_in_block): Use range_of_stmt
3999 instead of calling fold_range directly.
4001 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4003 * gimple-range-path.cc (path_range_query::range_of_expr): Set
4004 m_undefined_path when appropriate.
4005 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
4006 (path_range_query::unreachable_path_p): New.
4007 (path_range_query::precompute_ranges): Set m_undefined_path.
4008 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
4009 (path_range_query::internal_range_of_expr): New.
4010 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
4011 Use unreachable_path_p.
4013 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4015 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4016 Remove argument and call find_taken_edge.
4017 (back_threader::resolve_phi): Do not calculate taken edge before
4018 calling maybe_register_path.
4019 (back_threader::find_paths_to_names): Same.
4021 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
4023 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
4024 * config/h8300/testcompare.md (store_c): Update name, use new
4026 (store_neg_c, store_shifted_c): New patterns.
4028 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4031 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
4032 instead of r12 for CR save, in all cases.
4034 2021-09-03 Andrew Pinski <apinski@marvell.com>
4036 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
4037 Handle error_mark_node as the type of the type_decl.
4039 2021-09-03 Andrew Pinski <apinski@marvell.com>
4041 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
4043 (aarch64_simd_types): Likewise.
4044 (aarch64_simd_intOI_type_node): Likewise.
4045 (aarch64_simd_intCI_type_node): Likewise.
4046 (aarch64_simd_intXI_type_node): Likewise.
4047 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
4048 (aarch64_fp16_ptr_type_node): Likewise.
4049 (aarch64_bf16_type_node): Likewise.
4050 (aarch64_bf16_ptr_type_node): Likewise.
4052 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4054 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
4056 (minus_op1_op2_relation_effect): ...here.
4057 (class operator_pointer_diff): New.
4058 (operator_pointer_diff::op1_op2_relation_effect): Call
4059 minus_op1_op2_relation_effect.
4060 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
4062 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4064 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
4065 Add may_peel_loop_headers.
4066 (back_threader_registry::thread_through_all_blocks): Same.
4067 (try_thread_blocks): Pass may_peel_loop_headers argument.
4068 (pass_early_thread_jumps::execute): Same.
4070 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4072 * tree-ssa-threadedge.c (has_phis_p): New.
4073 (forwarder_block_p): New.
4074 (potentially_threadable_block): Call forwarder_block_p.
4075 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
4076 (jump_threader::thread_through_normal_block): Call
4079 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4081 * tree-ssa-threadbackward.c (back_threader::dump): New.
4082 (back_threader::debug): New.
4083 (back_threader_profitability::profitable_path_p): Dump blocks
4084 even if we are bailing early.
4086 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4088 * tree-ssa-threadupdate.c (cancel_thread): New.
4089 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
4090 (jump_thread_path_registry::mark_threaded_blocks): Same.
4091 (jump_thread_path_registry::register_jump_thread): Same.
4093 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4095 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
4096 which objects are available.
4097 (jt_state::pop): Same.
4098 (jt_state::register_equiv): Same.
4099 (jt_state::register_equivs_on_edge): Same.
4101 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4103 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
4104 Move pop until after a thread is registered.
4106 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4108 * tree-ssa-threadupdate.c (debug): New.
4110 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4112 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
4113 (push_dump_file::~push_dump_file): New.
4114 (dump_ranger): Change dump_file temporarily while dumping
4116 * gimple-range-trace.h (class push_dump_file): New.
4118 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4120 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4121 (dump_ranger): Dump function name.
4123 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4125 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4126 Adjust for non-null.
4127 (path_range_query::adjust_for_non_null_uses): New.
4128 (path_range_query::precompute_ranges): Call
4129 adjust_for_non_null_uses.
4130 * gimple-range-path.h: Add m_non_null and
4131 adjust_for_non_null_uses.
4133 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4135 * gimple-range-path.cc (path_range_query::dump): Dump path
4137 (path_range_query::precompute_ranges): Dump entire path.
4139 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4141 * value-relation.cc (relation_oracle::debug): New.
4142 * value-relation.h (relation_oracle::debug): New.
4144 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4146 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4148 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4150 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4151 Skip statements with no defining BB.
4152 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4153 Do not get confused by statements with no defining BB.
4155 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4157 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4158 gimple_range_adjustment. Add support for constants.
4159 (adjust_realpart_expr): New.
4160 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4161 adjust_imagpart_expr.
4162 * range-op.cc (integral_table::integral_table): Add entry for
4165 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4167 * omp-expand.c (expand_omp_atomic_pipeline): Use
4168 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4169 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4172 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4175 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4176 that is false if DECL_BIT_FIELD.
4177 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4178 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4179 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4180 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4181 of DECL_FIELD_ABI_IGNORED.
4182 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4183 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4184 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4185 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4186 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4187 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4189 2021-09-03 liuhongt <hongtao.liu@intel.com>
4192 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4193 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4194 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4196 2021-09-02 Martin Sebor <msebor@redhat.com>
4198 PR tree-optimization/17506
4200 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4202 2021-09-02 Richard Biener <rguenther@suse.de>
4204 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4205 fix for PR78185 and continue processing when leaving
4208 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4210 PR tree-optimization/99591
4211 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4214 2021-09-02 Richard Biener <rguenther@suse.de>
4217 2021-09-02 Richard Biener <rguenther@suse.de>
4219 PR tree-optimization/102155
4220 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4221 over a part of the RPO array and do not recurse here.
4222 Dump blocks marked as always executed.
4223 (fill_always_executed_in): Walk over the RPO array and
4224 process loops whose header we run into.
4225 (loop_invariant_motion_in_fun): Compute the first RPO
4226 using rev_post_order_and_mark_dfs_back_seme in iteration
4227 order and pass that to fill_always_executed_in.
4229 2021-09-02 liuhongt <hongtao.liu@intel.com>
4231 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4232 * config/i386/i386.c (enum x86_64_reg_class): Add
4234 (merge_classes): Handle X86_64_SSEHF_CLASS.
4235 (examine_argument): Ditto.
4236 (construct_container): Ditto.
4237 (classify_argument): Ditto, and set HFmode/HCmode to
4239 (function_value_32): Return _FLoat16/Complex Float16 by
4241 (function_value_64): Return _Float16/Complex Float16 by SSE
4243 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
4244 (ix86_secondary_reload): Require gpr as intermediate register
4245 to store _Float16 from sse register when sse4 is not
4247 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
4249 (ix86_scalar_mode_supported_p): Ditto.
4250 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
4251 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
4252 (VALID_INT_MODE_P): Add HFmode and HCmode.
4253 * config/i386/i386.md (*pushhf_rex64): New define_insn.
4255 (*movhf_internal): Ditto.
4256 * doc/extend.texi (Half-Precision Floating Point): Documemt
4259 2021-09-02 Richard Biener <rguenther@suse.de>
4261 PR tree-optimization/102155
4262 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4263 over a part of the RPO array and do not recurse here.
4264 Dump blocks marked as always executed.
4265 (fill_always_executed_in): Walk over the RPO array and
4266 process loops whose header we run into.
4267 (loop_invariant_motion_in_fun): Compute the first RPO
4268 using rev_post_order_and_mark_dfs_back_seme in iteration
4269 order and pass that to fill_always_executed_in.
4271 2021-09-02 YunQiang Su <syq@debian.org>
4274 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4276 * config/mips/mips.c (mips_module_isa_name): New.
4277 mips_file_start: add .module mipsREV to all asm output
4279 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4281 PR tree-optimization/102152
4282 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4283 comparison to a scalar comparison before calling
4284 update_stmt_if_modified.
4286 2021-09-01 Andrew Pinski <apinski@marvell.com>
4289 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4290 Check STRICT_ALIGNMENT before creating an overlapping
4293 2021-09-01 Martin Sebor <msebor@redhat.com>
4295 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4296 (check_access): Pass additional argument.
4297 (check_memop_access): Remove template and make a member function.
4298 (maybe_check_dealloc_call): Make a pass_waccess member function.
4299 (class pass_waccess): Add, rename, and remove members.
4300 (pass_waccess::pass_waccess): Adjust to name change.
4301 (pass_waccess::~pass_waccess): Same.
4302 (check_alloca): Make a member function.
4303 (check_alloc_size_call): Same.
4304 (check_strcat): Same.
4305 (check_strncat): Same.
4306 (check_stxcpy): Same.
4307 (check_stxncpy): Same.
4308 (check_strncmp): Same.
4309 (maybe_warn_rdwr_sizes): Rename...
4310 (pass_waccess::maybe_check_access_sizes): ...to this.
4311 (pass_waccess::check_call): Adjust to name changes.
4312 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4314 (pass_waccess::execute): Adjust to name changes.
4315 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4316 * pointer-query.cc (access_ref::phi): Handle null pointer.
4317 (access_ref::inform_access): Same.
4318 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4319 (pointer_query::dump): New function.
4320 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4322 * pointer-query.h (pointer_query::dump): Declare.
4323 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4324 (dump_strlen_info): Use function query.
4325 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4327 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4329 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4331 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4333 * doc/extend.texi: Document unavailable attribute.
4334 * print-tree.c (print_node): Handle unavailable attribute.
4335 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4336 * tree.c (error_unavailable_use): New.
4337 * tree.h (TREE_UNAVAILABLE): New.
4338 (error_unavailable_use): New.
4340 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4342 PR tree-optimization/102124
4343 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4344 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4345 add an extra cast to signed variant of itype to ensure sign-extension.
4347 2021-09-01 Martin Liska <mliska@suse.cz>
4349 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4350 edges and rather use colors for TRUE and FALSE edges.
4352 2021-09-01 Richard Biener <rguenther@suse.de>
4354 PR tree-optimization/93491
4355 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4356 after processing the stmt itself. Do not consider
4357 pure functions possibly not returning. Properly avoid
4358 adding possibly trapping calls to EXP_GEN when there's
4359 a preceeding possibly not returning call.
4360 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4363 2021-09-01 Richard Biener <rguenther@suse.de>
4365 PR tree-optimization/102139
4366 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4367 type to record a std::pair of the stmt-info and the innermost
4369 (dr_vec_info::group): New member.
4370 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4371 (vect_compute_data_ref_alignment): Verify the recorded
4372 base alignment can be used.
4373 (data_ref_pair): Remove.
4374 (dr_group_sort_cmp): Adjust.
4375 (vect_analyze_data_ref_accesses): Store the group-ID in the
4376 dr_vec_info and operate on a vector of dr_vec_infos.
4378 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4380 * read-md.c (md_reader::handle_enum): support value assignation.
4381 * doc/md.texi: record define_c_enum value assignation support.
4383 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4385 PR tree-optimization/102141
4386 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4387 argument. If false, emit stmts after gsi instead of before, and
4389 (bswap_replace): Adjust callers. When converting output of bswap,
4390 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4393 2021-09-01 liuhongt <hongtao.liu@intel.com>
4395 * emit-rtl.c (validate_subreg): Get rid of all float-int
4398 2021-09-01 liuhongt <hongtao.liu@intel.com>
4401 2021-08-30 liuhongt <hongtao.liu@intel.com>
4403 * expmed.c (extract_bit_field_1): Make sure we're playing with
4404 integral modes before call extract_integral_bit_field.
4405 (extract_integral_bit_field): Add a parameter of type
4406 scalar_int_mode which corresponds to of tmode.
4407 And call extract_and_convert_fixed_bit_field instead of
4408 extract_fixed_bit_field and convert_extracted_bit_field.
4409 (extract_and_convert_fixed_bit_field): New function, it's a
4410 combination of extract_fixed_bit_field and
4411 convert_extracted_bit_field.
4413 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4415 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4417 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4419 * omp-general.h (omp_is_reference): Rename to...
4420 (omp_privatize_by_reference): ... this. Adjust all users...
4421 * omp-general.c: ... here, ...
4422 * gimplify.c: ... here, ...
4423 * omp-expand.c: ... here, ...
4424 * omp-low.c: ... here.
4426 2021-08-31 Martin Sebor <msebor@redhat.com>
4428 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4429 pointer element for equality to zero, not that of the cotaining
4432 2021-08-31 Martin Sebor <msebor@redhat.com>
4434 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4436 2021-08-31 Martin Sebor <msebor@redhat.com>
4438 * function.h (function): Add comments.
4439 (get_range_query): Same. Add attribute returns nonnull.
4441 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4443 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4444 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4445 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4447 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4449 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4451 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4452 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4453 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4454 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4455 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4456 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4457 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4458 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4459 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4460 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4462 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4464 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4465 allowed on target constructs and only with particular other clauses.
4466 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4468 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4469 structs when 'ancestor' is used.
4470 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4472 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4473 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4475 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4477 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4478 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4480 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4481 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4482 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4483 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4484 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4485 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4486 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4487 would be paradoxical.
4489 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4491 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4492 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4493 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4495 2021-08-31 Richard Biener <rguenther@suse.de>
4497 PR tree-optimization/102142
4498 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4499 condition under which to unset the visited flag.
4501 2021-08-31 Richard Biener <rguenther@suse.de>
4503 PR middle-end/102129
4504 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4505 possibly trapping expressions across calls.
4507 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4509 PR tree-optimization/102134
4510 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4511 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4514 2021-08-31 Andrew Pinski <apinski@marvell.com>
4517 * collect-utils.c (setup_signals): New declaration.
4518 * collect-utils.h (setup_signals): New function.
4519 * collect2.c (handler): Delete.
4520 (main): Instead of manually setting up the signals,
4521 just call setup_signals.
4522 * lto-wrapper.c (main): Likewise.
4524 2021-08-31 Andrew Pinski <apinski@marvell.com>
4527 * config/i386/i386-protos.h (x86_output_aligned_bss):
4528 Change align argument to unsigned type.
4529 (x86_elf_aligned_decl_common): Likewise.
4530 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4531 (x86_output_aligned_bss): Likewise.
4533 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4535 * config/mips/mips.c (mips_module_isa_name): New.
4536 mips_file_start: add .module mipsREV to all asm output
4538 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4540 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4541 use enum instead of int for 'isa' member.
4542 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4543 config/mips/netbsd.h: replace hardcoded numbers with enum.
4545 2021-08-31 liuhongt <hongtao.liu@intel.com>
4547 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4548 define_split to define_insn_and_split.
4549 (*avx2_eq<mode>3): Removed.
4550 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4551 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4552 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4554 (*avx2_gt<mode>3): Removed.
4555 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4556 define_insn to define_expand, and adjust pattern.
4557 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4559 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4562 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4564 2021-08-30 Jason Merrill <jason@redhat.com>
4566 * doc/invoke.texi: Document -Wmissing-requires.
4568 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4570 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4571 TARGET_EXTRA_BUILTINS guard.
4573 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4575 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4576 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4578 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4580 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4581 decl when new_builtins_are_live.
4582 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4585 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4587 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4588 OPTION_MASK_P10_FUSION_2STORE.
4589 (POWERPC_MASKS): Likewise.
4590 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4591 store fusion for Power10.
4592 (is_fusable_store): New.
4593 (power10_sched_reorder): Likewise.
4594 (rs6000_sched_reorder): Do Power10 specific reordering.
4595 (rs6000_sched_reorder2): Likewise.
4596 * config/rs6000/rs6000.opt: Add new option.
4598 2021-08-30 Richard Biener <rguenther@suse.de>
4600 PR tree-optimization/102128
4601 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4602 Move scanning for if-converted scalar code to the caller
4603 and instead delay clearing the visited flag for profitable
4605 (vect_slp_region): Cost all subgraphs before scheduling.
4606 For if-converted BB vectorization scan for scalar COND_EXPRs
4607 and do not vectorize if any found and the cost model is
4610 2021-08-30 Richard Biener <rguenther@suse.de>
4612 * common.opt (fexceptions): Mark
4613 EnabledBy(fnon-call-exceptions).
4614 * doc/invoke.texi (fnon-call-exceptions): Document this
4615 enables -fexceptions.
4617 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
4619 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
4620 is not already defined.
4622 2021-08-30 liuhongt <hongtao.liu@intel.com>
4624 * expmed.c (extract_bit_field_1): Make sure we're playing with
4625 integral modes before call extract_integral_bit_field.
4626 (extract_integral_bit_field): Add a parameter of type
4627 scalar_int_mode which corresponds to of tmode.
4628 And call extract_and_convert_fixed_bit_field instead of
4629 extract_fixed_bit_field and convert_extracted_bit_field.
4630 (extract_and_convert_fixed_bit_field): New function, it's a
4631 combination of extract_fixed_bit_field and
4632 convert_extracted_bit_field.
4634 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4636 * config/darwin.c (darwin_libc_has_function): Do not run
4637 the checks for x86 or modern Darwin. Make sure that there
4638 is a value set for darwin_macosx_version_min before testing.
4640 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4642 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
4644 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
4646 * ipa-modref-tree.h (modref_access_node::merge): Break out
4647 logic combining offsets and logic merging ranges to ...
4648 (modref_access_node::combined_offsets): ... here
4649 (modref_access_node::update2): ... here
4650 (modref_access_node::closer_pair_p): New member function.
4651 (modref_access_node::forced_merge): New member function.
4652 (modre_ref_node::insert): Do merging when table is full.
4654 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
4657 * config.gcc: MIPS: use N64 ABI by default if the triple end
4658 with -gnuabi64, which is used by Debian since 2013.
4660 2021-08-28 Alexandre Oliva <oliva@adacore.com>
4662 * ipa-modref.c (analyze_function): Skip debug stmts.
4663 * tree-inline.c (estimate_num_insn): Consider builtins even
4664 without a cgraph_node.
4666 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4668 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
4669 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
4670 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
4671 * config/h8300/h8300.md (eqne): New code iterator.
4672 (geultu, geultu_to_c): Similarly.
4673 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
4674 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
4675 (cmp<mode>_c): New pattern
4677 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4679 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
4681 (dom_opt_dom_walker::optimize_stmt): Use it.
4683 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4685 * config/darwin.c (finalize_ctors): Add a section-start linker-
4687 (finalize_dtors): Likewise.
4688 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
4690 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
4692 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
4693 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
4694 old initialization logic when new builtins are enabled.
4695 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
4696 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
4697 (write_init_file): Likewise.
4699 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4701 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
4702 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
4703 (amdgcn-* | gcn-*) Likewise.
4705 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
4707 * symbol-summary.h: Added missing template keyword.
4709 2021-08-27 Richard Biener <rguenther@suse.de>
4711 PR tree-optimization/45178
4712 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
4713 infinite loops without exit do not mark control dependent
4714 edges of the latch necessary.
4716 2021-08-27 konglin1 <lingling.kong@intel.com>
4719 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
4721 (<avx512>scattersi<mode>): Likewise.
4722 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
4723 (*avx512f_scatterdi<VI48F:mode>): Likewise
4725 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4727 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
4728 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
4729 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
4730 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
4731 P10_BUILTIN_PEXTD on Power10.
4733 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4735 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
4736 signedness for some Power10 bifs.
4738 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
4741 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
4742 computed alignment if the entire struct has attribute packed.
4744 2021-08-27 liuhongt <hongtao.liu@intel.com>
4748 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
4749 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
4750 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
4751 IX86_BUILTIN_SHUFPS256.
4752 (ix86_masked_all_ones): New function.
4754 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4756 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
4757 (conditional moves with memory inputs splitters): Ditto.
4758 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
4760 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4762 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
4763 search after merging.
4765 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4767 * config/rs6000/rs6000-overload.def: Add remaining overloads.
4769 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4771 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
4773 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4775 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
4776 crypto, and htm stanzas.
4778 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4780 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
4782 2021-08-26 Martin Sebor <msebor@redhat.com>
4784 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
4785 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
4786 (warn_uninitialized_vars): Same. Reduce visibility of locals.
4787 (warn_uninitialized_phi): Same.
4789 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4791 * tree-ssa-ccp.c (get_individual_bits): Helper function to
4792 extract the individual bits from a widest_int constant (mask).
4793 (gray_code_bit_flips): New read-only table for effiently
4794 enumerating permutations/combinations of bits.
4795 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
4796 by unknown counts that are guaranteed less than the target
4797 precision and four or fewer unknown bits by enumeration.
4798 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
4799 enumeration under the same conditions. Handle remaining
4800 shifts as a mask based upon the minimum possible shift value.
4802 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4803 Richard Biener <rguenther@suse.de>
4805 * match.pd (shift transformations): Remove a redundant
4806 !POINTER_TYPE_P check.
4808 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4811 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
4812 Set all_regs to true in the call to replace_rtx.
4814 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4816 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
4817 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
4818 hiting --param modref-max-refs limit.
4819 (modref_tree:insert_base): Be smarter when hitting
4820 --param modref-max-bases limit. Add new parameter REF.
4821 (modref_tree:insert): Update.
4822 (modref_tree:merge): Update.
4823 * ipa-modref.c (read_modref_records): Update.
4825 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4827 * params.opt: (modref-max-adjustments): Add full stop.
4829 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4831 * ipa-modref-tree.h (modref_ref_node::verify): New member
4833 (modref_ref_node::insert): Use it.
4834 (modref_ref_node::try_mere_with): Fix off by one error.
4836 2021-08-26 Martin Liska <mliska@suse.cz>
4837 Stefan Kneifel <stefan.kneifel@bluewin.ch>
4839 * cgraph.h (create_version_clone_with_body): Add new parameter.
4840 * cgraphclones.c: Likewise.
4841 * multiple_target.c (create_dispatcher_calls): Do not use
4843 (create_target_clone): Likewise here.
4845 2021-08-26 Jonathan Yong <10walls@gmail.com>
4847 * doc/extend.texi: Add note about reserved priorities
4848 to the constructor attribute.
4850 2021-08-25 Martin Sebor <msebor@redhat.com>
4852 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
4853 range table header alone.
4854 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
4856 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
4858 * doc/invoke.texi: Document --param modref-max-adjustments.
4859 * ipa-modref-tree.c (test_insert_search_collapse): Update.
4860 (test_merge): Update.
4861 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
4862 (modref_access_node::operator==): Fix handling of access ranges.
4863 (modref_access_node::contains): Constify parameter; handle also
4864 mismatched parm offsets.
4865 (modref_access_node::update): New function.
4866 (modref_access_node::merge): New function.
4867 (unspecified_modref_access_node): Update constructor.
4868 (modref_ref_node::insert_access): Add record_adjustments parameter;
4870 (modref_ref_node::try_merge_with): New private function.
4871 (modref_tree::insert): New record_adjustments parameter.
4872 (modref_tree::merge): New record_adjustments parameter.
4873 (modref_tree::copy_from): Update.
4874 * ipa-modref.c (dump_access): Dump adjustments field.
4875 (get_access): Update constructor.
4876 (record_access): Update call of insert.
4877 (record_access_lto): Update call of insert.
4878 (merge_call_side_effects): Add record_adjustments parameter.
4879 (get_access_for_fnspec): Update.
4880 (process_fnspec): Update.
4881 (analyze_call): Update.
4882 (analyze_function): Update.
4883 (read_modref_records): Update.
4884 (ipa_merge_modref_summary_after_inlining): Update.
4885 (propagate_unknown_call): Update.
4886 (modref_propagate_in_scc): Update.
4887 * params.opt (param-max-modref-adjustments=): New.
4889 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
4891 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
4893 (xxspltiw_v4si): Use vecperm type attribute.
4894 (xxspltiw_v4si_inst): Use vecperm type attribute.
4895 (xxspltiw_v4sf_inst): Likewise.
4896 (xxspltidp_v2df): Use vecperm type attribute. Use
4897 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
4898 (xxspltidp_v2df_inst): Likewise.
4899 (xxsplti32dx_v4si): Use vecperm type attribute.
4900 (xxsplti32dx_v4si_inst): Likewise.
4901 (xxsplti32dx_v4sf_inst): Likewise.
4902 (xxblend_<mode>): Likewise.
4903 (xxpermx): Likewise.
4904 (xxpermx_inst): Likewise.
4907 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
4910 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
4911 * diagnostic.c (diagnostic_initialize_input_context): New function.
4912 * diagnostic.h (diagnostic_initialize_input_context): Declare.
4913 * input.c (default_charset_callback): New function.
4914 (file_cache::initialize_input_context): New function.
4915 (file_cache_slot::create): Added ability to convert the input
4916 according to the input context.
4917 (file_cache::file_cache): Initialize the new input context.
4918 (class file_cache_slot): Added new m_alloc_offset member.
4919 (file_cache_slot::file_cache_slot): Initialize the new member.
4920 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
4921 (file_cache_slot::maybe_grow): Likewise.
4922 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
4923 (file_cache_slot::get_next_line): Likewise.
4924 * input.h (class file_cache): Added input context member.
4926 2021-08-25 Richard Biener <rguenther@suse.de>
4928 PR tree-optimization/102046
4929 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
4930 update ->any_pattern when swapping operands.
4932 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
4935 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
4936 (ix86_decompose_address): Remove retval check for ASHIFT,
4937 allow non-canonical zero extend if AND mask covers ASHIFT
4939 (ix86_legitimate_address_p): Adjust condition for decompose.
4940 (ix86_rtx_costs): Adjust cost for lea with non-canonical
4942 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
4944 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
4946 PR tree-optimization/101145
4947 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
4949 (number_of_iterations_lt): Invoke above function.
4950 (adjust_cond_for_loop_until_wrap):
4951 Merge to number_of_iterations_until_wrap.
4952 (number_of_iterations_cond): Update invokes for
4953 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
4955 2021-08-25 konglin1 <lingling.kong@intel.com>
4958 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
4960 (_mm512_mask_fpclass_ps_mask): Ditto.
4962 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
4964 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
4965 (vec_unpacku_hi_v8hi): Likewise.
4966 (vec_unpacku_lo_v16qi): Likewise.
4967 (vec_unpacku_lo_v8hi): Likewise.
4968 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
4969 (vec_unpacku_lo_<VP_small_lc>): Likewise.
4971 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
4973 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
4974 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
4975 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
4976 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
4978 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
4980 PR middle-end/102031
4981 * simplify-rtx.c (simplify_truncation): When comparing precisions
4982 use "subreg_prec" variable, not "subreg_mode".
4984 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4986 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
4989 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4991 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
4992 various pointer type nodes.
4993 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
4994 values for various pointer types.
4995 (ptr_V16QI_type_node): New macro.
4996 (ptr_V1TI_type_node): New macro.
4997 (ptr_V2DI_type_node): New macro.
4998 (ptr_V2DF_type_node): New macro.
4999 (ptr_V4SI_type_node): New macro.
5000 (ptr_V4SF_type_node): New macro.
5001 (ptr_V8HI_type_node): New macro.
5002 (ptr_unsigned_V16QI_type_node): New macro.
5003 (ptr_unsigned_V1TI_type_node): New macro.
5004 (ptr_unsigned_V8HI_type_node): New macro.
5005 (ptr_unsigned_V4SI_type_node): New macro.
5006 (ptr_unsigned_V2DI_type_node): New macro.
5007 (ptr_bool_V16QI_type_node): New macro.
5008 (ptr_bool_V8HI_type_node): New macro.
5009 (ptr_bool_V4SI_type_node): New macro.
5010 (ptr_bool_V2DI_type_node): New macro.
5011 (ptr_bool_V1TI_type_node): New macro.
5012 (ptr_pixel_type_node): New macro.
5013 (ptr_intQI_type_node): New macro.
5014 (ptr_uintQI_type_node): New macro.
5015 (ptr_intHI_type_node): New macro.
5016 (ptr_uintHI_type_node): New macro.
5017 (ptr_intSI_type_node): New macro.
5018 (ptr_uintSI_type_node): New macro.
5019 (ptr_intDI_type_node): New macro.
5020 (ptr_uintDI_type_node): New macro.
5021 (ptr_intTI_type_node): New macro.
5022 (ptr_uintTI_type_node): New macro.
5023 (ptr_long_integer_type_node): New macro.
5024 (ptr_long_unsigned_type_node): New macro.
5025 (ptr_float_type_node): New macro.
5026 (ptr_double_type_node): New macro.
5027 (ptr_long_double_type_node): New macro.
5028 (ptr_dfloat64_type_node): New macro.
5029 (ptr_dfloat128_type_node): New macro.
5030 (ptr_ieee128_type_node): New macro.
5031 (ptr_ibm128_type_node): New macro.
5032 (ptr_vector_pair_type_node): New macro.
5033 (ptr_vector_quad_type_node): New macro.
5034 (ptr_long_long_integer_type_node): New macro.
5035 (ptr_long_long_unsigned_type_node): New macro.
5037 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5039 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
5040 and power9-64 stanzas.
5042 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5043 Tom de Vries <tdevries@suse.de>
5045 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
5046 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
5047 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
5048 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
5049 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
5050 * config/nvptx/nvptx-c.c: New source file.
5051 (nvptx_cpu_cpp_builtins): Move implementation here.
5053 2021-08-24 Martin Sebor <msebor@redhat.com>
5055 PR middle-end/101600
5056 PR middle-end/101977
5057 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
5058 the phrasing of a warning.
5059 (check_access): Use the remaining size after subtracting any offset
5060 rather than the whole object size.
5061 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
5062 clear for any nonnull PHI argument.
5063 (compute_objsize): Clear argument.
5065 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5067 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
5069 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5071 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
5074 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
5076 * value-relation.cc (rr_transitive_table): New.
5077 (relation_transitive): New.
5078 (value_relation::swap): Remove.
5079 (value_relation::apply_transitive): New.
5080 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
5081 (relation_oracle::register_relation): Call register_transitives.
5082 (relation_oracle::register_transitives): New.
5083 * value-relation.h (relation_oracle): Add new temporary bitmap and
5086 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
5089 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
5090 from integer to a pseudo vector register.
5092 2021-08-24 Richard Biener <rguenther@suse.de>
5094 PR tree-optimization/100089
5095 * tree-vectorizer.h (vect_slp_bb): Rename to ...
5096 (vect_slp_if_converted_bb): ... this and get the original
5097 loop as new argument.
5098 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
5099 pass original loop to vect_slp_if_converted_bb.
5100 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5101 If orig_loop was passed scan the not vectorized stmts
5102 for COND_EXPRs and force not profitable if found.
5103 (vect_slp_region): Pass down all SLP instances to costing
5104 if orig_loop was specified.
5105 (vect_slp_bbs): Pass through orig_loop.
5106 (vect_slp_bb): Rename to ...
5107 (vect_slp_if_converted_bb): ... this and get the original
5108 loop as new argument.
5109 (vect_slp_function): Adjust.
5111 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5114 * config/arm/arm.md (attribute arch): Add fix_vlldm.
5115 (arch_enabled): Use it.
5116 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5117 use when erratum mitigation is needed.
5119 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5122 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5123 * doc/invoke.texi (Arm Options): Document it.
5124 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5125 (ALL_QUIRKS): Add quirk_vlldm.
5126 (cortex-m33): Add quirk_vlldm.
5127 (cortex-m35p, cortex-m55): Likewise.
5128 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5129 targetting an affected CPU and not explicitly controlled on
5132 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5134 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5135 (lazy_load_multiple_insn): Likewise.
5137 2021-08-24 liuhongt <hongtao.liu@intel.com>
5140 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5141 Enable avx512 embedded broadcast.
5142 (*<avx512>_vternlog<mode>_all): Ditto.
5143 (<avx512>_vternlog<mode>_mask): Ditto.
5145 2021-08-24 liuhongt <hongtao.liu@intel.com>
5148 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5150 * config/i386/i386.h (STRIP_UNARY): New macro.
5151 * config/i386/predicates.md (reg_or_notreg_operand): New
5153 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5154 (*<avx512>_vternlog<mode>_1): New pre_reload
5155 define_insn_and_split.
5156 (*<avx512>_vternlog<mode>_2): Ditto.
5157 (*<avx512>_vternlog<mode>_3): Ditto.
5158 (any_logic1,any_logic2): New code iterator.
5159 (logic_op): New code attribute.
5160 (ternlogsuffix): Extend to VNxDF and VNxSF.
5162 2021-08-24 Richard Biener <rguenther@suse.de>
5164 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5165 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5167 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5168 inner_loop_cost_factor to the minimum of the estimated number
5169 of iterations of the inner loop and vect-inner-loop-cost-factor.
5171 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5172 Richard Biener <rguenther@suse.de>
5174 * config/i386/i386-features.c (compute_convert_gain): Provide
5175 more accurate values for CONST_INT, when optimizing for size.
5176 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5177 * config/i386/i386.h (COSTS_N_BYTES): to here.
5179 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5180 Jakub Jelinek <jakub@redhat.com>
5182 PR middle-end/102029
5183 * match.pd (shift transformations): Add an additional check for
5184 !POINTER_TYPE_P in the recently added left shift transformation.
5186 2021-08-24 liuhongt <hongtao.liu@intel.com>
5188 PR tree-optimization/100089
5189 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5190 loop vectorizer when cost model is very-cheap.
5192 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5194 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5195 asprintf, which is not available on AIX.
5197 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5199 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5200 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5202 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5204 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5205 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5207 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5208 (rbtree.o): Likewise.
5209 (rs6000-gen-builtins): Likewise.
5210 (rs6000-builtins.c): Likewise.
5211 (rs6000-builtins.h): Likewise.
5212 (rs6000.o): Add dependency.
5213 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5214 (rs6000-vecdefines.h): New target.
5215 (rs6000-builtins.o): Likewise.
5216 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5217 (rs6000-c.o): Likewise.
5219 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5222 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5223 Diagnose buffer overrun.
5224 (safe_inc_pos): Fix overrun detection.
5225 (match_identifier): Diagnose buffer overrun.
5226 (match_integer): Likewise.
5227 (match_to_right_bracket): Likewise.
5229 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5231 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5232 Improve range compare.
5233 (modref_access_node::contains): New member function.
5234 (modref_access_node::search): Remove.
5235 (modref_access_node::insert): Be smarter about subaccesses.
5237 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
5239 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
5240 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
5241 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
5244 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
5246 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
5248 * config/h8300/jumpcall.md (call, call_value): Restrict to
5249 !SIBLING_CALL_P cases.
5250 (subcall, sibcall_value): New patterns & expanders.
5251 * config/h8300/proepi.md (epilogue): Pass new argument to
5252 h8300_expand_epilogue.
5253 (sibcall_epilogue): New expander.
5254 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
5256 (h8300_ok_for_sibcall_p): New function.
5257 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
5259 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5261 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
5262 Handle case where the operand is already the desired mode.
5264 2021-08-23 Richard Biener <rguenther@suse.de>
5267 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5268 in addition to has_gimple_body.
5270 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5272 PR middle-end/101949
5273 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5276 2021-08-23 Martin Liska <mliska@suse.cz>
5278 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5281 2021-08-23 Richard Biener <rguenther@suse.de>
5283 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5284 Properly scale the inner loop cost only once.
5286 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5288 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5289 Provide bounds for unsigned (and signed with non-negative operands)
5290 division and modulus.
5292 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5294 * simplify-rtx.c (simplify_truncation): Generalize simplification
5295 of (truncate:A (subreg:B X)).
5296 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5297 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5298 already has the desired machine mode.
5299 (test_scalar_int_ops): Add tests that useless extensions and
5300 truncations are optimized away.
5301 (test_scalar_int_ext_ops): New self-test function to confirm
5302 that truncations of extensions are correctly simplified.
5303 (test_scalar_int_ext_ops2): New self-test function to check
5304 truncations of truncations, extensions of extensions, and
5305 truncations of extensions.
5306 (test_scalar_ops): Call the above two functions with a
5307 representative sampling of integer machine modes.
5309 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5311 * match.pd (shift transformations): Change the sign of an
5312 LSHIFT_EXPR if it reduces the number of explicit conversions.
5314 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5316 PR tree-optimization/86723
5317 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5318 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5319 non-memory permutation of 64-bit src which only has bytes of
5320 0 or [5..8] and n->range is 4.
5321 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5322 find_bswap_or_nop_finalize caller, support bswap with some bytes
5323 zeroed, as long as at least two bytes are not zeroed.
5324 (bswap_replace): Add mask argument and handle masking of bswap
5326 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5327 caller, punt if cast64_to_32 or mask is not all ones.
5328 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5329 caller, for now punt if cast64_to_32.
5331 2021-08-23 Richard Biener <rguenther@suse.de>
5333 PR tree-optimization/79334
5334 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5335 a type also for COMPONENT_REFs.
5336 (vn_reference_may_trap): Check ARRAY_REF with constant index
5337 against the array domain.
5339 2021-08-23 liuhongt <hongtao.liu@intel.com>
5342 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5343 TARGET_AVX512BW to condition.
5345 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5348 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5349 register variables already during early_dwarf if possible.
5351 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5353 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5355 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5358 * config/arm/arm.opt: Fix typo.
5359 * config/arm/t-rmprofile: Fix typo.
5361 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5363 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5364 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5365 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5366 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5367 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5368 if either grainsize or num_tasks clause has the strict modifier.
5370 2021-08-23 Martin Liska <mliska@suse.cz>
5372 * dbgcnt.def (DEBUG_COUNTER): New counter.
5373 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5375 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5377 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5379 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5382 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5384 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5386 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5387 vlshr<mode>3): Call it.
5389 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5390 Martin Liska <mliska@suse.cz>
5392 PR middle-end/101949
5393 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5396 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5398 * config/mips/mips.c (mips_function_rodata_section,
5399 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5401 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5403 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5405 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5407 * configure.ac (thread-local storage support): Remove tls_first_major
5408 and tls_first_minor. Use "$conftest_s" to check support.
5409 * configure: Regenerate.
5411 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5413 * configure.ac: Fixup formatting.
5415 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5417 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5418 * configure.ac: ... update all callers.
5420 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5423 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5424 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5425 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5426 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5427 Remove remaining checks for in-tree assembler.
5428 * configure: Regenerate.
5430 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5432 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5433 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5435 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5436 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5437 Improve logical shifts by 27 bits for H8/S.
5438 (get_shift_alg): Corresponding changes.
5439 (h8300_option_override): Revert to loops for -Os when profitable.
5441 2021-08-20 Richard Biener <rguenther@suse.de>
5443 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5445 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5446 index as group_id when dataref_groups were not computed.
5447 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5448 we advace to the next BB.
5450 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5452 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5455 2021-08-20 Martin Liska <mliska@suse.cz>
5457 PR gcov-profile/89961
5458 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5459 (mangle_name): Simplify, do not used the second argument.
5460 (strip_extention): New function.
5461 (get_md5sum): Likewise.
5462 (get_gcov_intermediate_filename): Handle properly -p and -x
5464 (output_gcov_file): Use string type.
5465 (generate_results): Likewise.
5466 (md5sum_to_hex): Remove.
5468 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5470 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5471 (UNSPEC_XXSPLTIW): Move to vsx.md.
5472 (UNSPEC_XXSPLTID): Move to vsx.md.
5473 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5474 (UNSPEC_XXBLEND): Move to vsx.md.
5475 (UNSPEC_XXPERMX): Move to vsx.md.
5476 (VM3): Move to vsx.md.
5477 (VM3_char): Move to vsx.md.
5478 (xxspltiw_v4si): Move to vsx.md.
5479 (xxspltiw_v4sf): Move to vsx.md.
5480 (xxspltiw_v4sf_inst): Move to vsx.md.
5481 (xxspltidp_v2df): Move to vsx.md.
5482 (xxspltidp_v2df_inst): Move to vsx.md.
5483 (xxsplti32dx_v4si_inst): Move to vsx.md.
5484 (xxsplti32dx_v4sf): Move to vsx.md.
5485 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5486 (xxblend_<mode>): Move to vsx.md.
5487 (xxpermx): Move to vsx.md.
5488 (xxpermx_inst): Move to vsx.md.
5489 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5490 (UNSPEC_XXSPLTIW): Move from altivec.md.
5491 (UNSPEC_XXSPLTID): Move from altivec.md.
5492 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5493 (UNSPEC_XXBLEND): Move from altivec.md.
5494 (UNSPEC_XXPERMX): Move from altivec.md.
5495 (VM3): Move from altivec.md.
5496 (VM3_char): Move from altivec.md.
5497 (xxspltiw_v4si): Move from altivec.md.
5498 (xxspltiw_v4sf): Move from altivec.md.
5499 (xxspltiw_v4sf_inst): Move from altivec.md.
5500 (xxspltidp_v2df): Move from altivec.md.
5501 (xxspltidp_v2df_inst): Move from altivec.md.
5502 (xxsplti32dx_v4si_inst): Move from altivec.md.
5503 (xxsplti32dx_v4sf): Move from altivec.md.
5504 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5505 (xxblend_<mode>): Move from altivec.md.
5506 (xxpermx): Move from altivec.md.
5507 (xxpermx_inst): Move from altivec.md.
5509 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5511 * tree-vect-generic.c (expand_vector_operations_1): Use either
5512 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5513 when constructing scalar splat expressions.
5515 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5518 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5519 pointer to __vector_pair *.
5521 2021-08-19 Martin Sebor <msebor@redhat.com>
5523 * gimple-range.cc: Add comments.
5524 * gimple-range.h: Same.
5526 2021-08-19 Martin Sebor <msebor@redhat.com>
5528 PR middle-end/101984
5529 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5532 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5534 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5535 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5536 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5538 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5540 PR middle-end/101950
5541 * optabs.c (expand_clrsb_using_clz): New function.
5542 (expand_unop): Use it as another clrsb expansion fallback.
5544 2021-08-19 liuhongt <hongtao.liu@intel.com>
5547 2021-07-28 liuhongt <hongtao.liu@intel.com>
5550 * config/i386/i386.h (processor_costs): Add new member
5552 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5553 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5554 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5555 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5556 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5557 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5558 generic_cost, core_cost): Initialize integer_to_sse same value
5560 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5561 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5562 Use integer_to_sse instead of sse_op to calculate the cost of
5565 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5567 * config.gcc: Include rpath.opt for Darwin.
5568 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5570 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5573 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5576 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5578 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5579 pattern and type cast.
5580 (vld4_lane_f32): Use float RTL pattern.
5581 (vld4q_lane_f64): Use float type cast.
5583 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5585 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5588 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5590 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5591 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5592 (hash_map_tests_c_tests): Call it.
5594 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5596 * ggc.h (enum ggc_collect): New.
5597 (ggc_collect): Use it.
5598 * ggc-page.c: Adjust.
5599 * ggc-common.c: Likewise.
5600 * ggc-tests.c: Likewise.
5601 * read-rtl-function.c: Likewise.
5602 * selftest-run-tests.c: Likewise.
5603 * doc/gty.texi (Invoking the garbage collector): Likewise.
5605 2021-08-18 liuhongt <hongtao.liu@intel.com>
5608 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
5610 * config/i386/sse.md (*sse3_haddv2df3_low): Add
5611 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
5612 (*sse3_hsubv2df3_low): Ditto.
5613 * config/i386/x86-tune.def
5614 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
5616 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5618 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
5619 (gori_compute::compute_operand_range): Add tracing.
5620 (gori_compute::logical_combine): Ditto.
5621 (gori_compute::compute_logical_operands): Ditto.
5622 (gori_compute::compute_operand1_range): Ditto.
5623 (gori_compute::compute_operand2_range): Ditto.
5624 (gori_compute::outgoing_edge_range_p): Ditto.
5625 * gimple-range-gori.h (class gori_compute): Add range_tracer.
5627 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5629 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
5630 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
5631 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
5632 * params.opt (--param=evrp-mode): Adjust options.
5634 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5636 * Makefile.in (OBJS): Add gimple-range-trace.o.
5637 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
5638 * gimple-range-fold.cc: Adjust headers.
5639 * gimple-range-trace.cc: New.
5640 * gimple-range-trace.h: New.
5641 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
5642 (gimple_ranger::range_of_expr): Add tracing.
5643 (gimple_ranger::range_on_entry): Ditto.
5644 (gimple_ranger::range_on_exit): Ditto.
5645 (gimple_ranger::range_on_edge): Ditto.
5646 (gimple_ranger::fold_range_internal): Ditto.
5647 (gimple_ranger::dump_bb): Do not calculate edge range twice.
5648 (trace_ranger::*): Remove.
5649 (enable_ranger): Never create a trace_ranger.
5650 (debug_seed_ranger): Move to gimple-range-trace.cc.
5651 (dump_ranger): Ditto.
5652 (debug_ranger): Ditto.
5653 * gimple-range.h: Include gimple-range-trace.h.
5654 (range_on_entry, range_on_exit): No longer virtual.
5655 (class trace_ranger): Remove.
5656 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
5658 2021-08-17 Martin Sebor <msebor@redhat.com>
5660 PR middle-end/101854
5661 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
5662 in gimple-ssa-warn-access.cc.
5663 * calls.c (alloc_max_size): Move code to check_alloca.
5664 (get_size_range): Move to pointer-query.cc.
5665 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
5666 (get_attr_nonstring_decl): Move to tree.c.
5667 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
5668 (append_attrname): Same.
5669 (maybe_warn_rdwr_sizes): Same.
5670 (initialize_argument_information): Move code to
5671 gimple-ssa-warn-access.cc.
5672 * calls.h (maybe_warn_alloc_args_overflow): Move to
5673 gimple-ssa-warn-access.h.
5674 (get_attr_nonstring_decl): Move to tree.h.
5675 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
5676 (enum size_range_flags): Move to pointer-query.h.
5677 (get_size_range): Same.
5678 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
5679 to avoid Clang -Wunused-function.
5680 (get_size_range): Declare static.
5681 (maybe_emit_free_warning): Rename...
5682 (maybe_check_dealloc_call): ...to this for consistency.
5683 (class pass_waccess): Add members.
5684 (pass_waccess::~pass_waccess): Defined.
5685 (alloc_max_size): Move here from calls.c.
5686 (maybe_warn_alloc_args_overflow): Same.
5687 (check_alloca): New function.
5688 (check_alloc_size_call): New function.
5689 (check_strncat): Handle another warning flag.
5690 (pass_waccess::check_builtin): Handle alloca.
5691 (fntype_argno_type): Move here from calls.c.
5692 (append_attrname): Same.
5693 (maybe_warn_rdwr_sizes): Same.
5694 (pass_waccess::check_call): Define.
5695 (check_nonstring_args): New function.
5696 (pass_waccess::check): Call new member functions.
5697 (pass_waccess::execute): Enable ranger.
5698 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
5699 (maybe_warn_nonstring_arg): Same.
5700 * gimple-ssa-warn-restrict.c: Remove #include.
5701 * pointer-query.cc (get_size_range): Move here from calls.c.
5702 * pointer-query.h (enum size_range_flags): Same.
5703 (get_size_range): Same.
5704 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
5705 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
5707 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
5709 * ggc.h (ggc_collect): Add 'force_collect' parameter.
5710 * ggc-page.c (ggc_collect): Use that one instead of global
5711 'ggc_force_collect'. Adjust all users.
5712 * doc/gty.texi (Invoking the garbage collector): Update.
5713 * ggc-internal.h (ggc_force_collect): Remove.
5714 * ggc-common.c (ggc_force_collect): Likewise.
5715 * selftest.h (forcibly_ggc_collect): Remove.
5716 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
5717 * read-rtl-function.c (test_loading_labels): Adjust.
5718 * selftest-run-tests.c (run_tests): Likewise.
5720 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5722 * config/darwin.c (darwin_file_end): Reset and reclaim the
5723 section names table at the end of compile.
5725 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5728 * config.in: Regenerate.
5729 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
5730 (ASM_SPEC): Pass options to disable branch shortening where
5732 * configure: Regenerate.
5733 * configure.ac: Detect versions of 'as' that support the
5734 optimisation which has the bug.
5736 2021-08-17 Richard Biener <rguenther@suse.de>
5738 * optabs-query.c (supports_vec_gather_load_p): Also check
5740 (supports_vec_scatter_store_p): Likewise.
5741 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
5742 back to masked variants if non-masked are not supported.
5743 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
5744 When we need to use masked gather/scatter but do not have
5745 a mask set up a constant true one.
5746 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
5749 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5751 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
5752 algorithm as PLUS_EXPR to improve subtraction bit bounds.
5753 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
5755 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5757 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
5758 calculate the mask-value pair result of a multiplication by an
5760 (bit_value_binop) [MULT_EXPR]: Call it from here for
5761 multiplications by (sparse) non-negative constants.
5763 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
5766 * config.gcc (gcc_cv_initfini_array): Leave undefined for
5767 uclinuxfdpiceabi targets.
5769 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5771 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
5772 reverse debug stmts.
5774 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5776 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
5778 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5780 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
5781 (__LD4Q_LANE_FUNC): Likewise.
5782 (vld4_lane_u8): Define without macro.
5783 (vld4_lane_u16): Likewise.
5784 (vld4_lane_u32): Likewise.
5785 (vld4_lane_u64): Likewise.
5786 (vld4_lane_s8): Likewise.
5787 (vld4_lane_s16): Likewise.
5788 (vld4_lane_s32): Likewise.
5789 (vld4_lane_s64): Likewise.
5790 (vld4_lane_f16): Likewise.
5791 (vld4_lane_f32): Likewise.
5792 (vld4_lane_f64): Likewise.
5793 (vld4_lane_p8): Likewise.
5794 (vld4_lane_p16): Likewise.
5795 (vld4_lane_p64): Likewise.
5796 (vld4q_lane_u8): Likewise.
5797 (vld4q_lane_u16): Likewise.
5798 (vld4q_lane_u32): Likewise.
5799 (vld4q_lane_u64): Likewise.
5800 (vld4q_lane_s8): Likewise.
5801 (vld4q_lane_s16): Likewise.
5802 (vld4q_lane_s32): Likewise.
5803 (vld4q_lane_s64): Likewise.
5804 (vld4q_lane_f16): Likewise.
5805 (vld4q_lane_f32): Likewise.
5806 (vld4q_lane_f64): Likewise.
5807 (vld4q_lane_p8): Likewise.
5808 (vld4q_lane_p16): Likewise.
5809 (vld4q_lane_p64): Likewise.
5810 (vld4_lane_bf16): Likewise.
5811 (vld4q_lane_bf16): Likewise.
5813 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5815 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
5816 (__LD3Q_LANE_FUNC): Delete.
5817 (vld3_lane_u8): Define without macro.
5818 (vld3_lane_u16): Likewise.
5819 (vld3_lane_u32): Likewise.
5820 (vld3_lane_u64): Likewise.
5821 (vld3_lane_s8): Likewise.
5822 (vld3_lane_s16): Likewise.
5823 (vld3_lane_s32): Likewise.
5824 (vld3_lane_s64): Likewise.
5825 (vld3_lane_f16): Likewise.
5826 (vld3_lane_f32): Likewise.
5827 (vld3_lane_f64): Likewise.
5828 (vld3_lane_p8): Likewise.
5829 (vld3_lane_p16): Likewise.
5830 (vld3_lane_p64): Likewise.
5831 (vld3q_lane_u8): Likewise.
5832 (vld3q_lane_u16): Likewise.
5833 (vld3q_lane_u32): Likewise.
5834 (vld3q_lane_u64): Likewise.
5835 (vld3q_lane_s8): Likewise.
5836 (vld3q_lane_s16): Likewise.
5837 (vld3q_lane_s32): Likewise.
5838 (vld3q_lane_s64): Likewise.
5839 (vld3q_lane_f16): Likewise.
5840 (vld3q_lane_f32): Likewise.
5841 (vld3q_lane_f64): Likewise.
5842 (vld3q_lane_p8): Likewise.
5843 (vld3q_lane_p16): Likewise.
5844 (vld3q_lane_p64): Likewise.
5845 (vld3_lane_bf16): Likewise.
5846 (vld3q_lane_bf16): Likewise.
5848 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5850 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
5851 (__LD2Q_LANE_FUNC): Likewise.
5852 (vld2_lane_u8): Define without macro.
5853 (vld2_lane_u16): Likewise.
5854 (vld2_lane_u32): Likewise.
5855 (vld2_lane_u64): Likewise.
5856 (vld2_lane_s8): Likewise.
5857 (vld2_lane_s16): Likewise.
5858 (vld2_lane_s32): Likewise.
5859 (vld2_lane_s64): Likewise.
5860 (vld2_lane_f16): Likewise.
5861 (vld2_lane_f32): Likewise.
5862 (vld2_lane_f64): Likewise.
5863 (vld2_lane_p8): Likewise.
5864 (vld2_lane_p16): Likewise.
5865 (vld2_lane_p64): Likewise.
5866 (vld2q_lane_u8): Likewise.
5867 (vld2q_lane_u16): Likewise.
5868 (vld2q_lane_u32): Likewise.
5869 (vld2q_lane_u64): Likewise.
5870 (vld2q_lane_s8): Likewise.
5871 (vld2q_lane_s16): Likewise.
5872 (vld2q_lane_s32): Likewise.
5873 (vld2q_lane_s64): Likewise.
5874 (vld2q_lane_f16): Likewise.
5875 (vld2q_lane_f32): Likewise.
5876 (vld2q_lane_f64): Likewise.
5877 (vld2q_lane_p8): Likewise.
5878 (vld2q_lane_p16): Likewise.
5879 (vld2q_lane_p64): Likewise.
5880 (vld2_lane_bf16): Likewise.
5881 (vld2q_lane_bf16): Likewise.
5883 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5885 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
5888 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5890 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
5891 (rank_for_schedule): Use it.
5893 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5895 PR rtl-optimization/91598
5896 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
5897 insns after memory reads and before memory writes.
5899 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
5901 * rtl.h (CONST_VECTOR_P): New macro.
5902 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
5903 code testing macros.
5904 (aarch64_ptrue_all_mode): Likewise.
5905 (aarch64_expand_mov_immediate): Likewise.
5906 (aarch64_const_vec_all_in_range_p): Likewise.
5907 (aarch64_rtx_costs): Likewise.
5908 (aarch64_legitimate_constant_p): Likewise.
5909 (aarch64_simd_valid_immediate): Likewise.
5910 (aarch64_simd_make_constant): Likewise.
5911 (aarch64_convert_mult_to_shift): Likewise.
5912 (aarch64_expand_sve_vec_perm): Likewise.
5913 (aarch64_vec_fpconst_pow_of_2): Likewise.
5915 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5917 PR tree-optimization/101938
5918 * range-op.cc (operator_abs::op1_range): Special case
5919 -TYPE_MIN_VALUE for flag_wrapv.
5921 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
5923 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
5926 2021-08-17 Jakub Jelinek <jakub@redhat.com>
5928 * tree.def (OMP_SCOPE): New tree code.
5929 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
5930 * tree-nested.c (convert_nonlocal_reference_stmt,
5931 convert_local_reference_stmt, convert_gimple_call): Handle
5933 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
5934 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
5935 * gimple.c (gimple_build_omp_scope): New function.
5936 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
5937 * gimple.h (gimple_build_omp_scope): Declare.
5938 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
5939 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
5940 gimple_omp_scope_set_clauses): New inline functions.
5941 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
5942 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
5943 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
5944 * gimple-walk.c (walk_gimple_stmt): Likewise.
5945 * gimple-low.c (lower_stmt): Likewise.
5946 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
5947 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
5948 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
5949 allowing task reductions. Reject inscan reductions on scope.
5950 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
5951 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
5952 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
5953 (estimate_num_insns): Likewise.
5954 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
5955 contexts if var isn't privatized there.
5956 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
5957 (scan_omp_1_stmt): Likewise.
5958 (maybe_add_implicit_barrier_cancel): Look through outer
5960 (lower_omp_scope): New function.
5961 (lower_omp_task_reductions): Handle OMP_SCOPE.
5962 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
5963 (diagnose_sb_1, diagnose_sb_2): Likewise.
5964 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
5965 (expand_omp): Handle GIMPLE_OMP_SCOPE.
5966 (omp_make_gimple_edges): Likewise.
5967 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
5969 2021-08-17 Richard Biener <rguenther@suse.de>
5971 PR tree-optimization/101925
5972 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
5973 reverse on COMPONENT_REF and ARRAY_REF according to
5974 what reverse_storage_order_for_component_p does.
5975 (vn_reference_eq): Compare reversed on reference ops.
5976 (reverse_storage_order_for_component_p): New overload.
5977 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
5978 on the reference looked up.
5980 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
5982 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
5984 (h8300_option_override): Use loops on H8/S more often when optimizing
5986 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
5987 accordingly. Handle various arithmetic right shifts with special
5988 sequences that we couldn't handle before.
5990 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
5992 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
5994 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
5996 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
5997 * config/sparc/sparc.c (sparc_gcov_type_size): New.
5998 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
5999 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
6000 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
6001 * doc/tm.texi.in: Regenerate.
6002 * target.def (gcov_type_size): New target hook.
6003 * targhooks.c (default_gcov_type_size): New.
6004 * targhooks.h (default_gcov_type_size): Declare.
6005 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
6007 (gimple_gen_time_profiler): Likewise.
6009 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6011 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
6013 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
6015 PR middle-end/101931
6016 * omp-low.c (omp_runtime_api_call): Update for routines
6017 added in the meanwhile.
6019 2021-08-16 Martin Liska <mliska@suse.cz>
6021 PR tree-optimization/100393
6022 * tree-switch-conversion.c (group_cluster::dump): Use
6023 get_comparison_count.
6024 (jump_table_cluster::find_jump_tables): Pre-compute number of
6025 comparisons and then decrement it. Cache also max_ratio.
6026 (jump_table_cluster::can_be_handled): Change signature.
6027 * tree-switch-conversion.h (get_comparison_count): New.
6029 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6031 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
6032 encodings only when -fgnat-encodings=all is specified.
6033 (add_bound_info): Likewise.
6034 (add_byte_size_attribute): Likewise.
6035 (gen_member_die): Likewise.
6037 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6039 * omp-oacc-neuter-broadcast.cc
6040 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
6042 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6044 * omp-oacc-neuter-broadcast.cc
6045 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
6048 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6050 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
6051 (execute_omp_oacc_neuter_broadcast): ... here.
6052 (install_var_field, build_receiver_ref, build_sender_ref): Take
6053 'field_map_t *' parameter. Adjust all users.
6054 (worker_single_copy, neuter_worker_single): Take a
6055 'record_field_map_t *' parameter. Adjust all users.
6057 2021-08-16 liuhongt <hongtao.liu@intel.com>
6060 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
6063 2021-08-16 Martin Liska <mliska@suse.cz>
6066 * multiple_target.c (create_dispatcher_calls): Make default
6067 function local only if it is a definition.
6069 2021-08-16 Martin Liska <mliska@suse.cz>
6072 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
6073 consider equal SSA_NAMEs when one is a param.
6075 2021-08-16 liuhongt <hongtao.liu@intel.com>
6078 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
6079 Support vpermi2b for V32QI/V16QImode.
6080 (ix86_extract_perm_from_pool_constant): New function.
6081 (ix86_expand_vec_one_operand_perm_avx512): Support
6082 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
6083 (expand_vec_perm_1): Adjust comments for upper.
6084 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
6086 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
6087 (pshufb_truncv4siv4hi_operand): Ditto.
6088 (pshufb_truncv8hiv8qi_operand): Ditto.
6089 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
6090 New pre_reload define_insn_and_split.
6091 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
6092 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
6093 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
6094 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
6095 (*avx512f_permvar_truncv4div4si_1): Ditto.
6096 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
6097 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
6098 (*avx512f_pshufd_truncv2div2si_1): Ditto.
6100 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
6102 * config/riscv/multilib-generator: Support code model option for
6104 * doc/install.texi: Add document of new option for
6105 --with-multilib-generator.
6107 2021-08-15 Clément Chigot <clement.chigot@atos.net>
6109 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
6110 (rs6000_legitimize_tls_address_aix): Use it.
6111 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
6112 xcoff_tls_exec_model_detected is true.
6114 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
6116 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6117 to allow a bit more code growth, saving many dozens of cycles.
6118 (h8300_option_override): Adjus shift_alg_si if optimizing for
6120 (get_shift_alg): Use special + inline shifts for residuals
6123 2021-08-14 Stafford Horne <shorne@gmail.com>
6126 * config/or1k/or1k-opts.h: New file.
6127 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6128 Support generating gotha relocations if -mcmodel=large is
6130 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6132 * config/or1k/or1k.opt (mcmodel=): New option.
6133 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6135 2021-08-14 Martin Sebor <msebor@redhat.com>
6137 PR middle-end/101791
6138 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6139 to valid_new_delete_pair_p.
6140 * tree.c (valid_new_delete_pair_p): Add argument.
6141 * tree.h (valid_new_delete_pair_p): Same.
6143 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6146 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6147 <case E_V64QImode>: For this mode assert
6148 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6150 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6153 * config/rs6000/altivec.md (xxeval): Use register_predicate
6154 instead of altivec_register_predicate.
6156 2021-08-13 Martin Sebor <msebor@redhat.com>
6158 PR middle-end/101734
6159 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6160 (maybe_warn_operand): Call it.
6162 2021-08-13 Martin Liska <mliska@suse.cz>
6165 * attribs.c (decl_attributes): Make naked functions "noipa"
6168 2021-08-13 Martin Liska <mliska@suse.cz>
6171 * symtab.c (symtab_node::noninterposable_alias): Do not create
6172 local aliases for target_clone functions as the clonning pass
6175 2021-08-13 Martin Liska <mliska@suse.cz>
6177 * opts.c (LIVE_PATCHING_OPTION): Define.
6178 (control_options_for_live_patching): Use it in error messages.
6180 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6182 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6183 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6184 ignore_stores_eaf_flags): New constants.
6185 (remove_useless_eaf_flags): New function.
6186 (eaf_flags_useful_p): Use it.
6187 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6189 (modref_lattice::init): Add EAF_NOREAD.
6190 (modref_lattice::add_escape_point): Do not reacord escape point if
6192 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6193 use remove_useless_eaf_flags.
6194 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6195 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6196 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6197 (analyze_parms): Use remove_useless_eaf_flags.
6198 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6199 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6200 use remove_useless_eaf_flags.
6201 (modref_propagate_flags_in_scc): Update.
6202 * ipa-modref.h: Turn eaf_flags_t back to char.
6203 * tree-core.h (EAF_NOT_RETURNED): Fix.
6204 (EAF_NOREAD): New constant
6205 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6207 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6208 (handle_pure_call): Likewise.
6210 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6212 * tree.def (OMP_MASKED): New tree code.
6213 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6214 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6215 OMP_CLAUSE_FILTER_EXPR): Define.
6216 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6217 (omp_clause_code_name): Likewise.
6218 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6219 * tree-nested.c (convert_nonlocal_omp_clauses,
6220 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6221 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6222 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6223 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6224 (dump_generic_node): Handle OMP_MASTER.
6225 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6226 * gimple.c (gimple_build_omp_masked): New function.
6227 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6228 * gimple.h (gimple_build_omp_masked): Declare.
6229 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6230 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6231 gimple_omp_masked_set_clauses): New inline functions.
6232 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6233 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6234 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6235 * gimple-walk.c (walk_gimple_stmt): Likewise.
6236 * gimple-low.c (lower_stmt): Likewise.
6237 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6238 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
6239 that take one expression rather than decl or constant, force
6240 gimplification of that into a SSA_NAME or temporary unless min
6242 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
6243 (gimplify_expr): Handle OMP_MASKED.
6244 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
6245 (estimate_num_insns): Likewise.
6246 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
6247 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
6248 diagnostics for existence of masked construct.
6249 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
6250 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
6251 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
6254 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
6257 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
6258 (ldexp<mode>3): Use avx512f_scalef<mode>2.
6259 (UNSPEC_SCALEF): Move from ...
6260 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
6262 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
6264 * ipa-split.c (consider_split): Fix condition testing void functions.
6266 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6268 * doc/invoke.texi: Remove docs for threader-mode param.
6269 * flag-types.h (enum threader_mode): Remove.
6270 * params.opt: Remove threader-mode param.
6271 * tree-ssa-threadbackward.c (class back_threader): Remove
6272 path_is_unreachable_p.
6273 Make find_paths private.
6274 Add maybe_thread and thread_through_all_blocks.
6275 Remove reference marker for m_registry.
6276 Remove reference marker for m_profit.
6277 (back_threader::back_threader): Adjust for registry and profit not
6279 (dump_path): Move down.
6281 (class thread_jumps): Remove.
6282 (class back_threader_registry): Remove m_all_paths.
6284 (thread_jumps::thread_through_all_blocks): Move to back_threader
6286 (fsm_find_thread_path): Remove
6287 (back_threader::maybe_thread): New.
6288 (back_threader::thread_through_all_blocks): Move from
6290 (back_threader_registry::back_threader_registry): Remove
6292 (back_threader_registry::~back_threader_registry): Remove.
6293 (thread_jumps::find_taken_edge): Remove.
6294 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6295 (thread_jumps::maybe_register_path): Remove.
6296 (thread_jumps::handle_phi): Remove.
6297 (handle_assignment_p): Remove.
6298 (thread_jumps::handle_assignment): Remove.
6299 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6300 (thread_jumps::find_jump_threads_backwards): Remove.
6301 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6302 (try_thread_blocks): Rename find_jump_threads_backwards to
6304 (pass_early_thread_jumps::execute): Same.
6306 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6308 * tree-core.h (omp_clause_proc_bind_kind): Add
6309 OMP_CLAUSE_PROC_BIND_PRIMARY.
6310 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6311 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6313 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6315 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6316 fno-common reference.
6317 * config/arc/arc.c (arc_override_options): Remove overriding of
6320 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6323 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6324 If d->testing_p, return true after performing checks instead of
6325 actually expanding the insn.
6326 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6327 !TARGET_AVX512BW and return false.
6329 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6331 * configure.ac (PE linker --disable-dynamicbase support): New check.
6332 * configure: Regenerate.
6333 * config.in: Likewise.
6334 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6335 (LINK_SPEC): Use it.
6336 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6337 (LINK_SPEC): Likewise.
6339 2021-08-12 liuhongt <hongtao.liu@intel.com>
6342 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6343 post_reload define_insn_and_split.
6344 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6345 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6346 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6347 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6348 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6349 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6350 (*avx2_zero_extendv4siv4di2_2): Ditto.
6351 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6352 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6355 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6357 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6360 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6362 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6364 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6366 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6367 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6368 initialization of pcvoid_type_node here...
6369 (altivec_init_builtins): ...from here.
6370 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6371 RS6000_BTI_const_ptr_void.
6372 (pcvoid_type_node): New macro.
6374 2021-08-11 Richard Biener <rguenther@suse.de>
6377 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6378 hard-register accesses.
6380 2021-08-11 Richard Biener <rguenther@suse.de>
6382 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6383 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6384 to determine has_volatile_ops.
6386 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6388 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6390 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6391 Alexandre Oliva <oliva@adacore.com>
6393 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6395 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6396 (update_escape_summary_1): Likewise.
6397 (update_escape_summary): Likewise.
6398 (ipa_merge_modref_summary_after_inlining): Likewise.
6400 2021-08-11 Richard Biener <rguenther@suse.de>
6402 PR middle-end/101858
6403 * fold-const.c (fold_binary_loc): Guard simplification
6404 of X < (cast) (1 << Y) to integer types.
6406 2021-08-11 Richard Biener <rguenther@suse.de>
6408 PR tree-optimization/101861
6409 * tree-vect-stmts.c (vectorizable_load): Fix error in
6410 previous change with regard to gather vectorization.
6412 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6415 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6417 (vdup_n_s16): Likewise.
6418 (vdup_n_s32): Likewise.
6419 (vdup_n_s64): Likewise.
6420 (vdup_n_u8): Likewise.
6421 (vdup_n_u16): Likewise.
6422 (vdup_n_u32): Likewise.
6423 (vdup_n_u64): Likewise.
6424 (vdup_n_p8): Likewise.
6425 (vdup_n_p16): Likewise.
6426 (vdup_n_p64): Likewise.
6427 (vdup_n_f16): Likewise.
6428 (vdup_n_f32): Likewise.
6429 (vdupq_n_s8): Likewise.
6430 (vdupq_n_s16): Likewise.
6431 (vdupq_n_s32): Likewise.
6432 (vdupq_n_s64): Likewise.
6433 (vdupq_n_u8): Likewise.
6434 (vdupq_n_u16): Likewise.
6435 (vdupq_n_u32): Likewise.
6436 (vdupq_n_u64): Likewise.
6437 (vdupq_n_p8): Likewise.
6438 (vdupq_n_p16): Likewise.
6439 (vdupq_n_p64): Likewise.
6440 (vdupq_n_f16): Likewise.
6441 (vdupq_n_f32): Likewise.
6442 (vmov_n_s8): Replace call to builtin with call to corresponding
6444 (vmov_n_s16): Likewise.
6445 (vmov_n_s32): Likewise.
6446 (vmov_n_s64): Likewise.
6447 (vmov_n_u8): Likewise.
6448 (vmov_n_u16): Likewise.
6449 (vmov_n_u32): Likewise.
6450 (vmov_n_u64): Likewise.
6451 (vmov_n_p8): Likewise.
6452 (vmov_n_p16): Likewise.
6453 (vmov_n_f16): Likewise.
6454 (vmov_n_f32): Likewise.
6455 (vmovq_n_s8): Likewise.
6456 (vmovq_n_s16): Likewise.
6457 (vmovq_n_s32): Likewise.
6458 (vmovq_n_s64): Likewise.
6459 (vmovq_n_u8): Likewise.
6460 (vmovq_n_u16): Likewise.
6461 (vmovq_n_u32): Likewise.
6462 (vmovq_n_u64): Likewise.
6463 (vmovq_n_p8): Likewise.
6464 (vmovq_n_p16): Likewise.
6465 (vmovq_n_f16): Likewise.
6466 (vmovq_n_f32): Likewise.
6467 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6469 2021-08-11 liuhongt <hongtao.liu@intel.com>
6472 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6473 when TARGET_AVX512F and TARGET_SSE_MATH.
6475 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6478 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6479 for V32HImode if !TARGET_AVX512BW.
6480 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6481 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6482 early, but actually check the permutation.
6484 2021-08-10 Richard Biener <rguenther@suse.de>
6486 PR tree-optimization/101809
6487 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6488 gathers with offset vector nunits being a constant multiple
6489 of the data vector nunits.
6490 (vect_get_gather_scatter_ops): Use the appropriate nunits
6491 for the offset vector defs.
6492 (vectorizable_store): Adjust call to
6493 vect_get_gather_scatter_ops.
6494 (vectorizable_load): Likewise. Handle the case of less
6495 offset vectors than data vectors.
6497 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6500 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6501 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6504 2021-08-10 Richard Biener <rguenther@suse.de>
6506 PR tree-optimization/101801
6507 PR tree-optimization/101819
6508 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6509 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6510 (vectorizable_reduction): Re-instantiate a check for emulated
6512 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6513 (vectorizable_operation): Likewise. Cost emulated vector
6514 operations according to the scalar sequence synthesized by
6517 2021-08-10 Richard Biener <rguenther@suse.de>
6519 PR middle-end/101824
6520 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6521 volatile in case the variable was.
6523 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6526 * config/i386/constraints.md (BC): Document for integer SSE
6527 constant all bits set operand.
6528 (BF): New constraint for const floating-point all bits set
6530 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6531 (standard_sse_constant_opcode): Likewise.
6532 * config/i386/sse.md (sseconstm1): New mode attribute.
6533 (mov<mode>_internal): Replace BC with <sseconstm1>.
6535 2021-08-10 liuhongt <hongtao.liu@intel.com>
6537 * config/i386/sse.md (cond_<insn><mode>): New expander.
6538 (VI248_AVX512VLBW): New mode iterator.
6539 * config/i386/predicates.md
6540 (nonimmediate_or_const_vec_dup_operand): New predicate.
6542 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6544 PR tree-optimization/101741
6545 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6546 type of parameter for toupper/tolower.
6548 2021-08-09 Martin Jambor <mjambor@suse.cz>
6551 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6553 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6555 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6557 (is_store_insn1): Verify source is a register.
6559 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6562 * config/i386/mmx.md (<any_logic:code>v2sf3):
6563 Rename from *mmx_<any_logic:code>v2sf3
6565 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6567 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6568 'gcc/omp-oacc-neuter-broadcast.cc'.
6569 * omp-low.c: Likewise.
6570 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6573 2021-08-09 Julian Brown <julian@codesourcery.com>
6574 Kwok Cheung Yeung <kcy@codesourcery.com>
6575 Thomas Schwinge <thomas@codesourcery.com>
6577 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6578 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6579 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6580 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6581 (gcn_fork_join): Update comment.
6582 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6583 (macc_experimental_workers): Remove unused option.
6585 2021-08-09 Julian Brown <julian@codesourcery.com>
6586 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6587 Kwok Cheung Yeung <kcy@codesourcery.com>
6588 Thomas Schwinge <thomas@codesourcery.com>
6590 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6591 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6592 Add documentation hook.
6593 * doc/tm.texi: Regenerate.
6594 * omp-oacc-neuter-broadcast.cc: New file.
6595 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6596 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6597 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6598 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
6599 * target.def (goacc.create_worker_broadcast_record): Add target
6601 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
6603 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
6604 Rename prototype to...
6605 (gcn_goacc_create_worker_broadcast_record): ... this.
6606 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
6608 (gcn_goacc_create_worker_broadcast_record): ... this.
6609 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
6611 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
6613 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
6616 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
6619 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6621 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
6623 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6625 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
6626 consider '-foffload-abi'.
6627 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
6628 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
6631 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6633 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
6636 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6638 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
6640 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6642 * doc/gty.texi (Files): Update.
6644 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6646 * doc/gty.texi (Files): Fix GTY header file example.
6648 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
6650 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
6651 determine the upper and lower bounds from a mask-value pair.
6652 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
6653 absolute value and unsigned absolute value expressions.
6654 (bit_value_binop): Initialize *VAL's precision.
6655 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
6656 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
6657 support when the operands are unknown but potentially equal.
6658 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
6660 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
6662 * config/aarch64/aarch64.md
6663 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
6665 2021-08-08 Sergei Trofimovich <siarheit@google.com>
6667 * lra-constraints.c: Fix s/otput/output/ typo.
6669 2021-08-06 Martin Sebor <msebor@redhat.com>
6671 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
6672 (expand_builtin_strcat): Same.
6673 (expand_builtin_stpncpy): Same.
6674 (expand_builtin_strncat): Same.
6675 (check_read_access): Same.
6676 (check_memop_access): Same.
6677 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
6678 (expand_builtin_strnlen): Same.
6679 (expand_builtin_memcpy): Same.
6680 (expand_builtin_memmove): Same.
6681 (expand_builtin_mempcpy): Same.
6682 (expand_builtin_strcpy): Same.
6683 (expand_builtin_strcpy_args): Same.
6684 (expand_builtin_stpcpy_1): Same.
6685 (expand_builtin_strncpy): Same.
6686 (expand_builtin_memset): Same.
6687 (expand_builtin_bzero): Same.
6688 (expand_builtin_strcmp): Same.
6689 (expand_builtin_strncmp): Same.
6690 (expand_builtin): Remove handlers.
6691 (fold_builtin_strlen): Add a comment.
6692 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
6693 * calls.c (maybe_warn_nonstring_arg): Same.
6694 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
6695 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
6696 (gimple_fold_builtin_stpcpy): Same.
6697 * gimple-ssa-warn-access.cc (has_location): New function.
6698 (get_location): Same.
6699 (get_callee_fndecl): Same.
6702 (warn_string_no_nul): Define.
6703 (unterminated_array): Same.
6704 (check_nul_terminated_array): Same.
6705 (maybe_warn_nonstring_arg): Same.
6706 (maybe_warn_for_bound): Same.
6707 (warn_for_access): Same.
6708 (check_access): Same.
6709 (check_memop_access): Same.
6710 (check_read_access): Same.
6711 (warn_dealloc_offset): Use helper functions.
6712 (maybe_emit_free_warning): Same.
6713 (class pass_waccess): Add members.
6714 (check_strcat): New function.
6715 (check_strncat): New function.
6716 (check_stxcpy): New function.
6717 (check_stxncpy): New function.
6718 (check_strncmp): New function.
6719 (pass_waccess::check_builtin): New function.
6720 (pass_waccess::check): Call it.
6721 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
6723 (maybe_warn_for_bound): Same.
6724 (check_access): Same.
6725 (check_memop_access): Same.
6726 (check_read_access): Same.
6727 * pointer-query.h (struct access_data): Define a ctor overload.
6729 2021-08-06 Richard Biener <rguenther@suse.de>
6731 PR tree-optimization/101801
6732 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
6733 (vect_can_vectorize_without_simd_p): ... to this.
6734 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
6735 (vect_can_vectorize_without_simd_p): ... to this and fold
6736 in vect_min_worthwhile_factor.
6737 (vect_min_worthwhile_factor): Remove.
6738 (vectorizable_reduction): Adjust and remove the cost part.
6739 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6740 (vectorizable_operation): Likewise.
6742 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
6745 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
6746 Add general_gr_operand predicate to operand 3.
6748 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
6750 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
6751 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
6753 2021-08-06 Tamar Christina <tamar.christina@arm.com>
6755 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
6756 register_svprfop): Pass vec<> by pointer.
6757 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
6758 * langhooks.c (lhd_simulate_enum_decl): Likewise.
6759 * langhooks.h (struct lang_hooks_for_types): Likewise.
6761 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6763 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
6764 __builtin_memcpy instead of constructing an additional
6765 __builtin_aarch64_simd_oi one vector at a time.
6766 (vst1q_bf16_x2): Likewise.
6767 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
6768 an additional __builtin_aarch64_simd_ci one vector at a time.
6769 (vst1q_bf16_x3): Likewise.
6770 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
6771 (vst1q_bf16_x4): Likewise.
6772 (vst2_bf16): Use __builtin_memcpy instead of constructing an
6773 additional __builtin_aarch64_simd_oi one vector at a time.
6774 (vst2q_bf16): Likewise.
6775 (vst3_bf16): Use __builtin_memcpy instead of constructing an
6776 additional __builtin_aarch64_simd_ci mode one vector at a
6778 (vst3q_bf16): Likewise.
6779 (vst4_bf16): Use __builtin_memcpy instead of constructing an
6780 additional __builtin_aarch64_simd_xi one vector at a time.
6781 (vst4q_bf16): Likewise.
6783 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6785 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
6786 (__ST2Q_LANE_FUNC): Delete.
6787 (vst2_lane_f16): Use __builtin_memcpy to copy vector
6788 structure instead of constructing __builtin_aarch64_simd_oi
6789 one vector at a time.
6790 (vst2_lane_f32): Likewise.
6791 (vst2_lane_f64): Likewise.
6792 (vst2_lane_p8): Likewise.
6793 (vst2_lane_p16): Likewise.
6794 (vst2_lane_p64): Likewise.
6795 (vst2_lane_s8): Likewise.
6796 (vst2_lane_s16): Likewise.
6797 (vst2_lane_s32): Likewise.
6798 (vst2_lane_s64): Likewise.
6799 (vst2_lane_u8): Likewise.
6800 (vst2_lane_u16): Likewise.
6801 (vst2_lane_u32): Likewise.
6802 (vst2_lane_u64): Likewise.
6803 (vst2_lane_bf16): Likewise.
6804 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
6805 structure instead of using a union.
6806 (vst2q_lane_f32): Likewise.
6807 (vst2q_lane_f64): Likewise.
6808 (vst2q_lane_p8): Likewise.
6809 (vst2q_lane_p16): Likewise.
6810 (vst2q_lane_p64): Likewise.
6811 (vst2q_lane_s8): Likewise.
6812 (vst2q_lane_s16): Likewise.
6813 (vst2q_lane_s32): Likewise.
6814 (vst2q_lane_s64): Likewise.
6815 (vst2q_lane_u8): Likewise.
6816 (vst2q_lane_u16): Likewise.
6817 (vst2q_lane_u32): Likewise.
6818 (vst2q_lane_u64): Likewise.
6819 (vst2q_lane_bf16): Likewise.
6821 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6823 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
6824 (__ST3Q_LANE_FUNC): Delete.
6825 (vst3_lane_f16): Use __builtin_memcpy to copy vector
6826 structure instead of constructing __builtin_aarch64_simd_ci
6827 one vector at a time.
6828 (vst3_lane_f32): Likewise.
6829 (vst3_lane_f64): Likewise.
6830 (vst3_lane_p8): Likewise.
6831 (vst3_lane_p16): Likewise.
6832 (vst3_lane_p64): Likewise.
6833 (vst3_lane_s8): Likewise.
6834 (vst3_lane_s16): Likewise.
6835 (vst3_lane_s32): Likewise.
6836 (vst3_lane_s64): Likewise.
6837 (vst3_lane_u8): Likewise.
6838 (vst3_lane_u16): Likewise.
6839 (vst3_lane_u32): Likewise.
6840 (vst3_lane_u64): Likewise.
6841 (vst3_lane_bf16): Likewise.
6842 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
6843 structure instead of using a union.
6844 (vst3q_lane_f32): Likewise.
6845 (vst3q_lane_f64): Likewise.
6846 (vst3q_lane_p8): Likewise.
6847 (vst3q_lane_p16): Likewise.
6848 (vst3q_lane_p64): Likewise.
6849 (vst3q_lane_s8): Likewise.
6850 (vst3q_lane_s16): Likewise.
6851 (vst3q_lane_s32): Likewise.
6852 (vst3q_lane_s64): Likewise.
6853 (vst3q_lane_u8): Likewise.
6854 (vst3q_lane_u16): Likewise.
6855 (vst3q_lane_u32): Likewise.
6856 (vst3q_lane_u64): Likewise.
6857 (vst3q_lane_bf16): Likewise.
6859 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6861 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
6862 (__ST4Q_LANE_FUNC): Delete.
6863 (vst4_lane_f16): Use __builtin_memcpy to copy vector
6864 structure instead of constructing __builtin_aarch64_simd_xi
6865 one vector at a time.
6866 (vst4_lane_f32): Likewise.
6867 (vst4_lane_f64): Likewise.
6868 (vst4_lane_p8): Likewise.
6869 (vst4_lane_p16): Likewise.
6870 (vst4_lane_p64): Likewise.
6871 (vst4_lane_s8): Likewise.
6872 (vst4_lane_s16): Likewise.
6873 (vst4_lane_s32): Likewise.
6874 (vst4_lane_s64): Likewise.
6875 (vst4_lane_u8): Likewise.
6876 (vst4_lane_u16): Likewise.
6877 (vst4_lane_u32): Likewise.
6878 (vst4_lane_u64): Likewise.
6879 (vst4_lane_bf16): Likewise.
6880 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
6881 structure instead of using a union.
6882 (vst4q_lane_f32): Likewise.
6883 (vst4q_lane_f64): Likewise.
6884 (vst4q_lane_p8): Likewise.
6885 (vst4q_lane_p16): Likewise.
6886 (vst4q_lane_p64): Likewise.
6887 (vst4q_lane_s8): Likewise.
6888 (vst4q_lane_s16): Likewise.
6889 (vst4q_lane_s32): Likewise.
6890 (vst4q_lane_s64): Likewise.
6891 (vst4q_lane_u8): Likewise.
6892 (vst4q_lane_u16): Likewise.
6893 (vst4q_lane_u32): Likewise.
6894 (vst4q_lane_u64): Likewise.
6895 (vst4q_lane_bf16): Likewise.
6897 2021-08-06 Martin Liska <mliska@suse.cz>
6899 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
6900 a target option is restored, it can have
6901 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
6902 and error should not be emitted.
6904 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
6906 * gcov-io.h (gcov_write): Declare.
6907 * gcov-io.c (gcov_write): New.
6908 (gcov_write_counter): Remove.
6909 (gcov_write_tag_length): Likewise.
6910 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
6911 gcov_write_unsigned().
6912 * doc/invoke.texi (fprofile-info-section): Mention
6913 __gcov_info_to_gdca().
6915 2021-08-06 Martin Sebor <msebor@redhat.com>
6917 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
6918 arguments to by-reference.
6919 (iterate_fix_dominators): Same.
6920 * dominance.h (iterate_fix_dominators): Same.
6921 * ipa-prop.h: Call auto_vec::to_vec_legacy.
6922 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
6923 arguments to by-reference.
6924 (debug_data_dependence_relation): Same.
6925 (dump_data_dependence_relations): Same.
6926 * tree-data-ref.h (debug_data_dependence_relation): Same.
6927 (dump_data_dependence_relations): Same.
6928 * tree-predcom.c (dump_chains): Same.
6929 (initialize_root_vars_lm): Same.
6930 (determine_unroll_factor): Same.
6931 (replace_phis_by_defined_names): Same.
6932 (insert_init_seqs): Same.
6933 (pcom_worker::tree_predictive_commoning_loop): Call
6934 auto_vec::to_vec_legacy.
6935 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
6936 arguments to by-reference.
6937 * tree-ssa-threadbackward.c (populate_worklist): Same.
6938 (back_threader::resolve_def): Same.
6939 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
6940 (vect_enhance_data_refs_alignment): Same.
6941 (vect_check_lower_bound): Same.
6942 (vect_prune_runtime_alias_test_list): Same.
6943 (vect_permute_store_chain): Same.
6944 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
6945 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
6946 * tree-vectorizer.h (vect_permute_store_chain): Same.
6947 * vec.c (test_init): New function.
6948 (vec_c_tests): Call new function.
6949 * vec.h (vec): Declare ctors, dtor, and assignment.
6950 (auto_vec::vec_to_legacy): New function.
6951 (vec::copy): Adjust initialization.
6953 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
6956 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
6957 callee only uses GPRs.
6958 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
6959 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
6960 * config/i386/x86gprintrin.h: Add
6961 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
6962 to disable non-GPR ISAs.
6964 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
6966 PR middle-end/101787
6967 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
6969 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
6971 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
6972 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
6973 (vect_is_extending_load, vect_is_integer_truncation): New functions,
6974 moved from aarch64.c but given different names.
6975 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
6976 (aarch64_is_reduction, aarch64_reduc_type)
6977 (aarch64_embedded_comparison_type, aarch64_comparison_type)
6978 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
6979 in favor of the above. Update callers accordingly.
6981 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
6984 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
6985 writing .cpu directive in asm output.
6986 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
6987 (arm_last_printed_arch_string): Delete.
6988 (arm_last-printed_fpu_string): Delete.
6989 (arm_configure_build_target): If use of floating-point/SIMD is
6990 disabled, remove all fp/simd related features from the target ISA.
6991 (last_arm_targ_options): New variable.
6992 (arm_print_asm_arch_directives): Add new parameters. Change order
6993 of emitted directives and handle all cases here.
6994 (arm_file_start): Always call arm_print_asm_arch_directives, move
6995 all generation of .arch/.arch_extension here.
6996 (arm_file_end): Call arm_print_asm_arch.
6997 (arm_declare_function_name): Call arm_print_asm_arch_directives
6998 instead of printing .arch/.fpu directives directly.
7000 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7002 * config/arm/arm.c (arm_configure_build_target): Don't call
7003 arm_option_reconfigure_globals.
7004 (arm_option_restore): Call arm_option_reconfigure_globals after
7005 reconfiguring the target.
7006 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
7008 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7010 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
7011 arch_name is always set.
7013 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7015 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7016 of vec_select high-half from being added into Neon subtract
7019 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7021 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7022 of vec_select high-half from being added into Neon add cost.
7024 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
7026 * cfgloop.h (loops_list::loops_list): Add one optional argument
7027 root and adjust accordingly, update loop tree walking and factor
7029 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
7031 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
7033 PR tree-optimization/101626
7034 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
7035 reverse scalar storage order on a pointer or vector component.
7037 2021-08-05 liuhongt <hongtao.liu@intel.com>
7039 * config/i386/sse.md (cond_<code><mode>): New expander.
7041 2021-08-05 liuhongt <hongtao.liu@intel.com>
7043 * config/i386/sse.md (cond_<code><mode>): New expander.
7045 2021-08-05 liuhongt <hongtao.liu@intel.com>
7047 * config/i386/sse.md (cond_<code><mode>): New expander.
7049 2021-08-04 David Malcolm <dmalcolm@redhat.com>
7052 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
7054 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7057 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
7058 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
7060 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7063 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7064 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7065 data with SSE register from one memory location to another.
7067 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7069 * config/s390/s390.c (expand_perm_with_vpdi): New function.
7070 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
7071 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
7072 parameterized expander.
7073 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
7075 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7077 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
7078 (struct expand_vec_perm_d): Define struct.
7079 (expand_perm_with_merge): New function.
7080 (vectorize_vec_perm_const_1): New function.
7081 (s390_vectorize_vec_perm_const): New function.
7082 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
7084 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7086 * config/s390/vector.md (V_HW_64): Remove mode iterator.
7087 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
7088 * config/s390/vx-builtins.md
7089 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
7092 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7094 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
7096 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
7098 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
7099 instead of an unspec.
7101 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7103 * config/s390/s390-modes.def: Add more vector modes to support
7104 concatenation of two vectors.
7105 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
7107 (s390_expand_merge): Likewise.
7108 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
7109 (s390_expand_merge): New function.
7110 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
7111 Remove constant definitions.
7112 * config/s390/vector.md (V_HW_2): Add mode iterators.
7113 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
7114 (vec_2x_nelts, vec_2x_wide): New mode attributes.
7115 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
7116 New pattern definitions.
7117 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7118 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7119 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7120 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7122 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7124 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7125 emit vec merge pattern.
7127 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7129 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7131 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7132 vec_select high-half from being added into Neon multiply
7134 * rtlanal.c (vec_series_highpart_p): Define.
7135 * rtlanal.h (vec_series_highpart_p): Declare.
7137 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7139 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7141 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7142 vec_select cost from being added into Neon multiply cost.
7144 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7146 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7147 which old_loop_vinfo is an epilogue loop that handles a constant
7148 number of iterations.
7150 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7152 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7153 when a reanalyzed loop fails to be cheaper than the current
7156 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7158 * config/aarch64/aarch64.c: Fix a typo.
7160 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7162 PR gcov-profile/101773
7163 * gcov-io.c (gcov_close): Check return code of a fclose.
7165 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7168 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7169 .file statement when needed.
7171 2021-08-04 Richard Biener <rguenther@suse.de>
7173 * tree-vect-data-refs.c (vect_check_gather_scatter):
7174 Include widening conversions only when the result is
7175 still handed by native gather or the current offset
7176 size not already matches the data size.
7177 Also succeed analysis in case there's no native support,
7178 noted by a IFN_LAST ifn and a NULL decl.
7179 (vect_analyze_data_refs): Always consider gathers.
7180 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7181 Test for no IFN gather rather than decl gather.
7182 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7183 gather-scatter info and cost emulated gathers accordingly.
7184 (vect_truncate_gather_scatter_offset): Properly test for
7186 (vect_use_strided_gather_scatters_p): Likewise.
7187 (get_load_store_type): Handle emulated gathers and its
7189 (vectorizable_load): Likewise. Emulate them by extracting
7190 scalar offsets, doing scalar loads and a vector construct.
7192 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7195 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7196 argument to set m_max_size.
7197 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7198 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7199 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7201 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7202 Marc Glisse <marc.glisse@inria.fr>
7204 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7205 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7206 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7208 2021-08-04 Richard Biener <rguenther@suse.de>
7210 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7211 out code to decompose vector loads ...
7212 (optimize_vector_load): ... here. Generalize it to
7213 handle intermediate widening and TARGET_MEM_REF loads
7214 and apply it to loads with a supported vector mode as well.
7216 2021-08-04 Richard Biener <rguenther@suse.de>
7218 PR tree-optimization/101756
7219 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7220 the result of the reduction epilogue is compatible to the original
7223 2021-08-04 liuhongt <hongtao.liu@intel.com>
7226 * config/i386/i386.md (peephole2): Refine predicate from
7227 register_operand to general_reg_operand.
7229 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7231 * gimple-range-path.h (path_range_query::dump): Mark override.
7233 2021-08-04 Richard Biener <rguenther@suse.de>
7235 PR tree-optimization/101769
7236 * tree-tailcall.c (eliminate_tail_call): Add the created loop
7237 for the first recursion and return it via the new output parameter.
7238 (optimize_tail_call): Pass through new output param.
7239 (tree_optimize_tail_calls_1): After creating all latches,
7240 add the created loop to the loop tree. Do not mark loops for fixup.
7242 2021-08-04 Martin Liska <mliska@suse.cz>
7244 * doc/invoke.texi: Document threader-mode param.
7246 2021-08-04 liuhongt <hongtao.liu@intel.com>
7248 * config/i386/sse.md (cond_fma<mode>): New expander.
7249 (cond_fms<mode>): Ditto.
7250 (cond_fnma<mode>): Ditto.
7251 (cond_fnms<mode>): Ditto.
7253 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7255 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
7257 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7259 * config/rs6000/constraints.md: Remove "e" from the list of available
7260 constraint characters.
7262 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
7264 PR gcov-profile/71672
7265 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7267 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7269 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7271 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7273 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7274 try XMM31 to avoid vzeroupper.
7276 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7278 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7279 -mcpu=neoverse-512tvb.
7280 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7281 * config/aarch64/aarch64-tune.md: Regenerate.
7282 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7283 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7284 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7285 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7286 (aarch64_adjust_body_cost): Likewise.
7288 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7290 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7291 record issue information for operations that occur in the
7294 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7296 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7297 parameter. Detect cases in which an Advanced SIMD MLA would almost
7298 certainly require a MOV.
7299 (aarch64_count_ops): Update accordingly.
7301 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7303 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7304 function, split out from...
7305 (aarch64_detect_vector_stmt_subtype): ...here.
7306 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7308 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7310 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7311 Add gather_load_x32_cost and gather_load_x64_cost.
7312 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7313 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7314 accordingly, using the values given by the scalar_load * number
7315 of elements calculation that we used previously.
7316 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7318 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7320 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7321 function, split out from...
7322 (aarch64_adjust_body_cost): ...here.
7324 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7326 * config/aarch64/fractional-cost.h: New file.
7327 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7328 and cost_fraction.h.
7329 (vec_cost_fraction): New typedef.
7330 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7331 (aarch64_detect_vector_stmt_subtype): Likewise.
7332 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7333 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7335 (aarch64_adjust_body_cost): Likewise.
7336 (aarch64_test_cost_fraction): New function.
7337 (aarch64_run_selftests): Call it.
7339 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7341 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7343 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7345 (aarch64_estimated_poly_value): Likewise. Use the least significant
7346 set bit for the minimum and likely values. Use the most significant
7347 set bit for the maximum value.
7349 2021-08-03 liuhongt <hongtao.liu@intel.com>
7351 * config/i386/sse.md (cond_<insn><mode>): New expander.
7352 (cond_mul<mode>): Ditto.
7354 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7356 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7358 2021-08-03 liuhongt <hongtao.liu@intel.com>
7360 * config/i386/sse.md (cond_<insn><mode>):New expander.
7361 (cond_mul<mode>): Ditto.
7362 (cond_div<mode>): Ditto.
7364 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7366 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7367 check stack_realign_needed for stack realignment.
7368 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7369 than the largest integer supported by vector register.
7370 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7371 (MOVE_MAX): Set to bytes of the largest integer supported by
7373 (STORE_MAX_PIECES): New.
7375 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7377 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7378 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7379 data from one memory location to another.
7381 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7384 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7386 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7388 PR tree-optimization/101724
7389 * params.opt: Remove --param=threader-iterative.
7390 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7393 2021-08-02 Tom de Vries <tdevries@suse.de>
7395 PR middle-end/101665
7396 * doc/extend.texi (nonnull attribute): Improve documentation.
7398 2021-08-02 Andrew Pinski <apinski@marvell.com>
7400 PR rtl-optimization/101683
7401 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7403 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7405 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7406 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7407 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7408 (tree_ssa_phiop_worker): Update call to function above.
7410 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7413 * config/i386/i386.md (bsr_rex64_1_zext): New.
7414 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7415 with gen_bsr_rex64_1_zext.
7417 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7420 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7421 define_insn patterns.
7422 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7423 Add combine splitters for constant - clz.
7424 (clz<mode>2): Use a temporary pseudo for bsr result.
7426 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7428 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7429 _mm_floor_sd, _mm_floor_ss): New.
7431 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7433 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7434 _mm_ceil_sd, _mm_ceil_ss): New.
7436 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7438 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7439 _mm_blend_ps, _mm_blendv_ps): New.
7441 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7442 Uroš Bizjak <ubizjak@gmail.com>
7444 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7445 to generate a conditional move using the carry flag after sub $1.
7446 (peephole2): Eliminate a register-to-register move by inverting
7447 the condition of a conditional move.
7449 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7451 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7452 ("*call_value_real"): Don't generate rtx mentioning the generic
7453 operands 1 and 2 to "call", and similarly for "call_value".
7454 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7455 (mmix_print_operand): Use '!' instead of 'p'.
7457 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7459 * doc/md.texi (call): Correct information about operand 2.
7460 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7462 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7464 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7466 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7468 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7470 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7472 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7474 (*::get_bb_range): Ditto.
7475 (*::bb_range_p): Ditto.
7476 * gimple-range-cache.h: Change prototypes.
7478 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7481 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7482 from scalar_int_mode to fixed_size_mode.
7483 (builtin_strncpy_read_str): Likewise.
7484 (gen_memset_value_from_prev): New function.
7485 (builtin_memset_read_str): Change the mode argument from
7486 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7487 and support CONST_VECTOR.
7488 (builtin_memset_gen_str): Likewise.
7489 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7491 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7492 with fixed_size_mode.
7493 (builtin_memset_read_str): Likewise.
7494 * expr.c (widest_int_mode_for_size): Renamed to ...
7495 (widest_fixed_size_mode_for_size): Add a bool argument to
7496 indicate if QI vector mode can be used.
7497 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7498 instead of widest_int_mode_for_size.
7499 (pieces_addr::adjust): Change the mode argument from
7500 scalar_int_mode to fixed_size_mode.
7501 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7502 m_qi_vector_mode, to indicate that QI vector mode can be used.
7503 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7504 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7505 instead of widest_int_mode_for_size.
7506 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7507 scalar_int_mode to fixed_size_mode. Call
7508 widest_fixed_size_mode_for_size instead of
7509 widest_int_mode_for_size.
7510 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7511 function to return the smallest integer or QI vector mode.
7512 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7513 instead of widest_int_mode_for_size. Call
7514 smallest_fixed_size_mode_for_size instead of
7515 smallest_int_mode_for_size.
7516 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7517 indicate that QI vector mode can be used and pass it to
7518 op_by_pieces_d::op_by_pieces_d.
7519 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7520 instead of widest_int_mode_for_size. Pass memsetp to
7521 widest_fixed_size_mode_for_size to support QI vector mode.
7522 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7523 (store_by_pieces): Pass memsetp to
7524 store_by_pieces_d::store_by_pieces_d.
7525 (clear_by_pieces_1): Removed.
7526 (clear_by_pieces): Replace clear_by_pieces_1 with
7527 builtin_memset_read_str and pass true to store_by_pieces_d to
7528 support vector mode broadcast.
7529 (string_cst_read_str): Change the mode argument from
7530 scalar_int_mode to fixed_size_mode.
7531 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7533 (by_pieces_prev): Likewise.
7534 * rtl.h (lowpart_subreg_regno): New.
7535 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7536 simplify_subreg_regno.
7537 * target.def (gen_memset_scratch_rtx): New hook.
7538 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7539 * doc/tm.texi: Regenerated.
7541 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7544 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7545 TARGET_EXPR instead of MODIFY_EXPR.
7547 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7550 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7551 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7552 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7554 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7556 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7559 * config/i386/i386-options.c (ix86_option_override_internal):
7560 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7562 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7565 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7566 explicitly dereferencing __a.
7567 (vld1_s64): Likewise.
7568 (vld1_u64): Likewise.
7569 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7570 and change to VAR13.
7572 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7574 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7575 use of m_range_analyzer.
7576 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7577 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7578 (loop_versioning::prune_loop_conditions): Replace vr_values use
7579 with range_query interface.
7580 (pass_loop_versioning::execute): Use ranger.
7582 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7585 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7586 enum values, and emit a warning if they mismatch.
7588 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7590 * cfgloop.h (as_const): New function.
7591 (class loop_iterator): Rename to ...
7592 (class loops_list): ... this.
7593 (loop_iterator::next): Rename to ...
7594 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7595 (loop_iterator::loop_iterator): Rename to ...
7596 (loops_list::loops_list): ... this and adjust.
7597 (loops_list::Iter): New class.
7598 (loops_list::iterator): New type.
7599 (loops_list::const_iterator): New type.
7600 (loops_list::begin): New function.
7601 (loops_list::end): Likewise.
7602 (loops_list::begin const): Likewise.
7603 (loops_list::end const): Likewise.
7604 (FOR_EACH_LOOP): Remove.
7605 (FOR_EACH_LOOP_FN): Remove.
7606 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
7607 for loop with loops_list instance.
7608 (sort_sibling_loops): Likewise.
7609 (disambiguate_loops_with_multiple_latches): Likewise.
7610 (verify_loop_structure): Likewise.
7611 * cfgloopmanip.c (create_preheaders): Likewise.
7612 (force_single_succ_latches): Likewise.
7613 * config/aarch64/falkor-tag-collision-avoidance.c
7614 (execute_tag_collision_avoidance): Likewise.
7615 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
7616 * config/s390/s390.c (s390_adjust_loops): Likewise.
7617 * doc/loop.texi: Likewise.
7618 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
7619 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
7620 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
7621 (loop_versioning::make_versioning_decisions): Likewise.
7622 * gimple-ssa-split-paths.c (split_paths): Likewise.
7623 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
7624 * graphite.c (canonicalize_loop_form): Likewise.
7625 (graphite_transform_loops): Likewise.
7626 * ipa-fnsummary.c (analyze_function_body): Likewise.
7627 * ipa-pure-const.c (analyze_function): Likewise.
7628 * loop-doloop.c (doloop_optimize_loops): Likewise.
7629 * loop-init.c (loop_optimizer_finalize): Likewise.
7630 (fix_loop_structure): Likewise.
7631 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
7632 (move_loop_invariants): Likewise.
7633 * loop-unroll.c (decide_unrolling): Likewise.
7634 (unroll_loops): Likewise.
7635 * modulo-sched.c (sms_schedule): Likewise.
7636 * predict.c (predict_loops): Likewise.
7637 (pass_profile::execute): Likewise.
7638 * profile.c (branch_prob): Likewise.
7639 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
7640 (sel_find_rgns): Likewise.
7641 * tree-cfg.c (replace_loop_annotate): Likewise.
7642 (replace_uses_by): Likewise.
7643 (move_sese_region_to_fn): Likewise.
7644 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
7645 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
7646 * tree-parloops.c (parallelize_loops): Likewise.
7647 * tree-predcom.c (tree_predictive_commoning): Likewise.
7648 * tree-scalar-evolution.c (scev_initialize): Likewise.
7649 (scev_reset): Likewise.
7650 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
7651 * tree-ssa-live.c (remove_unused_locals): Likewise.
7652 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
7653 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
7654 (tree_ssa_lim_initialize): Likewise.
7655 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
7656 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
7657 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
7658 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
7659 (free_numbers_of_iterations_estimates): Likewise.
7660 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
7661 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
7662 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
7663 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
7664 (pass_scev_cprop::execute): Likewise.
7665 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
7666 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
7667 * tree-ssa-threadupdate.c
7668 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
7669 * tree-vectorizer.c (vectorize_loops): Likewise.
7670 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
7672 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
7674 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
7675 generating a VOIDmode register for e.g the
7676 function_arg_info::end_marker.
7678 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
7680 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
7681 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
7682 (cc mode_attr): Similarly.
7683 (ccz subst_attr): Similarly.
7684 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
7685 * config/h8300/testcompare.md: Remove various cc0 based patterns
7686 that had been commented out. Add pattern to set CCZ from a bit
7689 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
7690 Julian Brown <julian@codesourcery.com>
7691 Kwok Cheung Yeung <kcy@codesourcery.com>
7693 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
7694 'update_stmt' after modification.
7695 (pass_oacc_loop_designation): New function, extracted out of...
7696 (pass_oacc_device_lower): ... this.
7697 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
7698 (make_pass_oacc_loop_designation): New
7699 * passes.def: Add it.
7700 * tree-parloops.c (create_parallel_loop): Adjust.
7701 * tree-pass.h (make_pass_oacc_loop_designation): New.
7703 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
7705 * flag-types.h (enum threader_mode): New.
7706 * params.opt: Add entry for --param=threader-mode.
7707 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
7708 (class back_threader): New.
7709 (back_threader::back_threader): New.
7710 (back_threader::~back_threader): New.
7711 (back_threader::maybe_register_path): New.
7712 (back_threader::find_taken_edge): New.
7713 (back_threader::find_taken_edge_switch): New.
7714 (back_threader::find_taken_edge_cond): New.
7715 (back_threader::resolve_def): New.
7716 (back_threader::resolve_phi): New.
7717 (back_threader::find_paths_to_names): New.
7718 (back_threader::find_paths): New.
7721 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
7722 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
7723 (pass_thread_jumps::execute): Abstract out code...
7724 (try_thread_blocks): ...here.
7725 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
7726 Abstract out threading candidate code to...
7727 (single_succ_to_potentially_threadable_block): ...here.
7728 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
7730 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
7731 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7732 Return bool from register_jump_thread.
7734 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
7736 * target.def: in0 and in1 do not need to be registers.
7737 * doc/tm.texi: Regenerate.
7739 2021-07-29 liuhongt <hongtao.liu@intel.com>
7742 * config/i386/i386.c (ix86_widen_mult_cost): New function.
7743 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
7746 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
7749 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
7750 (rs6000_preferred_doloop_mode): New hook.
7751 * doc/tm.texi: Regenerate.
7752 * doc/tm.texi.in: Add hook preferred_doloop_mode.
7753 * target.def (preferred_doloop_mode): New hook.
7754 * targhooks.c (default_preferred_doloop_mode): New hook.
7755 * targhooks.h (default_preferred_doloop_mode): New hook.
7756 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
7757 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
7758 and compute_doloop_base_on_mode.
7760 2021-07-28 Martin Sebor <msebor@redhat.com>
7762 PR middle-end/101494
7763 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
7764 and size computation.
7766 2021-07-28 Martin Sebor <msebor@redhat.com>
7768 PR middle-end/101601
7769 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
7771 Handle pointers to functions.
7773 2021-07-28 Martin Sebor <msebor@redhat.com>
7775 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
7776 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
7777 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
7778 (access_ref::access_ref): Same.
7779 (access_ref::phi): Same.
7780 (access_ref::get_ref): Same.
7781 (access_ref::size_remaining): Same.
7782 (access_ref::offset_in_range): Same.
7783 (access_ref::add_offset): Same.
7784 (access_ref::inform_access): Same.
7785 (ssa_name_limit_t::visit_phi): Same.
7786 (ssa_name_limit_t::leave_phi): Same.
7787 (ssa_name_limit_t::next): Same.
7788 (ssa_name_limit_t::next_phi): Same.
7789 (ssa_name_limit_t::~ssa_name_limit_t): Same.
7790 (pointer_query::pointer_query): Same.
7791 (pointer_query::get_ref): Same.
7792 (pointer_query::put_ref): Same.
7793 (pointer_query::flush_cache): Same.
7794 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
7795 (check_nul_terminated_array): Same.
7796 (unterminated_array): Same.
7797 (maybe_warn_for_bound): Same.
7798 (check_read_access): Same.
7799 (warn_for_access): Same.
7800 (get_size_range): Same.
7801 (check_access): Same.
7802 (gimple_call_alloc_size): Move to tree.c.
7803 (gimple_parm_array_size): Move to pointer-query.cc.
7804 (get_offset_range): Same.
7805 (gimple_call_return_array): Same.
7806 (handle_min_max_size): Same.
7807 (handle_array_ref): Same.
7808 (handle_mem_ref): Same.
7809 (compute_objsize): Same.
7810 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
7811 (call_dealloc_argno): Same.
7812 (fndecl_dealloc_argno): Same.
7813 (new_delete_mismatch_p): Same.
7814 (matching_alloc_calls_p): Same.
7815 (warn_dealloc_offset): Same.
7816 (maybe_emit_free_warning): Same.
7817 * builtins.h (check_nul_terminated_array): Move to
7818 gimple-ssa-warn-access.h.
7819 (check_nul_terminated_array): Same.
7820 (warn_string_no_nul): Same.
7821 (unterminated_array): Same.
7822 (class ssa_name_limit_t): Same.
7823 (class pointer_query): Same.
7824 (struct access_ref): Same.
7825 (class range_query): Same.
7826 (struct access_data): Same.
7827 (gimple_call_alloc_size): Same.
7828 (gimple_parm_array_size): Same.
7829 (compute_objsize): Same.
7830 (class access_data): Same.
7831 (maybe_emit_free_warning): Same.
7832 * calls.c (initialize_argument_information): Remove call to
7833 maybe_emit_free_warning.
7834 * gimple-array-bounds.cc: Include new header..
7835 * gimple-fold.c: Same.
7836 * gimple-ssa-sprintf.c: Same.
7837 * gimple-ssa-warn-restrict.c: Same.
7838 * passes.def: Add pass_warn_access.
7839 * tree-pass.h (make_pass_warn_access): Declare.
7840 * tree-ssa-strlen.c: Include new headers.
7841 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
7842 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
7843 * gimple-ssa-warn-access.cc: New file.
7844 * gimple-ssa-warn-access.h: New file.
7845 * pointer-query.cc: New file.
7846 * pointer-query.h: New file.
7848 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7850 PR middle-end/101624
7851 * ubsan.c (maybe_instrument_pointer_overflow,
7852 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
7853 PARM_DECLs or RESULT_DECLs.
7854 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
7856 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7858 PR middle-end/101642
7859 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
7860 to type of bswap16 for comparison.
7861 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
7863 2021-07-28 Richard Biener <rguenther@suse.de>
7865 PR tree-optimization/101615
7866 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
7867 at CTOR SLP graph entries.
7869 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7871 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
7872 Add "r,w" alternative.
7874 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
7877 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
7878 AVX_U128_DIRTY when all bits are zero.
7880 2021-07-28 Richard Biener <rguenther@suse.de>
7882 PR tree-optimization/101615
7883 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
7884 external nodes cannot be permuted so make them perm_out 0.
7886 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
7889 * config.in: Regenerate.
7890 * config/gcn/gcn-hsa.h (A_FIJI): New define.
7891 (A_900): New define.
7892 (A_906): New define.
7893 (A_908): New define.
7894 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
7895 * config/gcn/gcn.c (output_file_start): Adjust attributes according
7896 to the assembler capabilities.
7897 * config/gcn/mkoffload.c (main): Likewise.
7898 * configure: Regenerate.
7899 * configure.ac: Add tests for LLVM assembler attribute features.
7901 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
7903 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
7904 cond_false and cond_true on branches.
7906 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
7908 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
7911 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
7913 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
7915 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7918 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
7919 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
7920 and has special !TARGET_AVX512VL expansion.
7921 (vashrv2di3<mask_name>): Rename to ...
7922 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
7923 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
7925 2021-07-28 Martin Uecker <muecker@gwdg.de>
7927 * calls.c (maybe_warn_rdwr_sizes): Correct argument
7928 numbers in warning that were switched.
7930 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
7932 PR tree-optimization/101596
7933 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
7934 by using new_type's precision instead.
7936 2021-07-28 liuhongt <hongtao.liu@intel.com>
7939 * config/i386/i386.h (processor_costs): Add new member
7941 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
7942 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
7943 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
7944 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
7945 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
7946 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
7947 generic_cost, core_cost): Initialize integer_to_sse same value
7949 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
7950 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7951 Use integer_to_sse instead of sse_op to calculate the cost of
7954 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7956 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
7958 (write_init_file): Call write_ovld_static_init.
7960 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7962 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
7964 (write_init_file): Call write_bif_static_init.
7966 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7968 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
7969 (TYPE_MAP_SIZE): New macro.
7970 (type_map): New initialized variable.
7971 (typemap_cmp): New function.
7972 (write_type_node): Likewise.
7973 (write_fntype_init): Implement.
7975 2021-07-27 Martin Sebor <msebor@redhat.com>
7977 PR tree-optimization/101584
7978 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
7979 (check_defs): Call it.
7981 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
7983 * tree-ssa-dom.c (dom_jump_threader_simplifier):
7984 Put avail_exprs_stack in the class, instead of passing it to
7985 jump_threader_simplifier.
7986 (dom_jump_threader_simplifier::simplify): Add state argument.
7987 (dom_opt_dom_walker): Add state.
7988 (pass_dominator::execute): Pass state to threader.
7989 (dom_opt_dom_walker::before_dom_children): Use state.
7990 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
7992 (jump_threader::record_temporary_equivalences_from_phis):
7993 Register equivalences through the state variable.
7994 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
7995 Record ranges in a statement through the state variable.
7996 (jump_threader::simplify_control_stmt_condition): Pass state to
7998 (jump_threader::simplify_control_stmt_condition_1): Same.
7999 (jump_threader::thread_around_empty_blocks): Remove obsolete
8001 (jump_threader::thread_through_normal_block): Record equivalences
8002 on edge through the state variable.
8003 (jump_threader::thread_across_edge): Abstract state pushing.
8004 (jt_state::jt_state): New.
8005 (jt_state::push): New.
8006 (jt_state::pop): New.
8007 (jt_state::register_equiv): New.
8008 (jt_state::record_ranges_from_stmt): New.
8009 (jt_state::register_equivs_on_edge): New.
8010 (jump_threader_simplifier::jump_threader_simplifier): Move from
8012 (jump_threader_simplifier::simplify): Add state argument.
8013 * tree-ssa-threadedge.h (class jt_state): New.
8014 (class jump_threader): Add state to constructor.
8015 (class jump_threader_simplifier): Add state to simplify. Remove
8016 avail_exprs_stack from class.
8017 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
8019 (vrp_jump_threader::vrp_jump_threader): Add state.
8020 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
8022 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8024 * Makefile.in (OBJS): Add gimple-range-path.o.
8025 * gimple-range-path.cc: New file.
8026 * gimple-range-path.h: New file.
8028 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
8030 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
8031 inside vec_duplicate for all patterns.
8032 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
8033 Push sign/zero-extension inside vec_duplicate.
8035 2021-07-27 Richard Biener <rguenther@suse.de>
8037 PR tree-optimization/101573
8038 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
8039 looking at uninitialized PHI arg defs in some constrained cases.
8040 (warn_uninitialized_vars): Call it.
8041 (execute_early_warn_uninitialized): Calculate dominators.
8043 2021-07-27 Richard Biener <rguenther@suse.de>
8045 PR tree-optimization/39821
8046 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
8047 vector_stmt for widening arithmetic.
8048 (vectorizable_conversion): Adjust.
8050 2021-07-27 Martin Jambor <mjambor@suse.cz>
8052 * cgraph.h (ipa_replace_map): New field force_load_ref.
8053 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
8054 aded new flag load_dereferenced, adjusted comments.
8055 (ipa_get_param_dereferenced): New function.
8056 (ipa_set_param_dereferenced): Likewise.
8057 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
8058 * ipa-cp.c: Include gimple.h.
8059 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
8060 (get_replacement_map): New parameter force_load_ref, set the
8061 appropriate flag in ipa_replace_map if set.
8062 (struct symbol_and_index_together): New type.
8063 (adjust_refs_in_act_callers): New function.
8064 (adjust_references_in_caller): Likewise.
8065 (create_specialized_node): When appropriate, call
8066 adjust_references_in_caller and force only load references.
8067 * ipa-prop.c (load_from_dereferenced_name): New function.
8068 (ipa_analyze_controlled_uses): Also detect loads from a
8069 dereference, harden testing of call statements.
8070 (ipa_write_node_info): Stream the dereferenced flag.
8071 (ipa_read_node_info): Likewise.
8072 (ipa_set_jf_constant): Also create refdesc when jump function
8073 references a variable.
8074 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
8075 also on references of variables and return a symtab_node. Adjust
8077 (propagate_controlled_uses): Also remove references to VAR_DECLs.
8079 2021-07-27 Jakub Jelinek <jakub@redhat.com>
8081 PR middle-end/101586
8082 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
8083 positions above or equal to sz except for diagnostics of flexible
8086 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8088 PR tree-optimization/78888
8089 * gimple-range-fold.cc (get_letter_range): New.
8090 (fold_using_range::range_of_builtin_call): Call get_letter_range.
8092 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8094 PR tree-optimization/78888
8095 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
8096 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
8098 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
8099 Marc Glisse <marc.glisse@inria.fr>
8101 * match.pd (rotate): Simplify equality/inequality of rotations.
8102 (bswap): Simplify equality/inequality tests of byte swapping.
8104 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8106 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
8109 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8111 * range-op.cc (operator_lshift::fold_range): Pass rel to
8112 base class fold_range.
8113 (operator_rshift::fold_range): Same.
8115 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8118 * toplev.h (min_align_loops_log): Remove declaration.
8119 (min_align_jumps_log, min_align_labels_log): Likewise.
8120 (min_align_functions_log): Likewise.
8122 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8124 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8126 (execute_vrp): Abstract out simplification of conditionals...
8127 (simplify_casted_conds): ...here.
8129 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8131 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8132 Add gimple argument.
8133 (array_bounds_checker::check_array_ref): Same.
8134 (array_bounds_checker::check_addr_expr): Same.
8135 (array_bounds_checker::check_array_bounds): Pass statement to
8136 check_array_bounds and check_addr_expr.
8137 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8138 (check_addr_expr): Same.
8139 (get_value_range): Same.
8141 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8143 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8144 (sdot_prod, udot_prod): ... This.
8145 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8147 (<sur>dot_prod<vsi2qi>): ... this.
8148 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8149 Change operands order.
8150 (<sur>sadv16qi): Use new operands order.
8151 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8152 vdotq_s32): Use new RTL ordering.
8154 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8156 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8157 aarch64_types_ternop_suss_qualifiers): New.
8158 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8159 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8160 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8162 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8164 PR rtl-optimization/101562
8165 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8166 if the operand isn't paradoxical.
8168 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8170 * gimple-array-bounds.h (class array_bounds_checker): Change
8171 ranges type to range_query.
8173 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8175 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8176 __builtin_memcpy instead of constructing
8177 __builtin_aarch64_simd_oi one vector at a time.
8178 (vst1_u64_x2): Likewise.
8179 (vst1_f64_x2): Likewise.
8180 (vst1_s8_x2): Likewise.
8181 (vst1_p8_x2): Likewise.
8182 (vst1_s16_x2): Likewise.
8183 (vst1_p16_x2): Likewise.
8184 (vst1_s32_x2): Likewise.
8185 (vst1_u8_x2): Likewise.
8186 (vst1_u16_x2): Likewise.
8187 (vst1_u32_x2): Likewise.
8188 (vst1_f16_x2): Likewise.
8189 (vst1_f32_x2): Likewise.
8190 (vst1_p64_x2): Likewise.
8191 (vst1q_s8_x2): Likewise.
8192 (vst1q_p8_x2): Likewise.
8193 (vst1q_s16_x2): Likewise.
8194 (vst1q_p16_x2): Likewise.
8195 (vst1q_s32_x2): Likewise.
8196 (vst1q_s64_x2): Likewise.
8197 (vst1q_u8_x2): Likewise.
8198 (vst1q_u16_x2): Likewise.
8199 (vst1q_u32_x2): Likewise.
8200 (vst1q_u64_x2): Likewise.
8201 (vst1q_f16_x2): Likewise.
8202 (vst1q_f32_x2): Likewise.
8203 (vst1q_f64_x2): Likewise.
8204 (vst1q_p64_x2): Likewise.
8206 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8208 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8209 __builtin_memcpy instead of constructing
8210 __builtin_aarch64_simd_ci one vector at a time.
8211 (vst1_u64_x3): Likewise.
8212 (vst1_f64_x3): Likewise.
8213 (vst1_s8_x3): Likewise.
8214 (vst1_p8_x3): Likewise.
8215 (vst1_s16_x3): Likewise.
8216 (vst1_p16_x3): Likewise.
8217 (vst1_s32_x3): Likewise.
8218 (vst1_u8_x3): Likewise.
8219 (vst1_u16_x3): Likewise.
8220 (vst1_u32_x3): Likewise.
8221 (vst1_f16_x3): Likewise.
8222 (vst1_f32_x3): Likewise.
8223 (vst1_p64_x3): Likewise.
8224 (vst1q_s8_x3): Likewise.
8225 (vst1q_p8_x3): Likewise.
8226 (vst1q_s16_x3): Likewise.
8227 (vst1q_p16_x3): Likewise.
8228 (vst1q_s32_x3): Likewise.
8229 (vst1q_s64_x3): Likewise.
8230 (vst1q_u8_x3): Likewise.
8231 (vst1q_u16_x3): Likewise.
8232 (vst1q_u32_x3): Likewise.
8233 (vst1q_u64_x3): Likewise.
8234 (vst1q_f16_x3): Likewise.
8235 (vst1q_f32_x3): Likewise.
8236 (vst1q_f64_x3): Likewise.
8237 (vst1q_p64_x3): Likewise.
8239 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
8242 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
8243 hard register when LRA is in progress.
8245 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8247 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
8248 __builtin_memcpy instead of using a union.
8249 (vst1q_s8_x4): Likewise.
8250 (vst1_s16_x4): Likewise.
8251 (vst1q_s16_x4): Likewise.
8252 (vst1_s32_x4): Likewise.
8253 (vst1q_s32_x4): Likewise.
8254 (vst1_u8_x4): Likewise.
8255 (vst1q_u8_x4): Likewise.
8256 (vst1_u16_x4): Likewise.
8257 (vst1q_u16_x4): Likewise.
8258 (vst1_u32_x4): Likewise.
8259 (vst1q_u32_x4): Likewise.
8260 (vst1_f16_x4): Likewise.
8261 (vst1q_f16_x4): Likewise.
8262 (vst1_f32_x4): Likewise.
8263 (vst1q_f32_x4): Likewise.
8264 (vst1_p8_x4): Likewise.
8265 (vst1q_p8_x4): Likewise.
8266 (vst1_p16_x4): Likewise.
8267 (vst1q_p16_x4): Likewise.
8268 (vst1_s64_x4): Likewise.
8269 (vst1_u64_x4): Likewise.
8270 (vst1_p64_x4): Likewise.
8271 (vst1q_s64_x4): Likewise.
8272 (vst1q_u64_x4): Likewise.
8273 (vst1q_p64_x4): Likewise.
8274 (vst1_f64_x4): Likewise.
8275 (vst1q_f64_x4): Likewise.
8277 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8279 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8280 instead of constructing __builtin_aarch64_simd_oi one vector
8282 (vst2_u64): Likewise.
8283 (vst2_f64): Likewise.
8284 (vst2_s8): Likewise.
8285 (vst2_p8): Likewise.
8286 (vst2_s16): Likewise.
8287 (vst2_p16): Likewise.
8288 (vst2_s32): Likewise.
8289 (vst2_u8): Likewise.
8290 (vst2_u16): Likewise.
8291 (vst2_u32): Likewise.
8292 (vst2_f16): Likewise.
8293 (vst2_f32): Likewise.
8294 (vst2_p64): Likewise.
8295 (vst2q_s8): Likewise.
8296 (vst2q_p8): Likewise.
8297 (vst2q_s16): Likewise.
8298 (vst2q_p16): Likewise.
8299 (vst2q_s32): Likewise.
8300 (vst2q_s64): Likewise.
8301 (vst2q_u8): Likewise.
8302 (vst2q_u16): Likewise.
8303 (vst2q_u32): Likewise.
8304 (vst2q_u64): Likewise.
8305 (vst2q_f16): Likewise.
8306 (vst2q_f32): Likewise.
8307 (vst2q_f64): Likewise.
8308 (vst2q_p64): Likewise.
8310 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8312 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8313 instead of constructing __builtin_aarch64_simd_ci one vector
8315 (vst3_u64): Likewise.
8316 (vst3_f64): Likewise.
8317 (vst3_s8): Likewise.
8318 (vst3_p8): Likewise.
8319 (vst3_s16): Likewise.
8320 (vst3_p16): Likewise.
8321 (vst3_s32): Likewise.
8322 (vst3_u8): Likewise.
8323 (vst3_u16): Likewise.
8324 (vst3_u32): Likewise.
8325 (vst3_f16): Likewise.
8326 (vst3_f32): Likewise.
8327 (vst3_p64): Likewise.
8328 (vst3q_s8): Likewise.
8329 (vst3q_p8): Likewise.
8330 (vst3q_s16): Likewise.
8331 (vst3q_p16): Likewise.
8332 (vst3q_s32): Likewise.
8333 (vst3q_s64): Likewise.
8334 (vst3q_u8): Likewise.
8335 (vst3q_u16): Likewise.
8336 (vst3q_u32): Likewise.
8337 (vst3q_u64): Likewise.
8338 (vst3q_f16): Likewise.
8339 (vst3q_f32): Likewise.
8340 (vst3q_f64): Likewise.
8341 (vst3q_p64): Likewise.
8343 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8345 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8346 instead of constructing __builtin_aarch64_simd_xi one vector
8348 (vst4_u64): Likewise.
8349 (vst4_f64): Likewise.
8350 (vst4_s8): Likewise.
8351 (vst4_p8): Likewise.
8352 (vst4_s16): Likewise.
8353 (vst4_p16): Likewise.
8354 (vst4_s32): Likewise.
8355 (vst4_u8): Likewise.
8356 (vst4_u16): Likewise.
8357 (vst4_u32): Likewise.
8358 (vst4_f16): Likewise.
8359 (vst4_f32): Likewise.
8360 (vst4_p64): Likewise.
8361 (vst4q_s8): Likewise.
8362 (vst4q_p8): Likewise.
8363 (vst4q_s16): Likewise.
8364 (vst4q_p16): Likewise.
8365 (vst4q_s32): Likewise.
8366 (vst4q_s64): Likewise.
8367 (vst4q_u8): Likewise.
8368 (vst4q_u16): Likewise.
8369 (vst4q_u32): Likewise.
8370 (vst4q_u64): Likewise.
8371 (vst4q_f16): Likewise.
8372 (vst4q_f32): Likewise.
8373 (vst4q_f64): Likewise.
8374 (vst4q_p64): Likewise.
8376 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8378 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8379 instead of constructing __builtin_aarch64_simd_oi one vector
8381 (vtbx4_u8): Likewise.
8382 (vtbx4_p8): Likewise.
8384 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8386 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8387 instead of constructing __builtin_aarch64_simd_oi one vector
8389 (vtbl3_u8): Likewise.
8390 (vtbl3_p8): Likewise.
8391 (vtbl4_s8): Likewise.
8392 (vtbl4_u8): Likewise.
8393 (vtbl4_p8): Likewise.
8395 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8397 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8398 instead of constructing __builtin_aarch64_simd_oi one vector
8400 (vqtbx2_u8): Likewise.
8401 (vqtbx2_p8): Likewise.
8402 (vqtbx2q_s8): Likewise.
8403 (vqtbx2q_u8): Likewise.
8404 (vqtbx2q_p8): Likewise.
8405 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8406 __builtin_aarch64_simd_ci one vector at a time.
8407 (vqtbx3_u8): Likewise.
8408 (vqtbx3_p8): Likewise.
8409 (vqtbx3q_s8): Likewise.
8410 (vqtbx3q_u8): Likewise.
8411 (vqtbx3q_p8): Likewise.
8412 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8413 __builtin_aarch64_simd_xi one vector at a time.
8414 (vqtbx4_u8): Likewise.
8415 (vqtbx4_p8): Likewise.
8416 (vqtbx4q_s8): Likewise.
8417 (vqtbx4q_u8): Likewise.
8418 (vqtbx4q_p8): Likewise.
8420 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8422 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8423 instead of constructing __builtin_aarch64_simd_oi one vector
8425 (vqtbl2_u8): Likewise.
8426 (vqtbl2_p8): Likewise.
8427 (vqtbl2q_s8): Likewise.
8428 (vqtbl2q_u8): Likewise.
8429 (vqtbl2q_p8): Likewise.
8430 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8431 __builtin_aarch64_simd_ci one vector at a time.
8432 (vqtbl3_u8): Likewise.
8433 (vqtbl3_p8): Likewise.
8434 (vqtbl3q_s8): Likewise.
8435 (vqtbl3q_u8): Likewise.
8436 (vqtbl3q_p8): Likewise.
8437 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8438 __builtin_aarch64_simd_xi one vector at a time.
8439 (vqtbl4_u8): Likewise.
8440 (vqtbl4_p8): Likewise.
8441 (vqtbl4q_s8): Likewise.
8442 (vqtbl4q_u8): Likewise.
8443 (vqtbl4q_p8): Likewise.
8445 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8448 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8450 2021-07-22 Andrew Pinski <apinski@marvell.com>
8452 PR tree-optimization/10153
8453 * tree-tailcall.c (create_tailcall_accumulator):
8454 Don't call fold_convert as the type should be correct already.
8455 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8456 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8458 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8460 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8461 varying_p check for null/non-null check.
8463 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8465 PR tree-optimization/101511
8466 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8467 is in ssa2's equiv set, and don't trap if so.
8469 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8471 PR tree-optimization/101497
8472 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8475 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8477 PR tree-optimization/101496
8478 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8479 first, then vrp_visit_cond_Stmt.
8481 2021-07-22 liuhongt <hongtao.liu@intel.com>
8483 * config/i386/i386-expand.c
8484 (ix86_broadcast_from_integer_constant): Rename to ..
8485 (ix86_broadcast_from_constant): .. this, and extend it to
8487 (ix86_expand_vector_move): Extend to float mode.
8488 * config/i386/i386-features.c
8489 (replace_constant_pool_with_broadcast): Remove.
8490 (remove_partial_avx_dependency_gate): Ditto.
8491 (constant_pool_broadcast): Ditto.
8492 (class pass_constant_pool_broadcast): Ditto.
8493 (make_pass_constant_pool_broadcast): Ditto.
8494 (remove_partial_avx_dependency): Adjust gate.
8495 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8496 * config/i386/i386-protos.h
8497 (make_pass_constant_pool_broadcast): Remove.
8499 2021-07-22 liuhongt <hongtao.liu@intel.com>
8501 * config/i386/constraints.md (Wb): New constraint.
8503 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8505 (*ashlqi3_1): Ditto.
8506 (*<insn><mode>3_1): Split to ..
8507 (*ashr<mode>3_1): this, ...
8508 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8510 (*<insn><mode>3_1): Split to ..
8511 (*ashr<mode>3_1): this, ...
8512 (*lshrqi3_1): and this, also extend this pattern to avx512
8514 (*lshrhi3_1): And this, also extend this pattern to avx512
8516 * config/i386/sse.md (k<code><mode>): New define_split after
8517 it to convert generic shift pattern to mask shift ones.
8519 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8520 Joseph Myers <joseph@codesourcery.com>
8521 Cesar Philippidis <cesar@codesourcery.com>
8523 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8524 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8526 * tree-pretty-print.c (dump_omp_clause): Likewise.
8527 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8528 * gimplify.c (gimplify_scan_omp_clauses)
8529 (gimplify_adjust_omp_clauses): Likewise.
8530 * tree-nested.c (convert_nonlocal_omp_clauses)
8531 (convert_local_omp_clauses): Likewise.
8532 * omp-low.c (scan_sharing_clauses): Likewise.
8533 * omp-offload.c (execute_oacc_device_lower): Update.
8535 2021-07-21 Martin Sebor <msebor@redhat.com>
8537 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8539 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8541 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8544 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8546 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8548 (write_fntype_init): New stub function.
8549 (write_init_bif_table): Likewise.
8550 (write_init_ovld_table): New function.
8551 (write_init_file): Implement.
8553 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8555 * config/rs6000/rs6000-gen-builtins.c
8556 (write_autogenerated_header): New function.
8557 (write_decls): Likewise.
8558 (write_extern_fntype): New callback function.
8559 (write_header_file): Implement.
8561 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8563 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8566 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8568 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8570 (complete_base_type): Likewise.
8571 (construct_fntype_id): Likewise.
8572 (parse_bif_entry): Call contruct_fntype_id.
8573 (parse_ovld_entry): Likewise.
8575 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8577 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8578 (MAXOVLDSTANZAS): New macro.
8579 (ovld_stanzas): New variable.
8580 (curr_ovld_stanza): Likewise.
8581 (MAXOVLDS): New macro.
8582 (ovlddata): New struct.
8583 (ovlds): New variable.
8584 (curr_ovld): Likewise.
8585 (max_ovld_args): Likewise.
8586 (parse_ovld_entry): New function.
8587 (parse_ovld_stanza): Likewise.
8588 (parse_ovld): Implement.
8590 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8592 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8595 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8597 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8598 (parse_prototype): Implement.
8600 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8602 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
8603 (curr_bif_stanza): New variable.
8604 (stanza_entry): New struct.
8605 (stanza_map): New initialized variable.
8606 (enable_string): Likewise.
8607 (fnkinds): New enum.
8608 (typelist): New struct.
8609 (attrinfo): Likewise.
8610 (MAXRESTROPNDS): New macro.
8611 (prototype): New struct.
8612 (MAXBIFS): New macro.
8613 (bifdata): New struct.
8614 (bifs): New variable.
8615 (curr_bif): Likewise.
8616 (bif_order): Likewise.
8617 (bif_index): Likewise.
8618 (fatal): New function.
8619 (stanza_name_to_stanza): Likewise.
8620 (parse_bif_attrs): New stub function.
8621 (parse_prototype): Likewise.
8622 (parse_bif_entry): New function.
8623 (parse_bif_stanza): Likewise.
8624 (parse_bif): Implement.
8625 (set_bif_order): New function.
8626 (create_bif_order): Implement.
8628 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8630 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
8631 (num_bifs): New variable.
8632 (num_ovld_stanzas): Likewise.
8633 (num_ovlds): Likewise.
8634 (parse_codes): New enum.
8635 (bif_rbt): New variable.
8636 (ovld_rbt): Likewise.
8637 (fntype_rbt): Likewise.
8638 (bifo_rbt): Likewise.
8639 (parse_bif): New stub function.
8640 (create_bif_order): Likewise.
8641 (parse_ovld): Likewise.
8642 (write_header_file): Likewise.
8643 (write_init_file): Likewise.
8644 (write_defines_file): Likewise.
8645 (delete_output_files): New function.
8648 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
8651 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
8652 from CRC32 _builtin functions.
8654 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
8656 * coverage.c (build_gcov_info_var_registration): Mark the object placed
8657 in the linker set as referenced so that it does not get optimized away.
8659 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
8662 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8664 * config.gcc (riscv*-*-*): Detect which python is available.
8666 2021-07-21 Jakub Jelinek <jakub@redhat.com>
8668 PR middle-end/101535
8669 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
8670 contexts in which decl isn't privatized and for ORT_TARGET return
8671 false if decl is mapped.
8673 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8675 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
8676 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
8677 has been successfully unrolled.
8679 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8681 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
8682 exit condition if no tail loop is needed, and if the original exit
8683 condition should therefore be kept as-is.
8684 (tree_transform_and_unroll_loop): Handle that case here too.
8686 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
8688 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
8690 (free_data_refs): Likewise.
8691 * tree-data-ref.h (free_dependence_relations): Likewise.
8692 (free_data_refs): Likewise.
8693 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
8695 (struct component): Likewise.
8696 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
8697 (pcom_worker::~pcom_worker): Likewise.
8698 (pcom_worker::release_chain): Adjust as auto_vec changes.
8699 (pcom_worker::loop): Rename to ...
8700 (pcom_worker::m_loop): ... this.
8701 (pcom_worker::datarefs): Rename to ...
8702 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
8703 (pcom_worker::dependences): Rename to ...
8704 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
8705 (pcom_worker::chains): Rename to ...
8706 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
8707 (pcom_worker::looparound_phis): Rename to ...
8708 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
8710 (pcom_worker::cache): Rename to ...
8711 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
8712 (pcom_worker::release_chain): Adjust for auto_vec changes.
8713 (pcom_worker::release_chains): Adjust for auto_vec and renaming
8715 (release_component): Remove.
8716 (release_components): Adjust for release_component removal.
8717 (component_of): Adjust to use vec.
8718 (merge_comps): Likewise.
8719 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
8720 (pcom_worker::determine_offset): Likewise.
8721 (class comp_ptrs): Remove.
8722 (pcom_worker::split_data_refs_to_components): Adjust for renaming
8723 changes, for comp_ptrs removal with auto_vec.
8724 (pcom_worker::suitable_component_p): Adjust for renaming changes.
8725 (pcom_worker::filter_suitable_components): Adjust for release_component
8727 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
8728 (pcom_worker::find_looparound_phi): Likewise.
8729 (pcom_worker::add_looparound_copies): Likewise.
8730 (pcom_worker::determine_roots_comp): Likewise.
8731 (pcom_worker::single_nonlooparound_use): Likewise.
8732 (pcom_worker::execute_pred_commoning_chain): Likewise.
8733 (pcom_worker::execute_pred_commoning): Likewise.
8734 (pcom_worker::try_combine_chains): Likewise.
8735 (pcom_worker::prepare_initializers_chain): Likewise.
8736 (pcom_worker::prepare_initializers): Likewise.
8737 (pcom_worker::prepare_finalizers_chain): Likewise.
8738 (pcom_worker::prepare_finalizers): Likewise.
8739 (pcom_worker::tree_predictive_commoning_loop): Likewise.
8741 2021-07-20 Martin Sebor <msebor@redhat.com>
8743 PR middle-end/101397
8744 * builtins.c (gimple_call_return_array): Add argument. Correct
8745 offsets for memchr, mempcpy, stpcpy, and stpncpy.
8746 (compute_objsize_r): Adjust offset computation for argument returning
8749 2021-07-20 Martin Sebor <msebor@redhat.com>
8751 PR middle-end/101300
8752 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
8754 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
8756 * function.c (assign_parm_setup_block): Use adjust_address instead
8757 of change_address to preserve MEM_EXPR and friends.
8759 2021-07-20 Martin Sebor <msebor@redhat.com>
8761 * cfgloop.h (single_likely_exit): Adjust by-value argument to
8763 * cfgloopanal.c (single_likely_exit): Same.
8764 * cgraph.h (struct cgraph_node): Same.
8765 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
8766 * genautomata.c (merge_states): Same.
8767 * genextract.c (VEC_char_to_string): Same.
8768 * genmatch.c (dt_node::gen_kids_1): Same.
8769 (walk_captures): Adjust by-value argument to by-reference.
8770 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
8771 to by-const-reference.
8772 * gimple.c (gimple_build_call_vec): Same.
8773 (gimple_build_call_internal_vec): Same.
8774 (gimple_build_switch): Same.
8775 (sort_case_labels): Same.
8776 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8778 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
8780 (gimple_build_call_internal_vec): Same.
8781 (gimple_build_switch): Same.
8782 (sort_case_labels): Same.
8783 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8785 * haifa-sched.c (calc_priorities): Adjust by-value argument to
8787 (sched_init_luids): Same.
8788 (haifa_init_h_i_d): Same.
8789 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
8790 (adjust_callers_for_value_intersection): Adjust by-value argument to
8792 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
8794 (find_more_contexts_for_caller_subset): Same.
8795 (find_aggregate_values_for_callers_subset): Same.
8796 (copy_useful_known_contexts): Same.
8797 * ipa-fnsummary.c (remap_edge_summaries): Same.
8798 (remap_freqcounting_predicate): Same.
8799 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
8801 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
8802 to by-const-reference.
8803 * ipa-predicate.h (predicate::remap_after_inlining): Same.
8804 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
8805 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
8806 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
8807 * read-rtl.c (add_overload_instance): Same.
8808 * rtl.h (native_decode_rtx): Same.
8809 (native_decode_vector_rtx): Same.
8810 * sched-int.h (sched_init_luids): Same.
8811 (haifa_init_h_i_d): Same.
8812 * simplify-rtx.c (native_decode_vector_rtx): Same.
8813 (native_decode_rtx): Same.
8814 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
8815 (shrink_wrap_one_built_in_call_with_conds): Same.
8816 (shrink_wrap_conditional_dead_built_in_calls): Same.
8817 * tree-data-ref.c (create_runtime_alias_checks): Same.
8818 (compute_all_dependences): Same.
8819 * tree-data-ref.h (compute_all_dependences): Same.
8820 (create_runtime_alias_checks): Same.
8821 (index_in_loop_nest): Same.
8822 * tree-if-conv.c (mask_exists): Same.
8823 * tree-loop-distribution.c (class loop_distribution): Same.
8824 (loop_distribution::create_rdg_vertices): Same.
8825 (dump_rdg_partitions): Same.
8826 (debug_rdg_partitions): Same.
8827 (partition_contains_all_rw): Same.
8828 (loop_distribution::distribute_loop): Same.
8829 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
8830 (oacc_entry_exit_single_gang): Same.
8831 * tree-ssa-loop-im.c (hoist_memory_references): Same.
8832 (loop_suitable_for_sm): Same.
8833 * tree-ssa-loop-niter.c (bound_index): Same.
8834 * tree-ssa-reassoc.c (update_ops): Same.
8835 (swap_ops_for_binary_stmt): Same.
8836 (rewrite_expr_tree): Same.
8837 (rewrite_expr_tree_parallel): Same.
8838 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
8839 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
8840 * tree-ssa-structalias.c (process_all_all_constraints): Same.
8841 (make_constraints_to): Same.
8842 (handle_lhs_call): Same.
8843 (find_func_aliases_for_builtin_call): Same.
8844 (sort_fieldstack): Same.
8845 (check_for_overlaps): Same.
8846 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
8847 (vect_create_cond_for_unequal_addrs): Same.
8848 (vect_create_cond_for_lower_bounds): Same.
8849 (vect_create_cond_for_alias_checks): Same.
8850 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
8851 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
8852 (vect_make_slp_decision): Same.
8853 (vect_slp_bbs): Same.
8854 (duplicate_and_interleave): Same.
8855 (vect_transform_slp_perm_load): Same.
8856 (vect_schedule_slp): Same.
8857 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
8858 (vect_schedule_slp): Same.
8859 (duplicate_and_interleave): Same.
8860 * tree.c (build_vector_from_ctor): Same.
8861 (build_vector): Same.
8862 (check_vector_cst): Same.
8863 (check_vector_cst_duplicate): Same.
8864 (check_vector_cst_fill): Same.
8865 (check_vector_cst_stepped): Same.
8866 * tree.h (build_vector_from_ctor): Same.
8868 2021-07-20 Jakub Jelinek <jakub@redhat.com>
8871 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
8872 type from bool to int.
8873 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
8874 EASY_VECTOR_MSB case if either step or copies is not 1.
8875 (vspltis_shifted): Fix comment typo.
8876 (easy_altivec_constant): Change return type from bool to int, instead
8877 of returning true return byte size of the element mode that should be
8878 used to synthetize the constant.
8879 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
8880 that vspltis_shifted is 0, handle the case where easy_altivec_constant
8881 assumes using different vector mode from CONST_VECTOR's mode.
8882 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
8883 easy_altivec_constant to determine mode in which -1 >> -1 should be
8884 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
8886 2021-07-20 Richard Biener <rguenther@suse.de>
8889 * dwarf2out.h (dwarf_file_data): Add key member.
8890 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
8891 (dwarf_file_hasher::hash): Hash key.
8892 (lookup_filename): Remap the filename and store it in the
8893 filename member of dwarf_file_data when creating a new
8895 (file_name_acquire): Do not remap the filename again.
8896 (maybe_emit_file): Likewise.
8898 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
8900 * config/aarch64/aarch64-simd-builtins.def: Use two variant
8901 generators for all TBL/TBX intrinsics and rename to
8902 consistent forms: qtbl[1234] or qtbx[1234].
8903 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
8905 (aarch64_qtbl1<mode>): This.
8906 (aarch64_tbx1<mode>): Rename to...
8907 (aarch64_qtbx1<mode>): This.
8908 (aarch64_tbl2v16qi): Delete.
8909 (aarch64_tbl3<mode>): Rename to...
8910 (aarch64_qtbl2<mode>): This.
8911 (aarch64_tbx4<mode>): Rename to...
8912 (aarch64_qtbx2<mode>): This.
8913 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
8914 renamed qtbl1 and qtbl2 RTL patterns.
8915 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
8917 (vqtbl1_s8): Likewise.
8918 (vqtbl1_u8): Likewise.
8919 (vqtbl1q_p8): Likewise.
8920 (vqtbl1q_s8): Likewise.
8921 (vqtbl1q_u8): Likewise.
8922 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
8923 (vqtbx1_u8): Likewise.
8924 (vqtbx1_p8): Likewise.
8925 (vqtbx1q_s8): Likewise.
8926 (vqtbx1q_u8): Likewise.
8927 (vqtbx1q_p8): Likewise.
8928 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
8929 (vtbl1_u8): Likewise.
8930 (vtbl1_p8): Likewise.
8931 (vtbl2_s8): Likewise
8932 (vtbl2_u8): Likewise.
8933 (vtbl2_p8): Likewise.
8934 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
8935 (vtbl3_u8): Likewise.
8936 (vtbl3_p8): Likewise.
8937 (vtbl4_s8): Likewise.
8938 (vtbl4_u8): Likewise.
8939 (vtbl4_p8): Likewise.
8940 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
8941 (vtbx2_u8): Likewise.
8942 (vtbx2_p8): Likewise.
8943 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
8944 (vqtbl2_u8): Likewise.
8945 (vqtbl2_p8): Likewise.
8946 (vqtbl2q_s8): Likewise.
8947 (vqtbl2q_u8): Likewise.
8948 (vqtbl2q_p8): Likewise.
8949 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
8950 (vqtbx2_u8): Likewise.
8951 (vqtbx2_p8): Likewise.
8952 (vqtbx2q_s8): Likewise.
8953 (vqtbx2q_u8): Likewise.
8954 (vqtbx2q_p8): Likewise.
8955 (vtbx4_s8): Likewise.
8956 (vtbx4_u8): Likewise.
8957 (vtbx4_p8): Likewise.
8959 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
8962 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
8964 (define_peephole2 atomic_loaddi_fpu): Ditto.
8966 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8968 * config.gcc (riscv*-*-*): Detect which python is available.
8970 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
8972 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
8973 (smul<mode>3_highpart): ... this.
8974 (mulhu_<mode>): Rename to...
8975 (umul<mode>3_highpart): ... this.
8976 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
8977 MULHU_V2DI, MULHU_V4SI): Adjust.
8979 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
8981 PR tree-optimization/100696
8982 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
8983 * internal-fn.def (IFN_MULH): New internal function.
8984 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
8985 recog normal multiply highpart as IFN_MULH.
8986 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
8989 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
8991 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
8992 (BTF_DEBUGGING_INFO): Likewise.
8993 * doc/tm.texi.in: Document the new macros.
8994 * doc/tm.texi: Regenerated.
8995 * toplev.c: Guard initialization of debug hooks.
8997 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
8999 * flags.h (ctf_debuginfo_p): New function declaration.
9000 * opts.c (ctf_debuginfo_p): New function definition.
9002 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
9005 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
9006 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
9007 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
9008 * config/gcn/gcn-valu.md: Add a warning comment.
9009 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
9010 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
9011 * config/gcn/gcn.md: Add a warning comment.
9012 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
9013 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
9015 (EF_AMDGPU_XNACK): New.
9016 (EF_AMDGPU_SRAM_ECC): New.
9018 (copy_early_debug_info): Use elf_flags.
9019 (main): Handle -mxnack and -msram-ecc options.
9020 * doc/invoke.texi: Document -mxnack and -msram-ecc.
9022 2021-07-19 Andrew Pinski <apinski@marvell.com>
9025 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
9026 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
9028 2021-07-19 Richard Biener <rguenther@suse.de>
9030 PR tree-optimization/101505
9031 * tree-vect-patterns.c (vect_determine_precisions): Walk
9032 PHIs also for loop vectorization.
9034 2021-07-19 Richard Biener <rguenther@suse.de>
9036 * gimple.h (gimple_expr_type): Remove.
9037 * doc/gimple.texi: Remove gimple_expr_type documentation.
9039 2021-07-19 Richard Biener <rguenther@suse.de>
9041 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
9042 (ao_ref_init_from_vn_reference): Likewise.
9043 (fully_constant_reference): Likewise.
9044 (vn_reference_lookup_call): Do not set vr->type to random
9046 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
9048 * tree-vect-generic.c (expand_vector_piecewise): Pass in
9049 whether we expanded parallel.
9050 (expand_vector_parallel): Adjust.
9051 (expand_vector_addition): Likewise.
9052 (expand_vector_comparison): Likewise.
9053 (expand_vector_operation): Likewise.
9054 (expand_vector_scalar_condition): Likewise.
9055 (expand_vector_conversion): Likewise.
9057 2021-07-19 Richard Biener <rguenther@suse.de>
9059 * tree-vrp.c (register_edge_assert_for_2): Use the
9061 (vrp_folder::fold_predicate_in): Likewise.
9062 * vr-values.c (gimple_assign_nonzero_p): Likewise.
9063 (vr_values::extract_range_from_comparison): Likewise.
9064 (vr_values::extract_range_from_ubsan_builtin): Use the
9065 type of the first operand.
9066 (vr_values::extract_range_basic): Push down type
9067 computation, use the appropriate LHS.
9068 (vr_values::extract_range_from_assignment): Use the
9071 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9074 * common/config/i386/i386-common.c (ix86_handle_option): For
9075 -mgeneral-regs-only, enable the GPR only instructions which are
9076 enabled implicitly by SSE ISAs unless they have been disabled
9079 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9082 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
9083 ix86_avx_u128_mode_needed.
9084 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
9085 returns AVX register.
9087 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
9089 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
9090 (handle_const_call): Liekise
9091 (handle_pure_call): Liekise
9093 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
9095 PR tree-optimization/96542
9096 * range-op.cc (range_operator::wi_fold_in_parts): New.
9097 (range_operator::fold_range): Call wi_fold_in_parts.
9098 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
9099 * range-op.h (wi_fold_in_parts): Add prototype.
9101 2021-07-16 David Malcolm <dmalcolm@redhat.com>
9103 * doc/analyzer.texi: Add __analyzer_dump_state.
9105 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9107 * config/rs6000/rbtree.c: New file.
9108 * config/rs6000/rbtree.h: New file.
9110 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9112 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
9113 (typeinfo): Add restr field.
9114 (match_bracketed_pair): New function.
9115 (match_const_restriction): Implement.
9117 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9119 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9121 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9123 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9124 (basetype): Likewise.
9125 (typeinfo): Likewise.
9126 (handle_pointer): New function.
9127 (match_basetype): New stub function.
9128 (match_const_restriction): Likewise.
9129 (match_type): New function.
9131 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9133 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9135 (advance_line): Likewise.
9136 (safe_inc_pos): Likewise.
9137 (match_identifier): Likewise.
9138 (match_integer): Likewise.
9139 (match_to_right_bracket): Likewise.
9141 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9143 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9144 (ovld_file): Likewise.
9145 (header_file): Likewise.
9146 (init_file): Likewise.
9147 (defines_file): Likewise.
9148 (pgm_path): Likewise.
9149 (bif_path): Likewise.
9150 (ovld_path): Likewise.
9151 (header_path): Likewise.
9152 (init_path): Likewise.
9153 (defines_path): Likewise.
9154 (LINELEN): New macro.
9155 (linebuf): New variable.
9159 (bif_diag): New function.
9160 (ovld_diag): Likewise.
9162 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9164 * config/rs6000/rs6000-builtin-new.def: New.
9165 * config/rs6000/rs6000-overload.def: New.
9167 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9169 * config/rs6000/rs6000-gen-builtins.c: New.
9171 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9173 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9174 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9175 * gengtype-state.c (state_writer::write_state_file_list): Add a
9176 parameter to the fileslist expression for the number of build
9178 (read_state_files_list): Detect build headers and strip the
9179 initial "./" or ".\" from their names.
9180 * gengtype.c (build_headers): New global variable.
9181 (num_build_headers): Likewise.
9182 (open_base_files): Emit #include for each build header.
9183 (main): Detect and count build headers.
9184 * gengtype.h (build_headers): New extern variable.
9185 (num_build_headers): Likewise.
9187 2021-07-16 Richard Biener <rguenther@suse.de>
9189 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9190 the type of the LHS.
9191 (find_bswap_or_nop_1): Likewise.
9192 (find_bswap_or_nop): Likewise.
9193 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9195 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9196 Remove unused parameters, pass in the scalar type. Fix
9197 internal store function handling.
9198 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9199 (vect_get_vector_types_for_stmt): Move down check for
9200 existing vector stmt after we've determined a scalar type.
9201 Pass down the used scalar type to vect_get_smallest_scalar_type.
9202 * tree-vect-generic.c (expand_vector_condition): Use
9203 the type of the LHS.
9204 (expand_vector_scalar_condition): Likewise.
9205 (expand_vector_operations_1): Likewise.
9206 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9207 (vect_recog_dot_prod_pattern): Likewise.
9208 (vect_recog_sad_pattern): Likewise.
9209 (vect_recog_widen_op_pattern): Likewise.
9210 (vect_recog_widen_sum_pattern): Likewise.
9211 (vect_recog_mixed_size_cond_pattern): Likewise.
9213 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9215 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9216 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9217 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9218 and EAF_NOT_RETURNED.
9219 (modref_summary::useful_p): Likewise.
9220 (modref_summary_lto::useful_p): Likewise.
9221 (struct) modref_summary_lto: Use eaf_fleags_t.
9222 (deref_flags): Handle EAF_NOT_RETURNED.
9223 (struct escape_point): Use min_flags.
9224 (modref_lattice::init): Add EAF_NOT_RETURNED.
9225 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9226 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9228 (analyze_parms): Also analyze const functions; update conition on
9230 (modref_write): Update streaming.
9231 (read_section): Update streaming.
9232 (remap_arg_flags): Use eaf_flags_t.
9233 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9234 * ipa-modref.h: (eaf_flags_t): New typedef.
9235 (struct modref_summary): Use eaf_flags_t.
9236 * tree-core.h (EAF_NOT_RETURNED): New constant.
9238 2021-07-16 Richard Biener <rguenther@suse.de>
9240 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
9241 the type of the LHS.
9242 (gimple_assign_nonnegative_warnv_p): Likewise.
9243 (gimple_call_nonnegative_warnv_p): Likewise. Return false
9244 if the call has no LHS.
9245 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
9246 * tree-eh.c (stmt_could_throw_1_p): Likewise.
9247 * tree-inline.c (insert_init_stmt): Likewise.
9248 * tree-ssa-loop-niter.c (get_val_for): Likewise.
9249 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
9251 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
9252 gassign *. Use the type of the lhs.
9253 (vn_nary_op_lookup_stmt): Adjust.
9254 (vn_nary_op_insert_stmt): Likewise.
9256 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9258 * config/s390/predicates.md (bras_sym_operand): Accept all
9259 functions in 64-bit mode, use UNSPEC_PLT31.
9260 (larl_operand): Use UNSPEC_PLT31.
9261 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
9262 (legitimize_pic_address): Likewise.
9263 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
9265 (s390_delegitimize_address): Use UNSPEC_PLT31.
9266 (s390_output_addr_const_extra): Likewise.
9267 (print_operand): Add @PLT to TLS calls, handle %K.
9268 (s390_function_profiler): Mark __fentry__/_mcount as function,
9269 use %K, use UNSPEC_PLT31.
9270 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9271 (s390_emit_call): Use UNSPEC_PLT31.
9272 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9273 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9274 (*movdi_64): Use %K.
9275 (reload_base_64): Likewise.
9276 (*sibcall_brc): Likewise.
9277 (*sibcall_brcl): Likewise.
9278 (*sibcall_value_brc): Likewise.
9279 (*sibcall_value_brcl): Likewise.
9282 (*bras_r): Likewise.
9283 (*brasl_r): Likewise.
9284 (*bras_tls): Likewise.
9285 (*brasl_tls): Likewise.
9286 (main_base_64): Likewise.
9287 (reload_base_64): Likewise.
9288 (@split_stack_call<mode>): Likewise.
9290 2021-07-16 Richard Biener <rguenther@suse.de>
9292 PR tree-optimization/101467
9293 * tree-vect-stmts.c (vect_gen_while): Properly guard
9294 make_temp_ssa_name usage.
9296 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9298 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9299 t-sysroot-suffix for C-SKY.
9300 * config/csky/print-sysroot-suffix.sh: Delete.
9301 * config/csky/t-csky-linux: Delete.
9302 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9303 instead of CSKY_MULTILIB_DIRNAMES.
9305 2021-07-16 Richard Biener <rguenther@suse.de>
9307 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9308 conversion issues with the partial reduction of the reused
9311 2021-07-16 Richard Biener <rguenther@suse.de>
9313 * config/i386/i386-options.c (ix86_option_override_internal): Set
9314 param_vect_partial_vector_usage to zero if not set.
9316 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9319 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9320 (VALID_INT_MODE_P): Add SDmode and DDmode.
9321 Add TDmode for TARGET_64BIT.
9322 (VALID_DFP_MODE_P): Remove.
9323 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9324 Do not use VALID_DFP_MODE_P.
9326 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9328 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9330 (fold_using_range::fold_stmt): Ditto.
9331 (fold_using_range::range_of_range_op): Ditto.
9332 (fold_using_range::range_of_phi): Ditto.
9333 (fold_using_range::range_of_call): Ditto.
9334 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9335 (fold_using_range::range_of_builtin_call): Ditto.
9336 (fold_using_range::range_of_cond_expr): Ditto.
9337 * gimple-range-fold.h (gimple_range_type): New.
9339 2021-07-15 Martin Sebor <msebor@redhat.com>
9342 * tree-ssa-strlen.c (handle_assign): New function.
9343 (maybe_warn_overflow): Add argument.
9344 (nonzero_bytes_for_type): New function.
9345 (count_nonzero_bytes): Handle more tree types. Call
9346 nonzero_bytes_for_tye.
9347 (count_nonzero_bytes): Handle types.
9348 (handle_store): Handle stores from function calls.
9349 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9350 it for assignments from function calls.
9352 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9357 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9359 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9361 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9363 2021-07-15 Martin Sebor <msebor@redhat.com>
9367 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9368 (operand_compare::verify_hash_value): Same.
9369 * tree-core.h (OEP_DECL_NAME): New.
9371 2021-07-15 Martin Jambor <mjambor@suse.cz>
9373 * profile-count.h (profile_count::value): Change the return type to
9375 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9377 * tree-cfg.c (dump_function_to_file): Likewise.
9379 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9382 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9383 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9386 2021-07-15 Richard Biener <rguenther@suse.de>
9388 * tree-vectorizer.h (vect_gen_while): Match up with
9390 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9391 of vect_gen_while_not.
9392 (vect_gen_while_not): Adjust.
9393 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9395 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9397 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9398 (ranger_cache::range_of_def): Call adjust_range.
9399 (ranger_cache::entry_range): Same.
9400 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9401 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9403 (gimple_ranger::range_on_entry): Same.
9405 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9408 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9410 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9412 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9415 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9417 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9418 (sdot_prod, udot_prod): ...These.
9419 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9420 (aarch64_<sur>dot<vsi2qi>): Rename to...
9421 (<sur>dot_prod<vsi2qi>): ...This.
9422 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9425 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9427 PR middle-end/101437
9428 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9429 types even if they have non-BLKmode TYPE_MODE.
9431 2021-07-15 Richard Biener <rguenther@suse.de>
9434 * gcc.c (process_command): Process -gtoggle like process_options
9435 would after parsing options.
9437 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9439 * cfgexpand.c (expand_asm_loc): Adjust.
9440 (expand_asm_stmt): Likewise.
9441 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9442 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9443 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9444 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9445 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9446 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9447 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9448 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9449 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9450 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9451 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9452 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9453 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9454 * doc/tm.texi: Regenerate.
9455 * target.def: Add location argument to md_asm_adjust.
9457 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9459 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9460 diagnostic's location, not input_location.
9462 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9464 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9466 (expand_asm_stmt): Likewise.
9468 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9470 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9471 addressed memory rtx, if any.
9472 (rs6000_split_multireg_move): Fix code formatting.
9473 Handle MMA build built-ins with operands in adjacent memory locations.
9475 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9477 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9480 2021-07-14 Jason Merrill <jason@redhat.com>
9482 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9484 2021-07-14 Jason Merrill <jason@redhat.com>
9486 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9489 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9491 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9492 Remove erroneous line.
9494 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9496 * params.opt (param_evrp_mode): Change default.
9498 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9500 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9501 (sdot_prod, udot_prod): ...These.
9502 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9503 (aarch64_<sur>dot<vsi2qi>): Rename to...
9504 (<sur>dot_prod<vsi2qi>): ...This.
9505 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9508 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9510 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9512 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9514 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9516 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9518 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9520 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9522 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9523 (usdot_prod<vsi2qi>): ... This.
9524 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9525 (usdot_prod): ...This.
9526 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9527 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9529 (@<sur>dot_prod<vsi2qi>): ...This.
9530 * config/aarch64/aarch64-sve-builtins-base.cc
9531 (svusdot_impl::expand): Use it.
9533 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9535 * optabs.def (usdot_prod_optab): New.
9536 * doc/md.texi: Document it and clarify other dot prod optabs.
9537 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9538 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9539 * optabs.c (expand_widen_pattern_expr): Likewise.
9540 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9541 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9542 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9544 (vect_widened_op_tree): Optionally ignore
9546 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9548 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9551 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9552 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9553 Enable UINTR only for 64-bit codegen.
9554 * config/i386/i386-options.c
9555 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9557 * config/i386/i386.h (ARCH_ARG): New.
9558 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9559 "[arch|tune] 64" for 64-bit codegen.
9561 2021-07-14 Richard Biener <rguenther@suse.de>
9563 PR tree-optimization/101445
9564 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9565 of the IV in the correct direction for negative stride
9568 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9571 * godump.c (godump_str_hash): New type.
9572 (godump_container::pot_dummy_types): Use string_hash instead of
9573 ptr_hash in the hash_set.
9575 2021-07-14 Richard Biener <rguenther@suse.de>
9577 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9578 vector types where the old vector type has a multiple of
9579 the new vector type elements.
9580 (vect_create_partial_epilog): New function, split out from...
9581 (vect_create_epilog_for_reduction): ... here.
9582 (vect_transform_cycle_phi): Reduce the re-used accumulator
9583 to the new vector type.
9585 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9587 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9588 non-'t'-sized arg check.
9590 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9592 * tree-cfg.c (cleanup_dead_labels_eh): Update
9593 post_landing_pad label upon change of landing pad block's
9595 (cleanup_dead_labels): Check that a removed label is not that
9598 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
9600 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
9602 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
9603 Add Neon to general purpose register case for zero-extend
9605 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
9606 case to prevent some cases opting to go through memory.
9607 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
9608 * rtl.c (rtvec_series_p): Define predicate to determine
9609 whether a vector contains a linear series of integers.
9610 * rtl.h (rtvec_series_p): Define.
9611 * rtlanal.c (vec_series_lowpart_p): Define predicate to
9612 determine if a vector selection is equivalent to the low part
9614 * rtlanal.h (vec_series_lowpart_p): Define.
9615 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
9616 Add vec_select -> subreg simplification.
9618 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
9620 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
9621 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
9622 _mm_test_mix_ones_zeros): New.
9624 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
9625 Richard Biener <rguenther@suse.de>
9627 * gimple.c (gimple_could_trap_p_1): Make S argument a
9628 "const gimple*". Preserve constness in call to
9629 gimple_asm_volatile_p.
9630 (gimple_could_trap_p): Make S argument a "const gimple*".
9631 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
9632 Update function prototypes.
9634 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9636 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
9637 (_loop_vec_info::main_loop_edge): New field.
9638 (_loop_vec_info::skip_main_loop_edge): Likewise.
9639 (_loop_vec_info::skip_this_loop_edge): Likewise.
9640 (_loop_vec_info::reusable_accumulators): Likewise.
9641 (_stmt_vec_info::reduc_scalar_results): Likewise.
9642 (_stmt_vec_info::reused_accumulator): Likewise.
9643 (vect_get_main_loop_result): Declare.
9644 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
9645 reduc_scalar_inputs.
9646 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
9647 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
9648 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
9649 skip_main_loop_edge and skip_this_loop_edge fields.
9650 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
9651 (vect_emit_reduction_init_stmts): New function.
9652 (get_initial_def_for_reduction): Use it.
9653 (get_initial_defs_for_reduction): Likewise. Change the vinfo
9654 parameter to a loop_vec_info.
9655 (vect_create_epilog_for_reduction): Store the scalar results
9656 in the reduc_info. If an epilogue loop is reusing an accumulator
9657 from the main loop, and if the epilogue loop can also be skipped,
9658 try to place the reduction code in the join block. Record
9659 accumulators that could potentially be reused by epilogue loops.
9660 (vect_transform_cycle_phi): When vectorizing epilogue loops,
9661 try to reuse accumulators from the main loop. Record the initial
9662 value in reduc_info for non-SLP reductions too.
9664 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9666 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
9667 adjustment handling. Take the neutral value as an argument,
9668 in place of the code argument.
9669 (vect_transform_cycle_phi): Update accordingly. Handle the
9670 initial values of cond reductions separately from code reductions.
9671 Choose the adjustment here rather than in
9672 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
9674 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9676 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
9677 (neutral_op_for_reduction): ...this, providing a more general
9679 (vect_create_epilog_for_reduction): Update accordingly.
9680 (vectorizable_reduction): Likewise.
9681 (vect_transform_cycle_phi): Likewise.
9683 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9685 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
9686 reduc_info instead of the original stmt_vec_info.
9687 (vect_transform_cycle_phi): Update accordingly.
9689 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9691 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
9692 reduc_info as an additional parameter.
9693 (vect_transform_cycle_phi): Update accordingly.
9695 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9697 * tree-vectorizer.h: Include tree-ssa-operands.h.
9698 (vect_phi_initial_value): New function.
9699 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
9700 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
9701 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9702 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
9704 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9706 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
9707 the phi results to vectype after creating them. Remove later
9708 conversion code that thus becomes redundant.
9710 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9712 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
9713 the new_phis vector with a reduc_inputs vector. Combine handling
9714 of reduction chains and ncopies > 1.
9716 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9718 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
9719 scalar_results to group_size elements after reducing down from
9720 N*group_size elements. Construct an array_slice of the live-out
9721 stmts and assert that there is one stmt per scalar result.
9723 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9725 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
9726 nested_in_vect_loop and use double_reduc everywhere. Remove dead
9727 assignment to "loop".
9729 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9731 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
9732 vector types first. For scalar types, consider both the preferred
9733 vector mode and the alternative vector modes.
9734 * optabs-query.c (can_vec_mask_load_store_p): Use the same
9735 structure as above, in particular using related_vector_mode
9736 for modes provided by autovectorize_vector_modes.
9738 2021-07-13 Jakub Jelinek <jakub@redhat.com>
9739 Richard Biener <rguenther@suse.de>
9741 PR tree-optimization/101419
9742 * tree-pass.h (PROP_objsz): Define.
9743 (make_pass_early_object_sizes): Declare.
9744 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
9745 there to pass_early_object_sizes, drop parameter.
9746 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
9747 drop parameter, move pass_post_ipa_warn right after that.
9748 * tree-object-size.c (pass_object_sizes::execute): Rename to...
9749 (object_sizes_execute): ... this. Add insert_min_max_p argument.
9750 (pass_data_object_sizes): Move after object_sizes_execute.
9751 (pass_object_sizes): Likewise. In execute method call
9752 object_sizes_execute, drop set_pass_param method and insert_min_max_p
9753 non-static data member and its initializer in the ctor.
9754 (pass_data_early_object_sizes, pass_early_object_sizes,
9755 make_pass_early_object_sizes): New.
9756 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
9757 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
9759 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
9762 * config/riscv/constraints.md ("S"): Update description and remove
9764 * doc/md.texi (Machine Constraints): Document the 'S' constraints
9767 2021-07-13 Richard Biener <rguenther@suse.de>
9770 2021-07-12 Richard Biener <rguenther@suse.de>
9772 * tree-vect-slp.c (vect_slp_region): Show the number of
9773 SLP graph entries in the optimization message.
9775 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
9777 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
9779 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
9780 return type to long.
9781 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
9784 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9786 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
9787 Query relation between the 2 operands and use it.
9789 2021-07-12 Sergei Trofimovich <siarheit@google.com>
9791 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
9793 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9796 * config/i386/predicates.md (vec_setm_sse41_operand):
9797 Rename from vec_setm_operand.
9798 (vec_setm_avx2_operand): New predicate.
9799 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
9800 Use vec_setm_sse41_operand as operand 2 predicate.
9801 (vec_set<V_256_512:mode): New expander.
9802 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
9803 as operand 2 predicate.
9805 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9807 PR tree-optimization/101335
9808 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
9810 2021-07-12 Andrew Pinski <apinski@marvell.com>
9812 * tree-ssa-phiopt.c (match_simplify_replacement): Move
9813 insert of the sequence before the movement of the
9814 statement. Check if to see if the statement is used
9815 outside of the original phi to see if we should move it.
9817 2021-07-12 Richard Biener <rguenther@suse.de>
9819 * dump-context.h (debug_dump_context::debug_dump_context):
9820 Add FILE * parameter defaulted to stderr.
9821 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
9822 * tree-vect-slp.c (dot_slp_tree): New functions.
9824 2021-07-12 Richard Biener <rguenther@suse.de>
9826 PR tree-optimization/101373
9827 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
9828 references when the BB may not return.
9829 (compute_avail): Pass in the function we're working on and
9830 replace cfun references with it. Externally throwing
9831 const calls also possibly terminate the function.
9832 (pass_pre::execute): Pass down the function we're working on.
9833 * gcse.c (compute_hash_table_work): Externally throwing
9834 const/pure calls also need record_last_mem_set_info.
9835 * postreload-gcse.c (record_opr_changes): Looping or externally
9836 throwing const/pure calls also need record_last_mem_set_info.
9838 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9840 * recog.c (memory_address_addr_space_p): Change the type to bool.
9841 Return true/false instead of 1/0.
9842 (offsettable_memref_p): Ditto.
9843 (offsettable_nonstrict_memref_p): Ditto.
9844 (offsettable_address_addr_space_p): Ditto.
9845 Change the type of addressp indirect function to bool.
9846 * recog.h (memory_address_addr_space_p): Change the type to bool.
9847 (strict_memory_address_addr_space_p): Ditto.
9848 (offsettable_memref_p): Ditto.
9849 (offsettable_nonstrict_memref_p): Ditto.
9850 (offsettable_address_addr_space_p): Ditto.
9851 * reload.c (maybe_memory_address_addr_space_p): Ditto.
9852 (strict_memory_address_addr_space_p): Change the type to bool.
9853 Return true/false instead of 1/0.
9854 (maybe_memory_address_addr_space_p): Change the type to bool.
9856 2021-07-12 Richard Biener <rguenther@suse.de>
9858 * tree-vect-slp.c (vect_slp_region): Show the number of
9859 SLP graph entries in the optimization message.
9861 2021-07-12 Richard Biener <rguenther@suse.de>
9863 PR tree-optimization/101394
9864 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
9865 copies from abnormals for a full redundancy.
9867 2021-07-12 Richard Biener <rguenther@suse.de>
9869 PR middle-end/101423
9870 * gimple.c (gimple_could_trap_p_1): Internal function calls
9872 * tree-eh.c (tree_could_trap_p): Likewise.
9874 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9877 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
9879 (vmulq_n_u32): Likewise.
9880 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
9881 (vmulq_n_f32): Likewise.
9882 (vmul_n_f16): Likewise.
9883 (vmulq_n_f16): Likewise.
9885 2021-07-12 Martin Liska <mliska@suse.cz>
9888 * gcc.c (check_offload_target_name): Call
9889 candidates_list_and_hint only if we have a candidate.
9891 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9894 * config/arm/neon.md (vec_init): Move to ...
9895 * config/arm/vec-common.md (vec_init): ... here.
9896 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
9898 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
9900 PR tree-optimization/101403
9901 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
9902 signedness of the shift is not the same as the signedness of
9905 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
9906 Uroš Bizjak <ubizjak@gmail.com>
9908 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
9909 divmod of a constant numerator with new define_insn_and_split.
9911 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
9914 * config/i386/i386-expand.c (ix86_expand_call): If a call is
9915 to a non-local-binding, or local but to a public symbol, then
9916 assume that it might be indirected via the lazy symbol binder.
9917 Mark R10 and R10 as clobbered in that case.
9919 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
9922 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
9923 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
9924 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
9926 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
9928 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
9929 log2 range of operands[3] to [1,31].
9930 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
9932 2021-07-09 Sergei Trofimovich <siarheit@google.com>
9934 * doc/md.texi: Don't split @smallexample in multiple @groups.
9936 2021-07-09 Sergei Trofimovich <siarheit@google.com>
9938 * doc/md.texi: Add missing 'see' word.
9940 2021-07-09 Andrew Pinski <apinski@marvell.com>
9942 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
9943 to take sequence and gimple_match_op. Accept the case where
9944 op is a SSA_NAME and one statement in the sequence.
9945 Also allow constants.
9946 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
9947 Update call to phiopt_early_allow. Discard the sequence if not
9950 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
9955 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
9956 instead of ISA_HAS_MSA.
9957 (mips_expand_vec_unpack): Likewise.
9958 (mips_expand_vector_init): Likewise.
9960 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
9962 * config/rs6000/vsx.md (mods_<mode>): Rename to...
9963 (mod<mode>3): ... this.
9964 (modu_<mode>): Rename to...
9965 (umod<mode>3): ... this.
9966 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
9969 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
9971 * config/h8300/shiftrotate.md (variable shifts): Expose condition
9972 code handling for the test before the loop.
9974 2021-07-08 Martin Jambor <mjambor@suse.cz>
9977 * ipa-sra.c (class isra_call_summary): New member
9978 m_before_any_store, initialize it in the constructor.
9979 (isra_call_summary::dump): Dump the new field.
9980 (ipa_sra_call_summaries::duplicate): Copy it.
9981 (process_scan_results): Set it.
9982 (isra_write_edge_summary): Stream it.
9983 (isra_read_edge_summary): Likewise.
9984 (param_splitting_across_edge): Only override
9985 safe_to_import_accesses if m_before_any_store is set.
9987 2021-07-08 Martin Sebor <msebor@redhat.com>
9990 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
9991 Use Object Size Type 0 instead of 1.
9993 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
9995 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
9998 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10000 * match.pd: Simplify an extend-operate-truncate sequence involving
10003 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
10004 Richard Biener <rguenther@suse.de>
10006 PR tree-optimization/40210
10007 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
10008 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
10009 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
10011 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
10014 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
10016 * config/i386/mmx.md (V_32): New mode iterator.
10017 (mov<V_32:mode>): Use V_32 mode iterator.
10018 (*mov<V_32:mode>_internal): Ditto.
10019 (*push<V_32:mode>2_rex64): Ditto.
10020 (*push<V_32:mode>2): Ditto.
10021 (movmisalign<V_32:mode>): Ditto.
10022 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
10023 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
10024 (vec_unpacks_lo_v4qi): New expander.
10025 (vec_unpacks_hi_v4qi): Ditto.
10026 (vec_unpacku_lo_v4qi): Ditto.
10027 (vec_unpacku_hi_v4qi): Ditto.
10028 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
10029 (VALID_INT_MODE_P): Ditto.
10031 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
10034 * config/rs6000/rs6000.md (udivti3): New insn.
10035 (divti3): New insn.
10036 (umodti3): New insn.
10037 (modti3): New insn.
10039 2021-07-07 Martin Sebor <msebor@redhat.com>
10041 PR tree-optimization/100137
10042 PR tree-optimization/99121
10043 PR tree-optimization/97027
10044 * builtins.c (access_ref::access_ref): Also set offmax.
10045 (access_ref::offset_in_range): Define new function.
10046 (access_ref::add_offset): Set offmax.
10047 (access_ref::inform_access): Handle access_none.
10048 (handle_mem_ref): Clear ostype.
10049 (compute_objsize_r): Handle ASSERT_EXPR.
10050 * builtins.h (struct access_ref): Add offmax member.
10051 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
10052 compute_objsize() and simplify.
10054 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
10056 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
10057 and VSX_BUILTIN_STXVP.
10059 2021-07-07 Martin Sebor <msebor@redhat.com>
10062 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
10063 a stray %K from error_at() missed in r12-2088.
10065 2021-07-07 Richard Biener <rguenther@suse.de>
10067 PR tree-optimization/99728
10068 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
10070 (mem_refs_may_alias_p): Add assert we handled aggregate
10072 (sm_seq_valid_bb): Give up when running into aggregate copies.
10073 (ref_indep_loop_p): Handle aggregate copies as never
10074 being invariant themselves but allow other refs to be
10075 disambiguated against them.
10076 (can_sm_ref_p): Do not try to apply store-motion to aggregate
10079 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
10082 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
10085 2021-07-06 Martin Sebor <msebor@redhat.com>
10087 * gimple-pretty-print.c (percent_G_format): Remove.
10088 * tree-diagnostic.c (default_tree_printer): Remove calls.
10089 * tree-pretty-print.c (percent_K_format): Remove.
10090 * tree-pretty-print.h (percent_K_format): Remove.
10092 2021-07-06 Martin Sebor <msebor@redhat.com>
10094 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
10095 Remove %K and use error_at.
10096 (aarch64_expand_fcmla_builtin): Same.
10097 (aarch64_expand_builtin_tme): Same.
10098 (aarch64_expand_builtin_memtag): Same.
10099 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
10100 (arm_expand_builtin): Same.
10101 * config/arm/arm.c (bounds_check): Same.
10103 2021-07-06 Martin Sebor <msebor@redhat.com>
10105 * builtins.c (warn_string_no_nul): Remove %G.
10106 (maybe_warn_for_bound): Same.
10107 (warn_for_access): Same.
10108 (check_access): Same.
10109 (check_strncat_sizes): Same.
10110 (expand_builtin_strncat): Same.
10111 (expand_builtin_strncmp): Same.
10112 (expand_builtin): Same.
10113 (expand_builtin_object_size): Same.
10114 (warn_dealloc_offset): Same.
10115 (maybe_emit_free_warning): Same.
10116 * calls.c (maybe_warn_alloc_args_overflow): Same.
10117 (maybe_warn_nonstring_arg): Same.
10118 (maybe_warn_rdwr_sizes): Same.
10119 * expr.c (expand_expr_real_1): Remove %K.
10120 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10121 (gimple_fold_builtin_strncat): Same.
10122 * gimple-ssa-sprintf.c (format_directive): Same.
10123 (handle_printf_call): Same.
10124 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10125 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10126 (maybe_diag_access_bounds): Same. Call gimple_location.
10127 (check_bounds_or_overlap): Same.
10128 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10129 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10130 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10131 (maybe_diag_stxncpy_trunc): Same.
10132 (handle_builtin_stxncpy_strncat): Same.
10133 (maybe_warn_pointless_strcmp): Same.
10134 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10136 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10139 * config/i386/predicates.md (vec_setm_operand): Enable
10140 register_operand for TARGET_SSE4_1.
10141 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10142 as operand 2 predicate. Call ix86_expand_vector_set_var
10143 for non-constant index operand.
10144 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10145 Call ix86_expand_vector_set_var for non-constant index operand.
10147 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10149 * config/h8300/jumpcall.md (*branch): When possible, generate
10150 the comparison in CCZN mode.
10151 * config/h8300/predicates.md (simple_memory_operand): Reject all
10152 auto-increment addressing modes.
10154 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10156 PR bootstrap/100246
10157 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10160 2021-07-06 Richard Biener <rguenther@suse.de>
10162 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10163 (vec_fmsubadd<mode>4): Likewise.
10164 * optabs.def (vec_fmaddsub$a4): Add.
10165 (vec_fmsubadd$a4): Likewise.
10166 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10167 (IFN_VEC_FMSUBADD): Likewise.
10168 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10169 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10170 (addsub_pattern::build): Likewise.
10171 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10172 and CFN_VEC_FMSUBADD are not transparent for permutes.
10173 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10174 (vec_fmsubadd<mode>4): Likewise.
10176 2021-07-06 Richard Biener <rguenther@suse.de>
10178 * doc/invoke.texi (fmove-loop-stores): Document.
10179 * common.opt (fmove-loop-stores): New option.
10180 * opts.c (default_options_table): Enable -fmove-loop-stores
10181 at -O1 but not -Og.
10182 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10183 flag_move_loop_stores instead of true to
10184 loop_invariant_motion_in_fun.
10186 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10188 * doc/install.texi: Document --with-dsymutil.
10190 2021-07-06 Andrew Pinski <apinski@marvell.com>
10192 PR tree-optimization/101256
10193 * dbgcnt.def (phiopt_edge_range): New counter.
10194 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10195 Check to make sure the new name is defined in the same
10196 bb as the conditional before duplicating range info.
10197 Also add debug counter.
10199 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10201 PR rtl-optimization/100328
10202 * config/i386/i386-options.c (ix86_option_override_internal):
10203 Set param_ira_consider_dup_in_all_alts to 0.
10205 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10207 PR rtl-optimization/100328
10208 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10210 * ira.c (ira_get_dup_out_num): Adjust as parameter
10211 param_ira_consider_dup_in_all_alts.
10212 * params.opt (ira-consider-dup-in-all-alts): New.
10213 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10214 single_input_op_has_cstr_p.
10215 (get_freq_for_shuffle_copy): New function.
10216 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10217 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10219 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10221 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10222 generate condition code aware RTL directly.
10224 2021-07-05 Andrew Pinski <apinski@marvell.com>
10226 PR tree-optimization/101039
10227 * match.pd (A CMP 0 ? A : -A): New patterns.
10228 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10229 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10230 Update comment about abs_replacement.
10232 2021-07-05 Andrew Pinski <apinski@marvell.com>
10234 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10235 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
10237 2021-07-05 Andrew Pinski <apinski@marvell.com>
10239 * tree-ssa-phiopt.c (match_simplify_replacement):
10240 Add early_p argument. Call gimple_simplify_phiopt
10241 instead of gimple_simplify.
10242 (tree_ssa_phiopt_worker): Update call to
10243 match_simplify_replacement and allow unconditionally.
10244 (phiopt_early_allow): New function.
10245 (gimple_simplify_phiopt): New function.
10247 2021-07-05 Andrew Pinski <apinski@marvell.com>
10249 PR middle-end/101237
10250 * fold-const.c (negate_expr_p): Remove call to element_mode
10251 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
10252 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
10253 (fold_negate_expr_1): Likewise.
10254 (const_unop): Likewise.
10255 (fold_cond_expr_with_comparison): Likewise.
10256 (fold_binary_loc): Likewise.
10257 (fold_ternary_loc): Likewise.
10258 (tree_call_nonnegative_warnv_p): Likewise.
10259 * match.pd (-(A + B) -> (-B) - A): Likewise.
10261 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
10263 * configure.ac: Handle --with-dsymutil in the same way as we
10264 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
10265 Extract the type and version for the dsymutil configured or
10266 found by the default searches.
10267 * config.in: Regenerated.
10268 * configure: Regenerated.
10269 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10270 same way as for the assembler and linker.
10271 * config/darwin.h (DSYMUTIL): Delete.
10272 * gcc.c: Report a configured dsymutil correctly.
10273 * exec-tool.in: Allow for dsymutil.
10275 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10277 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10278 Handle V4QI and V2HI modes.
10279 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10280 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10281 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10282 Handle 4-byte dual operands with TARGET_XOP and single operands
10283 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10284 mmx_pshufbv4qi3 for TARGET_SSSE3.
10285 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10286 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10287 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10288 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10289 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10290 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10291 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10292 (mmx_pshufbv4qi3): Ditto.
10293 (*mmx_pblendw32): Ditto.
10294 (*mmx_pblendw64): Rename from *mmx_pblendw.
10295 (mmx_punpckhbw_low): New insn_and_split pattern.
10296 (mmx_punpcklbw_low): Ditto.
10298 2021-07-05 Richard Biener <rguenther@suse.de>
10300 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10301 set LOOP_C_INFINITE on the vectorized loop.
10303 2021-07-05 Richard Biener <rguenther@suse.de>
10305 PR middle-end/101291
10306 * cfgloopmanip.c (loop_version): Set the loop copy of the
10307 versioned loop to the new loop.
10309 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10312 * config.gcc: Ensure that Darwin biarch definitions are
10313 added before i386.h.
10314 * config/i386/darwin.h (TARGET_64BIT): Remove.
10315 (PR80556_WORKAROUND): New.
10316 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10317 (DARWIN_SUBARCH_SPEC): New.
10318 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10319 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10320 (REAL_LIBGCC_SPEC): Remove.
10321 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10322 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10323 (REAL_LIBGCC_SPEC): Remove.
10325 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10327 PR middle-end/101294
10328 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10330 2021-07-02 Martin Sebor <msebor@redhat.com>
10332 PR middle-end/98871
10333 PR middle-end/98512
10334 * diagnostic.c (get_any_inlining_info): New.
10335 (update_effective_level_from_pragmas): Handle inlining context.
10336 (diagnostic_enabled): Same.
10337 (diagnostic_report_diagnostic): Same.
10338 * diagnostic.h (struct diagnostic_info): Add ctor.
10339 (struct diagnostic_context): Add new member.
10340 * tree-diagnostic.c (set_inlining_locations): New.
10341 (tree_diagnostics_defaults): Set new callback pointer.
10343 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10345 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10347 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10348 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10349 lxvp and stxvp built-ins.
10350 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10351 (builtin_function_type): Likewise.
10352 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10354 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10356 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10357 additional argument for the code.
10358 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10359 argument for the code. Just return if the ZN bits are useful or
10360 not rather than the old style CC_* enums.
10361 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10362 more generic shiftqi patterns.
10363 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10364 (shiftqi_noscratch_set_flags): New pattern.
10365 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10367 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10369 PR tree-optimization/101223
10370 * range-op.cc (build_lt): Add -1 for signed values.
10371 (built_gt): Subtract -1 for signed values.
10373 2021-07-02 David Faust <david.faust@oracle.com>
10375 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10376 (btf_asm_type): Likewise.
10378 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10380 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10381 an array of rtx rather than a pointer to rtx. Add code argument.
10382 (compute_a_shift_length): Similarly.
10383 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10384 shift itself isn't an operand. Create dummy operand[0] to carry
10385 a mode and pass a suitable rtx code to compute_a_shift_length.
10386 (get_shift_alg): Adjust operand number of clobber in output templates.
10387 (output_a_shift): Make first argument an array of rtx rather than
10388 a pointer to rtx. Add code argument for the type of shift.
10389 Adjust now that the shift itself is no longer an operand.
10390 (compute_a_shift_length): Similarly.
10391 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10392 iterator rather than nshift_operator.
10393 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10394 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10395 and compute_a_shift_length.
10396 (shiftqi_noscratch_clobber_flags): Likewise.
10397 (shifthi_noscratch_clobber_flags): Likewise.
10398 (shiftsi_noscratch_clobber_flags): Likewise.
10400 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10403 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10404 dsymutil for BTF/CTF.
10406 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10409 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10410 segment to include BTF.
10411 (BTF_INFO_SECTION_NAME): New.
10413 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10415 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10416 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10417 * config/m32r/m32r.c (call_operand): Adjust return type.
10418 (small_data_operand, memreg_operand): Likewise.
10420 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10422 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10424 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10425 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10426 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10427 (move_source_operand, move_destination_operand): Likewise.
10428 (condexec_source_operand, condexec_dest_operand): Likewise.
10429 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10430 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10431 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10432 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10433 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10434 (even_fpr_operand, odd_fpr_operand): Likewise.
10435 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10436 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10437 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10438 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10439 (relational_operator, float_relational_operator): Likewise.
10440 (ccr_eqne_operator, minmax_operator): Likewise.
10441 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10442 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10443 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10444 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10445 (quad_acc_operand, accg_operand): Likewise.
10447 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10449 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10450 return type to a bool.
10451 (nonimmediate_nonstack_operand): Likewise.
10452 (xstormy16_splittable_below100_operand): Likewise.
10453 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10455 (xstormy16_splittable_below100_operand): Likewise.
10457 2021-07-02 Richard Biener <rguenther@suse.de>
10459 PR tree-optimization/101293
10460 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10461 with combined offsets.
10462 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10463 combined with the rest of the offset.
10465 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10467 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10468 PIC encodings for PE-COFF targets.
10470 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10473 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10474 Return nullptr for TImode inner mode.
10476 2021-07-02 Richard Biener <rguenther@suse.de>
10478 PR tree-optimization/101280
10479 PR tree-optimization/101173
10480 * gimple-loop-interchange.cc
10481 (tree_loop_interchange::valid_data_dependences): Properly
10482 guard all dependence checks with DDR_REVERSED_P or its
10485 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10487 * config/i386/i386-expand.c (ix86_expand_builtin):
10488 Add branch to clear odata when ZF is set for asedecenc_expand
10489 and wideaesdecenc_expand.
10491 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10493 * config/i386/gcc-auto-profile: regenerate
10495 2021-07-02 liuhongt <hongtao.liu@intel.com>
10497 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10498 (trunc<mode><pmov_dst_4_lower>2): this.
10500 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10502 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10503 * input.c (class fcache): Rename to...
10504 (class file_cache_slot): ...this, making most members private and
10505 prefixing fields with "m_".
10506 (file_cache_slot::get_file_path): New accessor.
10507 (file_cache_slot::get_use_count): New accessor.
10508 (file_cache_slot::missing_trailing_newline_p): New accessor.
10509 (file_cache_slot::inc_use_count): New.
10510 (fcache_buffer_size): Move to...
10511 (file_cache_slot::buffer_size): ...here.
10512 (fcache_line_record_size): Move to...
10513 (file_cache_slot::line_record_size): ...here.
10514 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10515 (fcache_tab_size): Move to file_cache::num_file_slots.
10516 (diagnostic_file_cache_init): Update for move of fcache_tab
10517 to global_dc->m_file_cache.
10518 (diagnostic_file_cache_fini): Likewise.
10519 (lookup_file_in_cache_tab): Convert to...
10520 (file_cache::lookup_file): ...this.
10521 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10522 fcache_tab to global_dc->m_file_cache, moving most of
10523 implementation to...
10524 (file_cache::forcibly_evict_file): ...this new function and...
10525 (file_cache_slot::evict): ...this new function.
10526 (evicted_cache_tab_entry): Convert to...
10527 (file_cache::evicted_cache_tab_entry): ...this.
10528 (add_file_to_cache_tab): Convert to...
10529 (file_cache::add_file): ...this, moving bulk of implementation
10531 (file_cache_slot::create): ..this new function.
10532 (file_cache::file_cache): New.
10533 (file_cache::~file_cache): New.
10534 (lookup_or_add_file_to_cache_tab): Convert to...
10535 (file_cache::lookup_or_add_file): ..this new function.
10536 (fcache::fcache): Rename to...
10537 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10539 (fcache::~fcache): Rename to...
10540 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10542 (needs_read): Convert to...
10543 (file_cache_slot::needs_read_p): ...this.
10544 (needs_grow): Convert to...
10545 (file_cache_slot::needs_grow_p): ...this.
10546 (maybe_grow): Convert to...
10547 (file_cache_slot::maybe_grow): ...this.
10548 (read_data): Convert to...
10549 (file_cache_slot::read_data): ...this.
10550 (maybe_read_data): Convert to...
10551 (file_cache_slot::maybe_read_data): ...this.
10552 (get_next_line): Convert to...
10553 (file_cache_slot::get_next_line): ...this.
10554 (goto_next_line): Convert to...
10555 (file_cache_slot::goto_next_line): ...this.
10556 (read_line_num): Convert to...
10557 (file_cache_slot::read_line_num): ...this.
10558 (location_get_source_line): Update for moving of globals to
10559 global_dc->m_file_cache.
10560 (location_missing_trailing_newline): Likewise.
10561 * input.h (class file_cache_slot): New forward decl.
10562 (class file_cache): New.
10564 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10566 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10567 128-bit floating point conditional move support.
10568 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10570 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10571 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10572 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10573 (fpmask<mode>, IEEE128 iterator): New insn.
10574 (xxsel<mode>, IEEE128 iterator): New insn.
10576 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10579 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10581 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10583 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10585 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10587 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10588 (vec_duplicate<mode>): New expander.
10590 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10593 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10595 (ix86_byte_broadcast): New function.
10596 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10597 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10598 size is 16 bytes or bigger.
10599 (ix86_broadcast_from_integer_constant): New function.
10600 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
10601 to broadcast if mode size is 16 bytes or bigger.
10602 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
10604 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
10606 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10608 * config/i386/predicates.md (ix86_endbr_immediate_operand):
10609 Return true/false instead of 1/0.
10610 (movq_parallel): Ditto.
10612 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10614 * recog.c (general_operand): Return true/false instead of 1/0.
10615 (register_operand): Ditto.
10616 (immediate_operand): Ditto.
10617 (const_int_operand): Ditto.
10618 (const_scalar_int_operand): Ditto.
10619 (const_double_operand): Ditto.
10620 (push_operand): Ditto.
10621 (pop_operand): Ditto.
10622 (memory_operand): Ditto.
10623 (indirect_operand): Ditto.
10625 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10627 * genpreds.c (write_predicate_subfunction):
10628 Change the type of written subfunction to bool.
10629 (write_one_predicate_function):
10630 Change the type of written function to bool.
10631 (write_tm_preds_h): Ditto.
10632 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
10633 * recog.c (general_operand): Change the type to bool.
10634 (address_operand): Ditto.
10635 (register_operand): Ditto.
10636 (pmode_register_operand): Ditto.
10637 (scratch_operand): Ditto.
10638 (immediate_operand): Ditto.
10639 (const_int_operand): Ditto.
10640 (const_scalar_int_operand): Ditto.
10641 (const_double_operand): Ditto.
10642 (nonimmediate_operand): Ditto.
10643 (nonmemory_operand): Ditto.
10644 (push_operand): Ditto.
10645 (pop_operand): Ditto.
10646 (memory_operand): Ditto.
10647 (indirect_operand): Ditto.
10648 (ordered_comparison_operator): Ditto.
10649 (comparison_operator): Ditto.
10650 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
10651 Change the type of indirect predicate function to bool.
10652 * config/rs6000/rs6000.c (easy_vector_constant):
10653 Change the type to bool.
10654 * config/mips/mips-protos.h (m16_based_address_p):
10655 Change the type of operand 3 to bool.
10657 2021-07-01 Richard Biener <rguenther@suse.de>
10659 PR tree-optimization/101280
10660 PR tree-optimization/101173
10661 * gimple-loop-interchange.cc
10662 (tree_loop_interchange::valid_data_dependences): Revert
10663 previous change and instead correctly handle DDR_REVERSED_P
10666 2021-07-01 Richard Biener <rguenther@suse.de>
10668 PR tree-optimization/101278
10669 * tree-ssa-dse.c (dse_classify_store): First check for
10670 uses, then ignore stmt for chaining purposes.
10672 2021-07-01 Richard Biener <rguenther@suse.de>
10674 PR tree-optimization/100778
10675 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
10676 vectorized ops ahead of their scalar BB.
10678 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10681 * config/i386/i386.md (*nabs<dwi>2_doubleword):
10682 New insn_and_split pattern.
10683 (*nabs<dwi>2_1): Ditto.
10684 * config/i386/i386-features.c
10685 (general_scalar_chain::compute_convert_gain):
10686 Handle (NEG (ABS (...))) RTX. Rewrite src code
10687 scanner as switch statement.
10688 (general_scalar_chain::convert_insn):
10689 Handle (NEG (ABS (...))) RTX.
10690 (general_scalar_to_vector_candidate_p):
10691 Detect (NEG (ABS (...))) RTX. Reorder case statements
10692 for (AND (NOT (...) ...)) fallthrough.
10694 2021-07-01 Richard Biener <rguenther@suse.de>
10696 PR tree-optimization/101178
10697 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
10698 (slpg::perm_in): Add.
10699 (slpg::get_perm_in): Remove.
10700 (slpg::get_perm_materialized): Add.
10701 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
10702 during permute propagation and materialization.
10704 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10707 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
10709 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10711 PR middle-end/94366
10712 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
10713 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
10714 use boolean_type_node instead of integer_type_node as NE_EXPR type.
10715 (lower_reduction_clauses): Likewise.
10717 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10719 * config/gcn/gcn.c: Include dwarf2.h.
10720 (gcn_addr_space_debug): New function.
10721 (TARGET_ADDR_SPACE_DEBUG): New hook.
10723 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10725 * common/config/gcn/gcn-common.c
10726 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
10727 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
10729 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
10730 (gcn_frame_pointer_rqd): New function.
10731 (TARGET_FRAME_POINTER_REQUIRED): New hook.
10733 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10735 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
10736 prologue register saves.
10737 (gcn_debug_unwind_info): Use UI_DWARF2.
10738 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
10739 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
10740 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
10741 (DWARF_LINK_REGISTER): New define.
10742 (FIRST_PSEUDO_REGISTER): Increment.
10743 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
10744 (CALL_USED_REGISTERS): Likewise.
10745 (REGISTER_NAMES): Likewise.
10747 2021-06-30 Richard Biener <rguenther@suse.de>
10749 PR tree-optimization/101267
10750 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
10751 API and use SLP compatible interface of vect_is_simple_use.
10752 Reject not vectorized SLP defs for callers that do not support
10754 (vect_check_store_rhs): Handle masked stores and pass down
10755 the appropriate operator index.
10756 (vectorizable_call): Adjust.
10757 (vectorizable_store): Likewise.
10758 (vectorizable_load): Likewise. Handle SLP pecularity of
10760 (vect_is_simple_use): Remove special-casing of masked stores.
10762 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10764 * common.opt (foffload): Remove help as Driver only.
10765 * gcc.c (display_help): Add -foffload.
10767 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10769 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
10770 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
10771 a candidate list; better inform no offload target is configured
10772 and fix hint extraction when passed target is not '\0' at [len].
10773 * common.opt (foffload): Add tailing '.'.
10774 (foffload-options): Likewise; fix flag name in the help string.
10776 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10779 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
10780 (vcage_f32): Gate comparison on __FAST_MATH__.
10781 (vcageq_f32): Likewise.
10782 (vcale_f32): Likewise.
10783 (vcaleq_f32): Likewise.
10784 (vcagt_f32): Likewise.
10785 (vcagtq_f32): Likewise.
10786 (vcalt_f32): Likewise.
10787 (vcaltq_f32): Likewise.
10788 (vcage_f16): Likewise.
10789 (vcageq_f16): Likewise.
10790 (vcale_f16): Likewise.
10791 (vcaleq_f16): Likewise.
10792 (vcagt_f16): Likewise.
10793 (vcagtq_f16): Likewise.
10794 (vcalt_f16): Likewise.
10795 (vcaltq_f16): Likewise.
10797 2021-06-30 Richard Biener <rguenther@suse.de>
10799 PR tree-optimization/101264
10800 * tree-vect-slp.c (vect_optimize_slp): Propagate the
10801 computed perm_in to all "any" permute successors
10802 we cannot de-duplicate immediately.
10804 2021-06-30 liuhongt <hongtao.liu@intel.com>
10807 * config/i386/sse.md
10808 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
10810 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
10812 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
10813 * config/i386/subst.md (maskz_scalar): New define_subst.
10814 (maskz_scalar_name): New subst_attr.
10815 (maskz_scalar_op5): Ditto.
10816 (round_saeonly_maskz_scalar_op5): Ditto.
10817 (round_saeonly_maskz_scalar_operand5): Ditto.
10819 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
10821 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
10822 Increase code CSECT alignment to at least 32 bytes.
10823 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
10824 alignment designation.
10826 2021-06-29 Sergei Trofimovich <siarheit@google.com>
10828 * doc/generic.texi: Fix s/net yet/not yet/ typo.
10830 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10832 PR tree-optimization/101254
10833 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
10834 wrapping/non-wrapping when setting the result range.
10836 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10838 * value-query.cc (gimple_range_global): Allow phis.
10840 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10842 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
10843 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
10844 statement for location context.
10845 (check_for_binary_op_overflow): Ditto.
10846 (simplify_using_ranges::get_vr_for_comparison): Ditto.
10847 (simplify_using_ranges::compare_name_with_value): Ditto.
10848 (simplify_using_ranges::compare_names): Ditto.
10849 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
10850 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
10851 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
10852 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
10853 (simplify_using_ranges::two_valued_val_range_p): Ditto.
10854 (simplify_using_ranges::simplify): Ditto.
10855 * vr-values.h: Adjust prototypes.
10857 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
10860 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
10862 2021-06-29 Julian Brown <julian@codesourcery.com>
10864 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
10865 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
10866 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
10868 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
10869 (MAX_FIXED_MODE_SIZE): Change to 128.
10871 2021-06-29 Julian Brown <julian@codesourcery.com>
10873 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
10874 (s_mnemonic): Add clrsb.
10875 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
10876 (clrsb<mode>2): Add expander for SImode/DImode.
10878 2021-06-29 Julian Brown <julian@codesourcery.com>
10880 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
10881 muldi3): Add patterns.
10883 2021-06-29 Julian Brown <julian@codesourcery.com>
10885 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
10886 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
10888 2021-06-29 Julian Brown <julian@codesourcery.com>
10890 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
10892 2021-06-29 Joseph Myers <joseph@codesourcery.com>
10894 * btfout.c, ctfout.c: Include "memmodel.h".
10896 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10898 * gcc.c (check_offload_target_name): Cast len argument to
10899 %q.*s to 'int'; avoid -Wstringop-truncation warning.
10901 2021-06-29 Richard Biener <rguenther@suse.de>
10903 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
10904 to "any" permute nodes and relax "any" permute proapgation
10905 during iterative backward propagation.
10907 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10910 * common.opt (-foffload=): Update description.
10911 (-foffload-options=): New.
10912 * doc/invoke.texi (C Language Options): Document
10913 -foffload and -foffload-options.
10914 * gcc.c (check_offload_target_name): New, split off from
10915 handle_foffload_option.
10916 (check_foffload_target_names): New.
10917 (handle_foffload_option): Handle -foffload=default.
10918 (driver_handle_option): Update for -foffload-options.
10919 * lto-opts.c (lto_write_options): Use -foffload-options
10920 instead of -foffload.
10921 * lto-wrapper.c (merge_and_complain, append_offload_options):
10923 * opts.c (common_handle_option): Likewise.
10925 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10927 * doc/invoke.texi (C Language Options): Sort options
10928 alphabetically in optlist and also the description itself.
10929 Remove leftover -fallow-single-precision from and add missing
10930 -fgnu-tm to the optlist.
10932 2021-06-29 Richard Biener <rguenther@suse.de>
10934 * tree-vect-slp.c (slpg_vertex::visited): Remove.
10935 (vect_slp_perms_eq): Handle -1 permutes.
10936 (vect_optimize_slp): Rewrite permute propagation.
10938 2021-06-29 Jakub Jelinek <jakub@redhat.com>
10941 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
10942 perform the optimization in GENERIC when sanitizing and x has a
10945 2021-06-29 Richard Biener <rguenther@suse.de>
10947 PR tree-optimization/101242
10948 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
10949 PHIs with not represented initial values as leafs.
10951 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
10953 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
10954 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
10955 "register" keyword.
10956 (pdp11_initial_elimination_offset) Remove unused variable.
10957 (pdp11_cmp_length) Ditto.
10958 (pdp11_insn_cost): Ditto, and fix signedness warning.
10960 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
10962 * btfout.c: Include tm_p.h.
10965 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10967 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
10969 (bpf_expand_epilogue): Likewise.
10970 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
10971 Do not define DBX_DEBUGGING_INFO.
10973 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10975 * doc/invoke.texi: Document the CTF and BTF debug info options.
10977 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10978 David Faust <david.faust@oracle.com>
10979 Jose E. Marchesi <jose.marchesi@oracle.com>
10980 Weimin Pan <weimin.pan@oracle.com>
10982 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
10983 GTFILES. Add new object files.
10984 * common.opt: Add CTF and BTF debug info options.
10985 * btfout.c: New file.
10986 * ctfc.c: Likewise.
10987 * ctfc.h: Likewise.
10988 * ctfout.c: Likewise.
10989 * dwarf2ctf.c: Likewise.
10990 * dwarf2ctf.h: Likewise.
10991 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
10993 * dwarf2out.c (dwarf2out_source_line): Likewise.
10994 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
10996 (debug_format_do_cu): New function.
10997 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
10999 Include dwarf2ctf.c.
11000 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
11002 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
11003 (CTF_DEBUG): New bitmask.
11004 (BTF_DEBUG): Likewise.
11005 (enum ctf_debug_info_levels): New enum.
11006 * gengtype.c (open_base_files): Handle ctfc.h.
11007 (main): Handle uint32_t type.
11008 * flags.h (btf_debuginfo_p): New definition.
11009 (dwarf_based_debuginfo_p): Likewise.
11010 * opts.c (debug_type_names): Add entries for CTF and BTF.
11011 (btf_debuginfo_p): New function.
11012 (dwarf_based_debuginfo_p): Likewise.
11013 (common_handle_option): Handle -gctfN and -gbtf options.
11014 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
11015 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
11018 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
11020 * dwarf2out.c (AT_class): Function is no longer static.
11021 (AT_int): Likewise.
11022 (AT_unsigned): Likewise.
11023 (AT_loc): Likewise.
11024 (get_AT): Likewise.
11025 (get_AT_string): Likewise.
11026 (get_AT_flag): Likewise.
11027 (get_AT_unsigned): Likewise.
11028 (get_AT_ref): Likewise.
11029 (new_die_raw): Likewise.
11030 (lookup_decl_die): Likewise.
11031 (base_type_die): Likewise.
11032 (add_name_attribute): Likewise.
11033 (add_AT_int): Likewise.
11034 (add_AT_unsigned): Likewise.
11035 (add_AT_loc): Likewise.
11036 (dw_get_die_tag): New function.
11037 (dw_get_die_child): Likewise.
11038 (dw_get_die_sib): Likewise.
11039 (struct dwarf_file_data): Move from here to dwarf2out.h
11040 (struct dw_attr_struct): Likewise.
11041 * dwarf2out.h: Analogous changes.
11043 2021-06-28 Martin Jambor <mjambor@suse.cz>
11046 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
11047 members m_dead_stmts and m_dead_ssas.
11048 * ipa-param-manipulation.c
11049 (ipa_param_body_adjustments::mark_dead_statements): New function.
11050 (ipa_param_body_adjustments::common_initialization): Call it on
11051 all removed but not split parameters.
11052 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
11054 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
11056 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
11057 dead debug statements.
11058 (copy_phis_for_bb): Do not copy dead PHI nodes.
11060 2021-06-28 Martin Jambor <mjambor@suse.cz>
11063 * symtab-clones.h (clone_info): Removed member param_adjustments.
11064 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
11065 deal with pass-through splits now.
11066 (ipa_param_performed_split): Removed.
11067 (ipa_param_adjustments::modify_call): Adjusted parameters.
11068 (class ipa_param_body_adjustments): Adjusted parameters of
11069 register_replacement, modify_gimple_stmt and modify_call_stmt.
11070 (ipa_verify_edge_has_no_modifications): Declare.
11071 (ipa_edge_modifications_finalize): Declare.
11072 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
11073 performed_splits processing, pas only edge to padjs->modify_call,
11074 check that call arguments were not modified if they should not have
11076 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
11078 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
11079 (ipa_edge_modification_info): Likewise.
11080 (ipa_edge_modification_sum): Likewise.
11081 (ipa_edge_modifications): New edge summary.
11082 (ipa_verify_edge_has_no_modifications): New function.
11083 (transitive_split_p): Removed.
11084 (transitive_split_map): Likewise.
11085 (init_transitive_splits): Likewise.
11086 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
11087 summary instead of performed_splits.
11088 (ipa_param_body_adjustments::register_replacement): Drop dummy
11089 parameter, set base_index of the created ipa_param_body_replacement.
11090 (phi_arg_will_live_p): New function.
11091 (ipa_param_body_adjustments::common_initialization): Do not create
11092 IPA_SRA dummy decls.
11093 (simple_tree_swap_info): Removed.
11094 (remap_split_decl_to_dummy): Likewise.
11095 (record_argument_state_1): New function.
11096 (record_argument_state): Likewise.
11097 (ipa_param_body_adjustments::modify_call_stmt): New parameter
11098 orig_stmt. Do not work with dummy decls, save necessary info about
11099 changes to ipa_edge_modifications.
11100 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
11101 orig_stmt, pass it to modify_call_stmt.
11102 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
11103 modify_gimple_stmt.
11104 (ipa_edge_modifications_finalize): New function.
11105 * tree-inline.c (remap_gimple_stmt): Pass original statement to
11106 modify_gimple_stmt.
11107 (copy_phis_for_bb): Do not copy dead PHI nodes.
11108 (expand_call_inline): Do not remap performed_splits.
11109 (update_clone_info): Likewise.
11110 * toplev.c: Include ipa-param-manipulation.h.
11111 (toplev::finalize): Call ipa_edge_modifications_finalize.
11113 2021-06-28 Andrew Pinski <apinski@marvell.com>
11115 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
11116 info if we're the only things setting the target PHI.
11117 (value_replacement): Don't duplicate range here.
11118 (minmax_replacement): Likewise.
11120 2021-06-28 Richard Biener <rguenther@suse.de>
11122 PR tree-optimization/101229
11123 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11125 2021-06-28 Martin Liska <mliska@suse.cz>
11127 * config/v850/v850.c (construct_dispose_instruction): Allocate
11129 (construct_prepare_instruction): Likewise.
11131 2021-06-28 Martin Liska <mliska@suse.cz>
11133 * config/v850/v850.c (v850_option_override): Build default
11135 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11136 ignored for inlining.
11137 (TARGET_CAN_INLINE_P): New.
11139 2021-06-28 Richard Biener <rguenther@suse.de>
11141 PR tree-optimization/101207
11142 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11143 permute eliding for load permutations properly.
11145 2021-06-28 Richard Biener <rguenther@suse.de>
11147 PR tree-optimization/101173
11148 * gimple-loop-interchange.cc
11149 (tree_loop_interchange::valid_data_dependences): Disallow outer
11150 loop dependence distance of zero.
11152 2021-06-28 liuhongt <hongtao.liu@intel.com>
11155 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11156 define_insn_and_split.
11157 (*avx_cmp<mode>3_ltint): Ditto.
11158 (*avx2_pcmp<mode>3_3): Ditto.
11159 (*avx2_pcmp<mode>3_4): Ditto.
11160 (*avx2_pcmp<mode>3_5): Ditto.
11162 2021-06-28 liuhongt <hongtao.liu@intel.com>
11164 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11165 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11166 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11167 IX86_BUILTIN_PBLENDVB128): Replace icode with
11169 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11171 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11172 New pre_reload splitter.
11174 2021-06-27 Andrew Pinski <apinski@marvell.com>
11176 PR middle-end/101230
11177 * fold-const.c (fold_ternary_loc): Check
11178 the return value of invert_tree_comparison.
11180 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11182 * config.gcc: Add SPDX License Identifier.
11183 (powerpc-ibm-aix789): Default to aix73.h.
11184 (powerpc-ibm-aix7.2.*.*): New stanza.
11185 * config/rs6000/aix72.h: Add SPDX License Identifier.
11186 * config/rs6000/aix73.h: New file.
11188 2021-06-26 Jason Merrill <jason@redhat.com>
11190 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11192 2021-06-26 Andrew Pinski <apinski@marvell.com>
11194 * genmatch.c (lower_cond): Copy for_subst_vec
11195 for the simplify also.
11196 (lower): Swap the order for lower_for and lower_cond.
11198 2021-06-26 Andrew Pinski <apinski@marvell.com>
11200 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11201 flow senatitive info on the moved ssa set.
11203 2021-06-26 Andrew Pinski <apinski@marvell.com>
11205 * fold-const.c (fold_cond_expr_with_comparison):
11206 Exand arg0 into comp_code, arg00, and arg01.
11207 (fold_ternary_loc): Use invert_tree_comparison
11208 instead of fold_invert_truthvalue for the case
11209 where we have A CMP B ? C : A.
11211 2021-06-25 Martin Sebor <msebor@redhat.com>
11213 PR middle-end/101216
11214 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11216 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11218 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11220 2021-06-25 Richard Biener <rguenther@suse.de>
11222 PR tree-optimization/101202
11223 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11226 2021-06-25 Richard Biener <rguenther@suse.de>
11228 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11229 STMT_VINFO_REDUC_DEF from the original representative.
11231 2021-06-25 Martin Sebor <msebor@redhat.com>
11233 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11234 gimple_no_warning_p and gimple_set_no_warning with
11235 warning_suppressed_p, and suppress_warning.
11237 (maybe_warn_for_bound): Same.
11238 (warn_for_access): Same.
11239 (check_access): Same.
11240 (expand_builtin_strncmp): Same.
11241 (fold_builtin_varargs): Same.
11242 * calls.c (maybe_warn_nonstring_arg): Same.
11243 (maybe_warn_rdwr_sizes): Same.
11244 * cfgexpand.c (expand_call_stmt): Same.
11245 * cgraphunit.c (check_global_declaration): Same.
11246 * fold-const.c (fold_undefer_overflow_warnings): Same.
11247 (fold_truth_not_expr): Same.
11248 (fold_unary_loc): Same.
11249 (fold_checksum_tree): Same.
11250 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
11251 (array_bounds_checker::check_mem_ref): Same.
11252 (array_bounds_checker::check_addr_expr): Same.
11253 (array_bounds_checker::check_array_bounds): Same.
11254 * gimple-expr.c (copy_var_decl): Same.
11255 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
11256 (gimple_fold_builtin_strncat): Same.
11257 (gimple_fold_builtin_stxcpy_chk): Same.
11258 (gimple_fold_builtin_stpcpy): Same.
11259 (gimple_fold_builtin_sprintf): Same.
11260 (fold_stmt_1): Same.
11261 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
11262 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
11263 * gimple-ssa-sprintf.c (handle_printf_call): Same.
11264 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
11265 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11266 * gimple-ssa-warn-restrict.h: Adjust declarations.
11267 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11268 gimple_no_warning_p and gimple_set_no_warning with
11269 warning_suppressed_p, and suppress_warning.
11270 (check_call): Same.
11271 (check_bounds_or_overlap): Same.
11272 * gimple.c (gimple_build_call_from_tree): Same.
11273 * gimplify.c (gimplify_return_expr): Same.
11274 (gimplify_cond_expr): Same.
11275 (gimplify_modify_expr_complex_part): Same.
11276 (gimplify_modify_expr): Same.
11277 (gimple_push_cleanup): Same.
11278 (gimplify_expr): Same.
11279 * omp-expand.c (expand_omp_for_generic): Same.
11280 (expand_omp_taskloop_for_outer): Same.
11281 * omp-low.c (lower_rec_input_clauses): Same.
11282 (lower_lastprivate_clauses): Same.
11283 (lower_send_clauses): Same.
11284 (lower_omp_target): Same.
11285 * tree-cfg.c (pass_warn_function_return::execute): Same.
11286 * tree-complex.c (create_one_component_var): Same.
11287 * tree-inline.c (remap_gimple_op_r): Same.
11288 (copy_tree_body_r): Same.
11289 (declare_return_variable): Same.
11290 (expand_call_inline): Same.
11291 * tree-nested.c (lookup_field_for_decl): Same.
11292 * tree-sra.c (create_access_replacement): Same.
11293 (generate_subtree_copies): Same.
11294 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11295 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11296 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11297 * tree-ssa-loop-im.c (execute_sm): Same.
11298 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11299 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11300 (handle_builtin_strcpy): Same.
11301 (maybe_diag_stxncpy_trunc): Same.
11302 (handle_builtin_stxncpy_strncat): Same.
11303 (handle_builtin_strcat): Same.
11304 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11305 (set_no_uninit_warning): Same.
11306 (uninit_undefined_value_p): Same.
11307 (warn_uninit): Same.
11308 (maybe_warn_operand): Same.
11309 * tree-vrp.c (compare_values_warnv): Same.
11310 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11311 (test_for_singularity): Same.
11312 * gimple.h (warning_suppressed_p): New function.
11313 (suppress_warning): Same.
11314 (copy_no_warning): Same.
11315 (gimple_set_block): Call gimple_set_location.
11316 (gimple_set_location): Call copy_warning.
11318 2021-06-25 Martin Sebor <msebor@redhat.com>
11320 * tree.h (warning_suppressed_at, copy_warning,
11321 warning_suppressed_p, suppress_warning): New functions.
11323 2021-06-25 Martin Sebor <msebor@redhat.com>
11325 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11326 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11327 * diagnostic-spec.c: New file.
11328 * diagnostic-spec.h: New file.
11329 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11331 * warning-control.cc: New file.
11333 2021-06-25 liuhongt <hongtao.liu@intel.com>
11336 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11339 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11341 PR tree-optimization/101189
11342 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11343 LHS range of condition to postfold routine.
11344 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11345 FALSE edge if the LHS range supports it being taken.
11346 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11348 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11350 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11351 (relation_oracle::find_relation_block): Check correct bitmap.
11352 (relation_oracle::dump): Do not dump NULL blocks.
11354 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11356 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11357 range_on_edge instead of manually calculating.
11359 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11361 * range-op.cc: Fix comment.
11363 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11366 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11367 Handle V8QI and V4HI modes.
11368 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11370 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11371 (mmxpackmode): New mode attribute.
11372 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11373 (mmxunpackmode): New mode attribute.
11374 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11375 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11376 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11377 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11378 * config/i386/i386.md (extsuffix): Move from ...
11379 * config/i386/sse.md: ... here.
11381 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11383 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11384 (dwarf2out_finish): ...instead of here.
11386 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11388 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11389 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11390 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11391 * configure: Regenerate.
11393 2021-06-24 Richard Biener <rguenther@suse.de>
11395 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11396 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11397 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11399 2021-06-24 Richard Biener <rguenther@suse.de>
11401 * config/i386/sse.md (avx_addsubv4df3): Rename to
11403 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11404 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11405 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11406 * config/i386/i386-builtin.def: Adjust.
11407 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11408 * optabs.def (vec_addsub_optab): New optab.
11409 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11410 (slp_patterns): Add addsub_pattern.
11411 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11412 across CFN_VEC_ADDSUB.
11413 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11415 * doc/md.texi (vec_addsub<mode>3): Document.
11417 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11419 PR middle-end/101170
11420 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11421 where regno + subreg_regno_offset wraps around use 0 as starting
11424 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11426 PR middle-end/101172
11427 * stor-layout.c (finish_bitfield_representative): If nextf has
11428 error_mark_node type, set repr type to error_mark_node too.
11430 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11432 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11434 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11436 2021-06-24 Richard Biener <rguenther@suse.de>
11438 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11439 across operations that have different semantics on different
11442 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11444 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11445 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11446 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11447 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11448 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11449 initializer/combiner gimplification.
11450 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11451 on target constructs.
11452 (lower_rec_input_clauses): Likewise.
11453 (lower_omp_target): Likewise.
11454 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11455 on target if in_reduction is present.
11457 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11459 * tree-predcom.c (class pcom_worker): New class.
11460 (release_chain): Renamed to...
11461 (pcom_worker::release_chain): ...this.
11462 (release_chains): Renamed to...
11463 (pcom_worker::release_chains): ...this.
11464 (aff_combination_dr_offset): Renamed to...
11465 (pcom_worker::aff_combination_dr_offset): ...this.
11466 (determine_offset): Renamed to...
11467 (pcom_worker::determine_offset): ...this.
11468 (class comp_ptrs): New class.
11469 (split_data_refs_to_components): Renamed to...
11470 (pcom_worker::split_data_refs_to_components): ...this,
11471 and update with class comp_ptrs.
11472 (suitable_component_p): Renamed to...
11473 (pcom_worker::suitable_component_p): ...this.
11474 (filter_suitable_components): Renamed to...
11475 (pcom_worker::filter_suitable_components): ...this.
11476 (valid_initializer_p): Renamed to...
11477 (pcom_worker::valid_initializer_p): ...this.
11478 (find_looparound_phi): Renamed to...
11479 (pcom_worker::find_looparound_phi): ...this.
11480 (add_looparound_copies): Renamed to...
11481 (pcom_worker::add_looparound_copies): ...this.
11482 (determine_roots_comp): Renamed to...
11483 (pcom_worker::determine_roots_comp): ...this.
11484 (determine_roots): Renamed to...
11485 (pcom_worker::determine_roots): ...this.
11486 (single_nonlooparound_use): Renamed to...
11487 (pcom_worker::single_nonlooparound_use): ...this.
11488 (remove_stmt): Renamed to...
11489 (pcom_worker::remove_stmt): ...this.
11490 (execute_pred_commoning_chain): Renamed to...
11491 (pcom_worker::execute_pred_commoning_chain): ...this.
11492 (execute_pred_commoning): Renamed to...
11493 (pcom_worker::execute_pred_commoning): ...this.
11494 (struct epcc_data): New member worker.
11495 (execute_pred_commoning_cbck): Call execute_pred_commoning
11496 with pcom_worker pointer.
11497 (find_use_stmt): Renamed to...
11498 (pcom_worker::find_use_stmt): ...this.
11499 (find_associative_operation_root): Renamed to...
11500 (pcom_worker::find_associative_operation_root): ...this.
11501 (find_common_use_stmt): Renamed to...
11502 (pcom_worker::find_common_use_stmt): ...this.
11503 (combinable_refs_p): Renamed to...
11504 (pcom_worker::combinable_refs_p): ...this.
11505 (reassociate_to_the_same_stmt): Renamed to...
11506 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11507 (stmt_combining_refs): Renamed to...
11508 (pcom_worker::stmt_combining_refs): ...this.
11509 (combine_chains): Renamed to...
11510 (pcom_worker::combine_chains): ...this.
11511 (try_combine_chains): Renamed to...
11512 (pcom_worker::try_combine_chains): ...this.
11513 (prepare_initializers_chain): Renamed to...
11514 (pcom_worker::prepare_initializers_chain): ...this.
11515 (prepare_initializers): Renamed to...
11516 (pcom_worker::prepare_initializers): ...this.
11517 (prepare_finalizers_chain): Renamed to...
11518 (pcom_worker::prepare_finalizers_chain): ...this.
11519 (prepare_finalizers): Renamed to...
11520 (pcom_worker::prepare_finalizers): ...this.
11521 (tree_predictive_commoning_loop): Renamed to...
11522 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11523 some calls and remove some cleanup code.
11524 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11525 (static variable looparound_phis): Remove.
11526 (static variable name_expansions): Remove.
11528 2021-06-24 Richard Biener <rguenther@suse.de>
11530 * tree-vect-slp.c (slpg_vertex): New struct.
11531 (vect_slp_build_vertices): Adjust.
11532 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11533 and a materialized one.
11535 2021-06-24 Richard Biener <rguenther@suse.de>
11537 PR tree-optimization/101105
11538 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11539 Only ignore steps when they are equal or scalar order is preserved.
11541 2021-06-24 liuhongt <hongtao.liu@intel.com>
11544 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11545 Adjust comments for ix86_expand_vecop_qihi2.
11546 (ix86_expand_vecmul_qihi): Renamed to ..
11547 (ix86_expand_vecop_qihi2): Adjust function prototype to
11548 support shift operation, add static to definition.
11549 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11550 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11551 ix86_expand_vec_shift_qihi_constant.
11552 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11553 (ix86_expand_vec_shift_qihi_constant): Deleted.
11554 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11556 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11557 condition TARGET_64BIT.
11558 (mul<mode>3): Ditto.
11559 (<insn><mode>3): Ditto.
11560 (vlshr<mode>3): Extend to support avx512 vlshr.
11561 (v<insn><mode>3): New expander for
11563 (v<insn>v8qi3): Ditto.
11564 (vashrv8hi3<mask_name>): Renamed to ..
11565 (vashr<mode>3): And extend to support V16QImode for avx512.
11566 (vashrv16qi3): Deleted.
11567 (vashrv2di3<mask_name>): Extend expander to support avx512
11570 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11572 * doc/lto.texi (Design Overview): Update that slim objects are
11575 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11577 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11578 of OTHER_POWER10_MASKS so it will not be enabled by default.
11580 2021-06-23 Richard Biener <rguenther@suse.de>
11581 Martin Jambor <mjambor@suse.cz>
11583 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11584 param replacement unconditionally. Adjust comment.
11586 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11588 * Makefile.in (OBJS): Add gimple-range-fold.o
11589 * gimple-range-fold.cc: New.
11590 * gimple-range-fold.h: New.
11591 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11592 (gimple_range_calc_op2): Ditto.
11593 * gimple-range-gori.h: Move prototypes to here.
11594 * gimple-range.cc: Adjust include files.
11595 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11596 (fur_source::get_operand): Ditto.
11597 (fur_source::get_phi_operand): Ditto.
11598 (fur_source::query_relation): Ditto.
11599 (fur_source::register_relation): Ditto.
11600 (class fur_edge): Ditto.
11601 (fur_edge::fur_edge): Ditto.
11602 (fur_edge::get_operand): Ditto.
11603 (fur_edge::get_phi_operand): Ditto.
11604 (fur_stmt::fur_stmt): Ditto.
11605 (fur_stmt::get_operand): Ditto.
11606 (fur_stmt::get_phi_operand): Ditto.
11607 (fur_stmt::query_relation): Ditto.
11608 (class fur_depend): Relocate to gimple-range-fold.h.
11609 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
11610 (fur_depend::register_relation): Ditto.
11611 (fur_depend::register_relation): Ditto.
11612 (class fur_list): Ditto.
11613 (fur_list::fur_list): Ditto.
11614 (fur_list::get_operand): Ditto.
11615 (fur_list::get_phi_operand): Ditto.
11616 (fold_range): Ditto.
11617 (adjust_pointer_diff_expr): Ditto.
11618 (gimple_range_adjustment): Ditto.
11619 (gimple_range_base_of_assignment): Ditto.
11620 (gimple_range_operand1): Ditto.
11621 (gimple_range_operand2): Ditto.
11622 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
11623 (gimple_range_calc_op2): Ditto.
11624 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
11625 (fold_using_range::range_of_range_op): Ditto.
11626 (fold_using_range::range_of_address): Ditto.
11627 (fold_using_range::range_of_phi): Ditto.
11628 (fold_using_range::range_of_call): Ditto.
11629 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11630 (fold_using_range::range_of_builtin_call): Ditto.
11631 (fold_using_range::range_of_cond_expr): Ditto.
11632 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
11633 (fold_using_range::relation_fold_and_or): Ditto.
11634 (fold_using_range::postfold_gcond_edges): Ditto.
11635 * gimple-range.h: Add gimple-range-fold.h to include files. Change
11636 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
11637 (gimple_range_handler): Relocate to gimple-range-fold.h.
11638 (gimple_range_ssa_p): Ditto.
11639 (range_compatible_p): Ditto.
11640 (class fur_source): Ditto.
11641 (class fur_stmt): Ditto.
11642 (class fold_using_range): Ditto.
11643 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
11644 (gimple_range_calc_op2): Ditto.
11646 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11648 PR tree-optimization/101148
11649 PR tree-optimization/101014
11650 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
11651 (ranger_cache::~ranger_cache): Adjust.
11652 (ranger_cache::block_range): Check if propagation disallowed.
11653 (ranger_cache::propagate_cache): Disallow propagation if new value
11654 can't be stored properly.
11655 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
11657 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11659 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
11660 (sbr_vector::set_bb_range): Return true.
11661 (class sbr_sparse_bitmap): Adjust.
11662 (sbr_sparse_bitmap::set_bb_range): Return value.
11663 (block_range_cache::set_bb_range): Return value.
11664 (ranger_cache::propagate_cache): Use return value to print msg.
11665 * gimple-range-cache.h (class block_range_cache): Adjust.
11667 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11669 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
11671 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
11673 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
11674 so this pattern can be used for test/compare removal. Pass
11675 current insn to compute_logical_op_length and output_logical_op.
11676 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
11677 (h8300_and_costs): Add argument to compute_logical_op_length.
11678 (output_logical_op): Add new argument. Use it to determine if the
11679 condition codes are used and adjust the output accordingly.
11680 (compute_logical_op_length): Add new argument and update length
11681 computations when condition codes are used.
11682 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
11684 (output_logical_op): Likewise.
11686 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11689 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
11690 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
11691 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
11692 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
11693 * config/i386/sse.md (unspec): ... here.
11695 2021-06-23 Martin Liska <mliska@suse.cz>
11698 * optc-save-gen.awk: Put back arm_fp16_format to
11701 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11704 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
11707 (clz<mode>2): Update RTX pattern for additions.
11709 2021-06-23 Jakub Jelinek <jakub@redhat.com>
11711 PR middle-end/101167
11712 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
11713 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
11715 2021-06-22 Sergei Trofimovich <siarheit@google.com>
11717 * doc/rtl.texi: drop unbalanced parenthesis.
11719 2021-06-22 Richard Biener <rguenther@suse.de>
11721 PR middle-end/101156
11722 * gimplify.c (gimplify_expr): Remove premature incorrect
11725 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11727 PR tree-optimization/101159
11728 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
11731 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11733 PR middle-end/101160
11734 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
11735 clear crtl->return_rtx instead of keeping it referencing a pseudo.
11737 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11738 Andrew Pinski <apinski@marvell.com>
11740 PR tree-optimization/101162
11741 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
11744 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11746 * range-op.cc (range_relational_tests): New.
11747 (range_op_tests): Call range_relational_tests.
11749 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11751 * range-op.cc (operator_cast::lhs_op1_relation): New.
11752 (operator_identity::lhs_op1_relation): Mew.
11754 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11756 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
11758 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11760 * range-op.cc (operator_plus::lhs_op1_relation): New.
11761 (operator_plus::lhs_op2_relation): New.
11763 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11765 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
11766 relation_oracle if dominators exist.
11767 (ranger_cache::~ranger_cache): Dispose of oracle.
11768 (ranger_cache::dump_bb): Dump oracle.
11769 * gimple-range.cc (fur_source::fur_source): New.
11770 (fur_source::get_operand): Use mmeber query.
11771 (fur_source::get_phi_operand): Use member_query.
11772 (fur_source::query_relation): New.
11773 (fur_source::register_dependency): Delete.
11774 (fur_source::register_relation): New.
11775 (fur_edge::fur_edge): Adjust.
11776 (fur_edge::get_phi_operand): Fix comment.
11777 (fur_edge::query): Delete.
11778 (fur_stmt::fur_stmt): Adjust.
11779 (fur_stmt::query): Delete.
11780 (fur_depend::fur_depend): Adjust.
11781 (fur_depend::register_relation): New.
11782 (fur_depend::register_relation): New.
11783 (fur_list::fur_list): Adjust.
11784 (fur_list::get_operand): Use member query.
11785 (fold_using_range::range_of_range_op): Process and query relations.
11786 (fold_using_range::range_of_address): Adjust dependency call.
11787 (fold_using_range::range_of_phi): Ditto.
11788 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
11789 (fold_using_range::relation_fold_and_or): New.
11790 (fold_using_range::postfold_gcond_edges): New.
11791 * gimple-range.h (class gimple_ranger): Adjust.
11792 (class fur_source): Adjust members.
11793 (class fur_stmt): Ditto.
11794 (class fold_using_range): Ditto.
11796 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11798 * range-op.cc (range_operator::wi_fold): Apply relation effect.
11799 (range_operator::fold_range): Adjust and apply relation effect.
11800 (*::fold_range): Add relation parameters.
11801 (*::op1_range): Ditto.
11802 (*::op2_range): Ditto.
11803 (range_operator::lhs_op1_relation): New.
11804 (range_operator::lhs_op2_relation): New.
11805 (range_operator::op1_op2_relation): New.
11806 (range_operator::op1_op2_relation_effect): New.
11807 (relop_early_resolve): New.
11808 (operator_equal::op1_op2_relation): New.
11809 (operator_equal::fold_range): Call relop_early_resolve.
11810 (operator_not_equal::op1_op2_relation): New.
11811 (operator_not_equal::fold_range): Call relop_early_resolve.
11812 (operator_lt::op1_op2_relation): New.
11813 (operator_lt::fold_range): Call relop_early_resolve.
11814 (operator_le::op1_op2_relation): New.
11815 (operator_le::fold_range): Call relop_early_resolve.
11816 (operator_gt::op1_op2_relation): New.
11817 (operator_gt::fold_range): Call relop_early_resolve.
11818 (operator_ge::op1_op2_relation): New.
11819 (operator_ge::fold_range): Call relop_early_resolve.
11820 * range-op.h (class range_operator): Adjust parameters and methods.
11822 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11824 * Makefile.in (OBJS): Add value-relation.o.
11825 * gimple-range.h: Adjust include files.
11826 * tree-data-ref.c: Adjust include file order.
11827 * value-query.cc (range_query::get_value_range): Default to no oracle.
11828 (range_query::query_relation): New.
11829 (range_query::query_relation): New.
11830 * value-query.h (class range_query): Adjust.
11831 * value-relation.cc: New.
11832 * value-relation.h: New.
11834 2021-06-22 Richard Biener <rguenther@suse.de>
11836 PR tree-optimization/101151
11837 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
11840 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
11842 * config/riscv/riscv.c (thead_c906_tune_info): New.
11843 (riscv_tune_info_table): Use new tune.
11845 2021-06-22 Richard Biener <rguenther@suse.de>
11847 PR tree-optimization/101158
11848 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
11849 checking after checking for matching operation.
11851 2021-06-22 Richard Biener <rguenther@suse.de>
11853 PR tree-optimization/101159
11854 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
11855 missing NULL vectype check.
11857 2021-06-22 Richard Biener <rguenther@suse.de>
11859 PR tree-optimization/101154
11860 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
11862 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11865 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
11866 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
11867 * config/i386/i386.c (ix86_expand_prologue): Clear it.
11868 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
11869 Remove "" from match_operand. Emit new insns using emit_move_insn and
11870 set ix86_last_zero_store_uid to INSN_UID of the last store.
11871 Add peephole2s for 1/2/4 stores of const0_rtx following previous
11874 2021-06-22 Martin Liska <mliska@suse.cz>
11876 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
11879 2021-06-22 Martin Liska <mliska@suse.cz>
11881 * gcov-io.h: Remove padding entries.
11883 2021-06-22 liuhongt <hongtao.liu@intel.com>
11885 PR tree-optimization/97770
11886 * tree-vect-patterns.c (vect_recog_popcount_pattern):
11888 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
11890 2021-06-22 liuhongt <hongtao.liu@intel.com>
11893 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
11894 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
11895 (expand<mode>_mask): this ..
11896 (*expand<mode>_mask): New pre_reload splitter to transform
11897 v{,p}expand* to vmov* when mask is zero, all ones, or has all
11898 ones in it's lower part, otherwise still generate
11901 2021-06-22 liuhongt <hongtao.liu@intel.com>
11904 * config/i386/i386-expand.c
11905 (ix86_expand_special_args_builtin): Keep constm1_operand only
11906 if it satisfies insn's operand predicate.
11908 2021-06-21 Jason Merrill <jason@redhat.com>
11911 * df-scan.c (df_ref_record): Check that regno < endregno.
11912 * function.c (assign_parms, expand_function_end): Do nothing with a
11913 TYPE_EMPTY_P result.
11915 2021-06-21 Richard Biener <rguenther@suse.de>
11917 PR tree-optimization/101120
11918 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
11920 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
11921 DR chain DCE capability.
11922 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
11923 * tree-vect-stmts.c (vectorizable_load): Remove unused
11924 loads in the DR chain for SLP.
11926 2021-06-21 Jakub Jelinek <jakub@redhat.com>
11928 PR inline-asm/100785
11929 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
11930 output or input operands were already error_mark_node.
11931 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
11932 remove all inputs, outputs and clobbers from the asm and
11933 set template to "".
11935 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11937 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
11938 (vceq_s16): Likewise.
11939 (vceq_s32): Likewise.
11940 (vceq_u8): Likewise.
11941 (vceq_u16): Likewise.
11942 (vceq_u32): Likewise.
11943 (vceq_p8): Likewise.
11944 (vceqq_s8): Likewise.
11945 (vceqq_s16): Likewise.
11946 (vceqq_s32): Likewise.
11947 (vceqq_u8): Likewise.
11948 (vceqq_u16): Likewise.
11949 (vceqq_u32): Likewise.
11950 (vceqq_p8): Likewise.
11951 (vceq_f32): Gate __a == __b on __FAST_MATH__.
11952 (vceqq_f32): Likewise.
11953 (vceq_f16): Likewise.
11954 (vceqq_f16): Likewise.
11956 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11959 * config/arm/iterators.md (NEON_VACMP): Remove.
11960 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
11962 (neon_vca<cmp_op><mode>_insn): Likewise.
11963 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
11966 2021-06-21 Richard Biener <rguenther@suse.de>
11968 PR tree-optimization/101121
11969 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
11970 when we just lack a stmt with the desired op when doing permutation.
11971 (vect_build_slp_tree): When caching a failed SLP build attempt
11972 assert that at least one lane is marked as not matching.
11974 2021-06-21 liuhongt <hongtao.liu@intel.com>
11977 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
11978 register alternative.
11979 (*and<mode>_1): Ditto.
11981 (*andn<mode>_1): Ditto.
11982 (*<code><mode>_1): Ditto.
11983 (*<code>qi_1): Ditto.
11984 (*one_cmpl<mode>2_1): Ditto.
11985 (*one_cmplsi2_1_zext): Ditto.
11986 (*one_cmplqi2_1): Ditto.
11987 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
11988 the order of mask registers to be before general registers.
11990 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
11993 * config/i386/i386.md: New define_peephole2s to shrink writing
11994 1, 2 or 4 consecutive zeros to memory when optimizing for size.
11996 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
11998 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
11999 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
12000 more efficient code when the source can be trivially simplified.
12002 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12004 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
12005 a range if global is not available.
12006 (ranger_cache::entry_range): Fallback to range_of_def.
12007 * gimple-range-cache.h (range_of_def): Adjust prototype.
12009 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12011 PR tree-optimization/101014
12012 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
12014 (ranger_cache::~ranger_cache): Ditto.
12015 (ranger_cache::enable_new_values): Delete.
12016 (ranger_cache::push_poor_value): Delete.
12017 (ranger_cache::range_of_def): Remove poor value processing.
12018 (ranger_cache::entry_range): Ditto.
12019 (ranger_cache::fill_block_cache): Ditto.
12020 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
12021 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
12022 * gimple-range.h (class gimple_ranger): Adjust.
12024 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12027 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
12028 derived from arm_canon_arch.
12029 (arm_canon_arch_option): Call it.
12030 (arm_canon_arch_multilib_option): New function.
12031 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
12032 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
12033 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
12034 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
12035 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
12036 * config/arm/arm.opt (mlibarch): New option.
12037 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
12038 of march on RHS with mlibarch.
12040 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
12042 * config.in: Regenerate.
12043 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
12045 * configure: Regenerate.
12046 * configure.ac: Fix for global_load assembler functions.
12048 2021-06-18 Richard Biener <rguenther@suse.de>
12050 PR tree-optimization/101112
12051 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
12052 to lookup a pattern stmt def.
12054 2021-06-18 Jakub Jelinek <jakub@redhat.com>
12056 PR middle-end/101062
12057 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
12058 representatives in QUAL_UNION_TYPE.
12060 2021-06-18 Andrew Pinski <apinski@marvell.com>
12062 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
12063 Add counting of how many times it is done.
12064 (factor_out_conditional_conversion): Likewise.
12065 (match_simplify_replacement): Likewise.
12066 (value_replacement): Likewise.
12067 (spaceship_replacement): Likewise.
12068 (cond_store_replacement): Likewise.
12069 (cond_if_else_store_replacement_1): Likewise.
12070 (hoist_adjacent_loads): Likewise.
12072 2021-06-18 Andrew Pinski <apinski@marvell.com>
12074 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
12075 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
12076 (verify_gimple_assign_binary): Reject point and offset types on
12077 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
12078 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
12079 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
12081 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
12083 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
12084 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
12085 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
12088 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
12090 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
12091 earlyclobber to alts 0/1.
12092 (gen_addadd): Add earlyclobber to alts 0/1.
12093 * config/rs6000/fusion.md: Regenerate file.
12095 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12097 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
12099 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12101 * gimple-range-cache.cc: Comment cleanups.
12102 * gimple-range-gori.cc: Comment cleanups.
12103 * gimple-range.cc: Comment/spacing cleanups
12104 * value-range.h: Comment cleanups.
12106 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
12109 * calls.c (expand_call): Replace PUSH_ARGS with
12110 targetm.calls.push_argument (0).
12111 (emit_library_call_value_1): Likewise.
12112 * defaults.h (PUSH_ARGS): Removed.
12113 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
12114 targetm.calls.push_argument (0).
12115 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
12116 (emit_push_insn): Pass the number bytes to push to
12117 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12118 * hooks.c (hook_bool_uint_true): New.
12119 * hooks.h (hook_bool_uint_true): Likewise.
12120 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12121 targetm.calls.push_argument (0).
12122 * target.def (push_argument): Add a targetm.calls hook.
12123 * targhooks.c (default_push_argument): New.
12124 * targhooks.h (default_push_argument): Likewise.
12125 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12126 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12127 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12128 * config/i386/i386.c (ix86_push_argument): New.
12129 (TARGET_PUSH_ARGUMENT): Likewise.
12130 * config/i386/i386.h (PUSH_ARGS): Removed.
12131 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12132 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12133 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12134 * config/pru/pru.h (PUSH_ARGS): Likewise.
12135 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12136 TARGET_PUSH_ARGUMENT hook.
12137 * doc/tm.texi: Regenerated.
12139 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12142 * config/i386/i386-expand.c (expand_vector_set_var):
12143 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12144 ix86_expand_vector_init_duplicate.
12145 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12146 QImode for !TARGET_PARTIAL_REG_STALL.
12147 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12148 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12149 as operand 2 predicate. Call ix86_expand_vector_set_var
12150 for non-constant index operand.
12151 (vec_setv2si): Ditto.
12152 (vec_setv4hi): Ditto.
12153 (vec_setv8qi): ditto.
12155 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12157 PR tree-optimization/100790
12158 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12161 2021-06-17 Martin Liska <mliska@suse.cz>
12163 * doc/invoke.texi: Use consistently -O1 instead of -O.
12165 2021-06-17 Martin Liska <mliska@suse.cz>
12167 * gcov-io.h: Update documentation entry about string format.
12169 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12172 * config/s390/vecintrin.h (vec_doublee): Fix to use
12173 __builtin_s390_vflls.
12174 (vec_floate): Fix to use __builtin_s390_vflrd.
12176 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12178 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12179 * dominance.h (get_dominated_to_depth): Likewise.
12180 (get_all_dominated_blocks): Likewise.
12181 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12182 * gcse.c (hoist_code): Likewise.
12183 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12184 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12185 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12186 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12188 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12190 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12191 * dominance.h (get_dominated_by_region): Likewise.
12192 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12193 (gimple_duplicate_sese_tail): Likewise.
12194 (move_sese_region_to_fn): Likewise.
12196 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12198 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12199 * dominance.h (get_dominated_by): Likewise.
12200 * auto-profile.c (afdo_find_equiv_class): Adjust.
12201 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12202 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12203 * tree-cfg.c (test_linear_chain): Likewise.
12204 (test_diamond): Likewise.
12206 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12208 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12209 * cfgloopanal.c (get_loop_hot_path): Likewise.
12210 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12212 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12214 * cgraph.c (cgraph_node::collect_callers): Return
12215 auto_vec<cgraph_edge *>.
12216 * cgraph.h (cgraph_node::collect_callers): Likewise.
12217 * ipa-cp.c (create_specialized_node): Adjust.
12218 (decide_about_value): Likewise.
12219 (decide_whether_version_node): Likewise.
12220 * ipa-sra.c (process_isra_node_results): Likewise.
12222 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12224 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12225 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12227 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12230 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12232 * gimple-range.cc (debug_seed_ranger): New.
12233 (dump_ranger): New.
12234 (debug_ranger): New.
12236 2021-06-17 Richard Biener <rguenther@suse.de>
12238 PR tree-optimization/54400
12239 * tree-vectorizer.h (enum slp_instance_kind): Add
12240 slp_inst_kind_bb_reduc.
12241 (reduction_fn_for_scalar_code): Declare.
12242 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
12243 Check SLP_INSTANCE_KIND instead of looking at the
12245 (vect_slp_analyze_instance_alignment): Likewise.
12246 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
12247 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
12248 chain linearization from vect_build_slp_tree_2 and generalize
12249 for the use of BB reduction vectorization.
12250 (vect_build_slp_tree_2): Adjust accordingly.
12251 (vect_optimize_slp): Elide permutes at the root of BB reduction
12253 (vectorizable_bb_reduc_epilogue): New function.
12254 (vect_slp_prune_covered_roots): Likewise.
12255 (vect_slp_analyze_operations): Use them.
12256 (vect_slp_check_for_constructors): Recognize associatable
12257 chains for BB reduction vectorization.
12258 (vectorize_slp_instance_root_stmt): Generate code for the
12259 BB reduction epilogue.
12261 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12263 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
12265 (gori_compute::may_recompute_p): New.
12266 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12267 * gimple-range-gori.h (class gori_compute): Add prototype.
12269 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12271 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12272 true when a range can be calculated.
12273 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12275 2021-06-16 Martin Sebor <msebor@redhat.com>
12277 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12278 Correct documented defaults.
12280 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12282 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12283 m_new_value_p directly.
12285 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12288 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12289 Handle 64bit modes for TARGET_SSE4_1.
12290 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12291 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12292 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12293 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12294 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12295 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12297 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12299 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12300 Change to an expander that emits the correct instruction
12301 depending on endianness.
12302 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12303 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12305 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12307 * config/aarch64/aarch64-simd-builtins.def: Split generator
12308 for aarch64_<su>qmovn builtins into scalar and vector
12310 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12312 (aarch64_<su>qmovn<mode>_insn_be): Define.
12313 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12314 variants. Change vector variant to an expander that emits the
12315 correct instruction depending on endianness.
12317 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12319 * config/aarch64/aarch64-simd-builtins.def: Split generator
12320 for aarch64_sqmovun builtins into scalar and vector variants.
12321 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12322 Split into scalar and vector variants. Change vector variant
12323 to an expander that emits the correct instruction depending
12325 (aarch64_sqmovun<mode>_insn_le): Define.
12326 (aarch64_sqmovun<mode>_insn_be): Define.
12328 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12330 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12331 Define - modeling zero-high-half semantics.
12332 (aarch64_xtn<mode>): Change to an expander that emits the
12333 appropriate instruction depending on endianness.
12334 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12336 (aarch64_xtn2<mode>_le): Rename to...
12337 (aarch64_xtn2<mode>_insn_le): This.
12338 (aarch64_xtn2<mode>_be): Rename to...
12339 (aarch64_xtn2<mode>_insn_be): This.
12340 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12342 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12343 attribute iterator.
12345 2021-06-16 Martin Jambor <mjambor@suse.cz>
12347 PR tree-optimization/100453
12348 * tree-sra.c (create_access): Disqualify any const candidates
12349 which are written to.
12350 (sra_modify_expr): Do not store sub-replacements back to a const base.
12351 (handle_unscalarized_data_in_subtree): Likewise.
12352 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12353 instead of constant_decl_p.
12355 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12357 PR middle-end/101062
12358 * stor-layout.c (finish_bitfield_representative): For fields in unions
12359 assume nextf is always NULL.
12360 (finish_bitfield_layout): Compute bit field representatives also in
12361 unions, but handle it as if each bitfield was the only field in the
12364 2021-06-16 Richard Biener <rguenther@suse.de>
12366 PR tree-optimization/101088
12367 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12368 supported refs on edges. Do not assert same ref but
12369 different kind stores are unsuported but mark them so.
12370 (hoist_memory_references): Only look for supported refs
12373 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12375 PR rtl-optimization/46235
12376 * config/i386/i386.md: New define_split for bt followed by cmov.
12377 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12378 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12379 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12380 by setnc with zero extension.
12382 2021-06-16 Richard Biener <rguenther@suse.de>
12384 PR tree-optimization/101083
12385 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12386 vectype as argument.
12387 (vect_build_slp_tree_2): Adjust.
12389 2021-06-15 Martin Sebor <msebor@redhat.com>
12391 PR middle-end/100876
12392 * builtins.c: (gimple_call_return_array): Account for size_t
12393 mangling as either unsigned int or unsigned long
12395 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12397 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12398 up before eliminating comparisons.
12400 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12402 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12403 nonzero|X is nonzero.
12404 (range_op_bitwise_and_tests): Add tests for above.
12406 2021-06-15 Carl Love <cel@us.ibm.com>
12409 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12411 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12414 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12417 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12418 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12419 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12420 (omp_notice_variable): Update type classification for Fortran.
12421 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12422 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12423 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12424 * langhooks.c (lhd_omp_scalar_p): Likewise.
12425 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12426 (LANG_HOOKS_DECLS): Add them.
12427 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12428 omp_scalar_p pointer type to include the new bool argument.
12430 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12432 * doc/analyzer.texi
12433 (Special Functions for Debugging the Analyzer): Add
12434 __analyzer_dump_capacity.
12436 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12439 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12440 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12442 2021-06-15 Richard Biener <rguenther@suse.de>
12444 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12445 check to identify loop latches.
12446 * cfgloop.c (verify_loop_structure): Likewise.
12447 * loop-init.c (apply_loop_flags): Allow marked irreducible
12448 regions even with multiple latches.
12449 * predict.c (rebuild_frequencies): Simplify.
12451 2021-06-15 Richard Biener <rguenther@suse.de>
12453 * tree-ssa-threadupdate.c
12454 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12455 have marked irreducible regions.
12457 2021-06-14 Martin Sebor <msebor@redhat.com>
12460 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12461 Handle calls to placement new.
12462 (ndecl_dealloc_argno): Avoid placement delete.
12464 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12467 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12468 create_tmp_reg_or_ssa_name().
12470 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12472 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12473 (ranger_cache::enable_new_values): Set to specified value and
12474 return the old value.
12475 (ranger_cache::disable_new_values): Delete.
12476 (ranger_cache::fill_block_cache): Disable non 1st order derived
12478 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12479 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12481 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12484 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12485 Return true early when testing with V2HImode.
12486 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12488 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12490 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12491 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12492 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12493 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12494 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12496 (vec_unpack<US>_lo_<mode>): Likewise.
12497 (vec_pack_trunc_<mode>): Rename to
12498 neon_quad_vec_pack_trunc_<mode>.
12499 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12501 (vec_unpack<US>_lo_<mode>): New.
12502 (vec_pack_trunc_<mode>): New.
12504 2021-06-14 Richard Biener <rguenther@suse.de>
12506 PR tree-optimization/100934
12507 * tree-ssa-dom.c (pass_dominator::execute): Properly
12508 mark irreducible regions.
12510 2021-06-14 Martin Liska <mliska@suse.cz>
12512 * doc/invoke.texi: Put r{...} on the same line as @item.
12514 2021-06-14 Martin Liska <mliska@suse.cz>
12516 * doc/invoke.texi: Add missing newline.
12518 2021-06-14 Martin Liska <mliska@suse.cz>
12520 * doc/invoke.texi: Remove '+' charasters.
12522 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12524 * config.gcc (arc): Add support for with_cpu option.
12525 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12527 2021-06-14 Richard Biener <rguenther@suse.de>
12529 PR tree-optimization/101031
12530 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12531 instead of size when accounting for a possibly string
12534 2021-06-14 Martin Liska <mliska@suse.cz>
12536 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12538 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12540 * value-query.cc (gimple_range_global): Call get_range_global
12541 if called after inlining.
12543 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12546 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12547 Emit constant permutation insn directly from here.
12549 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12551 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12553 * auto-profile.c (afdo_find_equiv_class): Likewise.
12554 * gcc.c (do_specs_vec): Likewise.
12555 (do_spec_1): Likewise.
12556 (driver::set_up_specs): Likewise.
12557 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12558 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12559 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12560 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12561 (get_location_for_stmts): Likewise.
12562 * graphite-poly.c (print_iteration_domains): Likewise.
12563 (free_poly_bb): Likewise.
12564 (remove_gbbs_in_scop): Likewise.
12565 (free_scop): Likewise.
12566 (dump_gbb_cases): Likewise.
12567 (dump_gbb_conditions): Likewise.
12568 (print_pdrs): Likewise.
12569 (print_scop): Likewise.
12570 * ifcvt.c (cond_move_process_if_block): Likewise.
12571 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12572 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12573 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12574 * sel-sched-dump.c (dump_insn_vector): Likewise.
12575 * store-motion.c (store_ops_ok): Likewise.
12576 (store_killed_in_insn): Likewise.
12577 * timevar.c (timer::named_items::print): Likewise.
12578 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12579 (cleanup_tree_cfg_noloop): Likewise.
12580 * tree-data-ref.c (dump_data_references): Likewise.
12581 (print_dir_vectors): Likewise.
12582 (print_dist_vectors): Likewise.
12583 (dump_data_dependence_relations): Likewise.
12584 (dump_dist_dir_vectors): Likewise.
12585 (dump_ddrs): Likewise.
12586 (create_runtime_alias_checks): Likewise.
12587 (free_subscripts): Likewise.
12588 (save_dist_v): Likewise.
12589 (save_dir_v): Likewise.
12590 (invariant_access_functions): Likewise.
12591 (same_access_functions): Likewise.
12592 (access_functions_are_affine_or_constant_p): Likewise.
12593 (find_data_references_in_stmt): Likewise.
12594 (graphite_find_data_references_in_stmt): Likewise.
12595 (free_dependence_relations): Likewise.
12596 (free_data_refs): Likewise.
12597 * tree-inline.c (copy_debug_stmts): Likewise.
12598 * tree-into-ssa.c (dump_currdefs): Likewise.
12599 (rewrite_update_phi_arguments): Likewise.
12600 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12601 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12603 (vect_slp_analyze_node_dependences): Likewise.
12604 (vect_slp_analyze_instance_dependence): Likewise.
12605 (vect_record_base_alignments): Likewise.
12606 (vect_get_peeling_costs_all_drs): Likewise.
12607 (vect_peeling_supportable): Likewise.
12608 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
12609 (vec_info::free_stmt_vec_infos): Likewise.
12611 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
12613 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
12614 (andqi3_1<cczn>): Removed.
12615 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
12616 (H8/SX bit logicals): Split out from other patterns.
12617 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
12618 mulqihi3_const_clobber_flags.
12619 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
12621 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
12624 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
12625 to true if red zone is used.
12626 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
12627 ix86_red_zone_used.
12628 * config/i386/i386.h (machine_function): Add red_zone_used.
12629 (ix86_red_zone_size): Removed.
12630 (ix86_red_zone_used): New.
12631 * config/i386/i386.md (peephole2 patterns): Replace
12632 ix86_red_zone_size with ix86_red_zone_used.
12634 2021-06-12 Jason Merrill <jason@redhat.com>
12636 * doc/extend.texi (unused variable attribute): Applies to
12637 structure fields as well.
12639 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
12641 * auto-profile.c (read_profile): fix a typo in an error string
12643 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
12645 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
12647 * tree-pretty-print.c (dump_omp_clauses): Update.
12648 (dump_generic_node) <OMP_CLAUSE>: Use it.
12650 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12653 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
12654 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
12655 the polymorphic variants matching code.
12656 (__arm_vld1q_z): Likewise.
12657 (__arm_vld2q): Likewise.
12658 (__arm_vld4q): Likewise.
12659 (__arm_vldrbq_gather_offset): Likewise.
12660 (__arm_vldrbq_gather_offset_z): Likewise.
12662 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12664 PR tree-optimization/96392
12665 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
12667 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12669 PR tree-optimization/96392
12670 * fold-const.c (fold_real_zero_addition_p): Take both arguments
12671 of the addition or subtraction, not just the zero. Use this
12672 other argument in tests for signaling NaNs and signed zeros.
12673 (tree_expr_maybe_real_minus_zero_p): New predicate.
12674 * fold-const.h (fold_real_zero_addition_p): Update prototype.
12675 (tree_expr_maybe_real_minus_zero_p): New function prototype.
12676 * match.pd: Update calls to fold_real_zero_addition_p.
12677 Replace HONOR_NANS with tree_expr_maybe_nan_p.
12678 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
12679 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
12680 * tree-ssa-reassoc.c (eliminate_using_constants): Update
12681 call to fold_real_zero_addition_p.
12683 2021-06-11 Richard Biener <rguenther@suse.de>
12685 PR tree-optimization/101025
12686 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
12687 all refs that require dependence checking.
12689 2021-06-11 Richard Biener <rguenther@suse.de>
12691 PR tree-optimization/101028
12692 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
12693 reassoc discovery fails fatally, mark appropriate lanes
12696 2021-06-11 Richard Biener <rguenther@suse.de>
12698 PR tree-optimization/101026
12699 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
12700 have a representative for the associated chain nodes.
12702 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12704 PR rtl-optimization/101008
12705 * simplify-rtx.c (relational_result): New function.
12706 (simplify_logical_relational_operation,
12707 simplify_relational_operation): Use it.
12709 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12712 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
12714 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
12717 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
12718 false if the permutation can be implemented with constant
12719 permutation instruction in wider mode.
12720 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
12721 Handle V8QImode and V4HImode.
12723 2021-06-11 Martin Liska <mliska@suse.cz>
12725 PR gcov-profile/100788
12726 * common.opt: Add new option.
12727 * coverage.c (coverage_begin_function): Emit warning instead on
12728 the internal compiler error.
12729 * doc/invoke.texi: Document the option.
12730 * toplev.c (process_options): Enable it by default.
12732 2021-06-11 Richard Biener <rguenther@suse.de>
12734 PR middle-end/101009
12735 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
12736 to set *init_b to true when we encounter a constant equal
12738 (compute_affine_dependence): Also dump the actual DR_REF.
12740 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12742 PR tree-optimization/100984
12743 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
12744 replacements table.
12745 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
12747 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
12749 * config/rs6000/rs6000.md
12750 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
12751 define_insn_and_split.
12753 2021-06-11 Richard Biener <rguenther@suse.de>
12755 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
12756 to sort operands of the associative chain.
12758 2021-06-11 Richard Biener <rguenther@suse.de>
12760 * system.h (gcc_stablesort_r): Declare.
12761 * sort.cc (gcc_sort_r): Support stable sort.
12762 (gcc_stablesort_r): Define.
12763 * vec.h (vec<>::stablesort): Add.
12765 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
12768 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
12769 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
12770 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
12771 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
12772 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
12773 (expand_vec_perm_interleave2): Handle 64bit modes.
12774 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
12775 (expand_vec_perm_even_odd_1): Ditto.
12776 (ix86_vectorize_vec_perm_const): Ditto.
12777 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
12778 * config/i386/sse.md: ... here.
12779 * config/i386/mmx.md (*vec_interleave_lowv2sf):
12780 New insn_and_split pattern.
12781 (*vec_interleave_highv2sf): Ditto.
12782 (mmx_pshufbv8qi3): New insn pattern.
12783 (*mmx_pblendw): Ditto.
12785 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
12787 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
12788 (build_acc): Likewise.
12789 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
12790 source operands in little-endian mode.
12791 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
12792 (mma_init_builtins): Likewise.
12793 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
12794 ordering for the MMA assemble and build source operands.
12795 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
12797 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
12800 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
12802 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
12804 * config/h8300/extensions.md: Replace _clobber_flags patterns
12807 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
12809 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
12810 (vcond_mask_<mode><tointvec>): this.
12812 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
12813 Thomas Schwinge <thomas@codesourcery.com>
12815 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
12816 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
12817 * gimple.h (enum gf_mask): Split
12818 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
12819 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
12820 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
12821 (is_gimple_omp_oacc): Update.
12822 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
12823 * gimplify.c (gimplify_omp_target_update): Likewise.
12824 * omp-expand.c (expand_omp_target, build_omp_regions_1)
12825 (omp_make_gimple_edges): Likewise.
12826 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
12829 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12831 * value-query.cc (value_query::value_on_edge): Rename name to
12833 (range_query::range_on_edge): Same.
12834 (range_query::value_of_expr): Same.
12835 (range_query::value_on_edge): Same.
12836 * value-query.h (class value_query): Same.
12837 (class range_query): Same.
12839 2021-06-10 Richard Biener <rguenther@suse.de>
12841 PR tree-optimization/101003
12842 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
12843 use the pattern stmt defs when linearizing a chain.
12845 2021-06-10 Jakub Jelinek <jakub@redhat.com>
12848 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
12849 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
12851 2021-06-10 Clement Chigot <clement.chigot@atos.net>
12853 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
12854 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
12856 2021-06-09 Andrew Pinski <apinski@marvell.com>
12858 PR tree-optimization/100925
12859 * match.pd (a ? CST1 : CST2): Limit transformations
12860 that would produce a negative to integeral types only.
12861 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
12863 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12866 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12868 * doc/tm.texi: Correctly update.
12870 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12872 * doc/tm.texi: Correctly update.
12874 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
12877 * doc/tm.texi.in (Trampolines): Add a missing blank line.
12879 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
12882 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
12883 and -ftrampolines work only with Ada.
12884 * doc/tm.texi.in (Trampolines): Likewise.
12885 * doc/tm.texi: Regenerated.
12887 2021-06-09 Carl Love <cel@us.ibm.com>
12889 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
12890 Add define for new builtins.
12891 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
12892 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
12893 overloaded builtin definitions.
12894 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
12895 VSIGNEXTSD2Q): Add builtin expansions.
12896 (SIGNEXT): Add P10 overload definition.
12897 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
12898 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
12899 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
12900 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
12901 vsignextend_si_v2di)[VIlong]: Add define_expand.
12902 Make define_insn vsx_sign_extend_si_v2di visible.
12903 * doc/extend.texi: Add documentation for the vec_signexti,
12904 vec_signextll builtins and vec_signextq.
12906 2021-06-09 Carl Love <cel@us.ibm.com>
12908 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
12909 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
12910 __floattikf_sw, __floatuntikf_sw respectively.
12911 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
12912 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
12913 define_insn for mode IEEE 128.
12915 2021-06-09 Carl Love <cel@us.ibm.com>
12917 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
12918 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
12919 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
12920 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
12921 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
12922 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
12923 uses of VSX_TI to VEC_TI.
12925 2021-06-09 Carl Love <cel@us.ibm.com>
12927 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
12929 2021-06-09 Carl Love <cel@us.ibm.com>
12931 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
12933 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
12934 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
12935 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
12936 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
12937 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
12938 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
12939 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
12941 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
12942 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
12943 altivec_vrlqnm): New define_expands.
12944 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
12945 VCMPGTUT_P): Add macro expansions.
12946 (BU_P10V_AV_P): Add builtin predicate definition.
12947 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
12948 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
12949 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
12950 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
12951 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
12952 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
12953 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
12954 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
12955 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
12956 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
12957 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
12958 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
12959 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
12960 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
12961 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
12962 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
12963 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
12964 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
12965 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
12966 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
12967 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
12968 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
12969 P10V_BUILTIN_MODU_V1TI):
12970 New overloaded definitions.
12971 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
12972 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
12973 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
12974 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
12975 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
12976 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
12978 (altivec_init_builtins): New E_V1TImode case statement.
12979 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
12980 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
12981 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
12982 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
12983 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
12984 E_V1TImode]: New case statements.
12985 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
12986 value RS6000_BTI_bool_V1TI.
12987 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
12988 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
12989 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
12990 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
12991 vlshrv1ti3, vashrv1ti3): New define_expands.
12992 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
12993 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
12994 UNSPEC_VSX_MODUQ): New unspecs.
12995 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
12996 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
12998 (vcmpnet): New define_expand.
12999 * doc/extend.texi: Add documentation for the new builtins vec_rl,
13000 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
13001 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
13002 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
13003 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
13004 vec_any_ge, vec_any_le.
13006 2021-06-09 Carl Love <cel@us.ibm.com>
13008 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
13009 bug in argument generation.
13011 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13013 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
13015 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
13016 remove <supf> iterator.
13017 (mve_vclzq_u<mode>): New.
13018 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
13019 (neon_vclz<mode): Move to ...
13020 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
13021 * config/arm/vec-common.md: ... here. Add support for MVE.
13023 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13025 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
13026 (@mve_vrhaddq_<supf><mode): Likewise.
13027 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
13028 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
13029 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
13031 2021-06-09 imba-tjd <109224573@qq.com>
13033 * doc/invoke.texi: Fix typo.
13035 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
13037 PR middle-end/53267
13038 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
13039 Support evaluation of fmod/fmodf/fmodl at compile-time.
13041 2021-06-09 Richard Biener <rguenther@suse.de>
13043 PR tree-optimization/100981
13044 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
13045 gimple_get_lhs to also handle calls.
13046 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
13049 2021-06-09 Richard Biener <rguenther@suse.de>
13051 PR tree-optimization/97832
13052 * tree-vectorizer.h (_slp_tree::failed): New.
13053 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
13055 (_slp_tree::~_slp_tree): Free failed.
13056 (vect_build_slp_tree): Retain failed nodes and record
13057 matches in them, copying that back out when running
13058 into a cached fail. Dump start and end of discovery.
13059 (dt_sort_cmp): New.
13060 (vect_build_slp_tree_2): Handle associatable chains
13061 together doing more aggressive operand swapping.
13063 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13066 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
13068 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
13069 Linux and GNU targets.
13071 2021-06-09 Richard Biener <rguenther@suse.de>
13073 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
13076 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13078 * config/arc/arc.md (loop_end): Change it to
13079 define_insn_and_split.
13081 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13083 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
13084 (machi): New pattern.
13085 (umaddhisi4): Use VMAC2HU instruction.
13086 (umachi): New pattern.
13088 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13090 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
13091 * config/arc/arc.c (arc_split_move_p): New function.
13092 (arc_split_move): Clean up.
13093 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
13094 (movdf_insn): Likewise.
13095 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
13097 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
13100 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
13101 argument to "raw". Do not emit segment overrides when "raw" is true.
13103 2021-06-09 Martin Liska <mliska@suse.cz>
13105 * doc/gcov.texi: Create a proper JSON files.
13106 * doc/invoke.texi: Remove dots in order to make it a valid
13109 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
13111 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
13112 (insn_is_load_p): Use pattern_is_rotate64.
13113 (insn_is_swap_p): Likewise.
13114 (quad_aligned_load_p): Likewise.
13115 (const_load_sequence_p): Likewise.
13116 (replace_swapped_aligned_load): Likewise.
13117 (recombine_lvx_pattern): Likewise.
13118 (recombine_stvx_pattern): Likewise.
13120 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13122 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13123 fur_stmt source record.
13124 * gimple-range.cc (fur_source::get_operand): Generic range query.
13125 (fur_source::get_phi_operand): New.
13126 (fur_source::register_dependency): New.
13127 (fur_source::query): New.
13128 (class fur_edge): New. Edge source for operands.
13129 (fur_edge::fur_edge): New.
13130 (fur_edge::get_operand): New.
13131 (fur_edge::get_phi_operand): New.
13132 (fur_edge::query): New.
13133 (fur_stmt::fur_stmt): New.
13134 (fur_stmt::get_operand): New.
13135 (fur_stmt::get_phi_operand): New.
13136 (fur_stmt::query): New.
13137 (class fur_depend): New. Statement source and process dependencies.
13138 (fur_depend::fur_depend): New.
13139 (fur_depend::register_dependency): New.
13140 (class fur_list): New. List source for operands.
13141 (fur_list::fur_list): New.
13142 (fur_list::get_operand): New.
13143 (fur_list::get_phi_operand): New.
13144 (fold_range): New. Instantiate appropriate fur_source class and fold.
13145 (fold_using_range::range_of_range_op): Use new API.
13146 (fold_using_range::range_of_address): Ditto.
13147 (fold_using_range::range_of_phi): Ditto.
13148 (imple_ranger::fold_range_internal): Use fur_depend class.
13149 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13150 * gimple-range.h (class fur_source): Now a base class.
13151 (class fur_stmt): New.
13152 (fold_range): New prototypes.
13153 (fur_source::fur_source): Delete.
13155 2021-06-08 Andrew Pinski <apinski@marvell.com>
13157 PR tree-optimization/25290
13158 * tree-ssa-phiopt.c (xor_replacement): Delete.
13159 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13160 (match_simplify_replacement): Allow one cheap preparation
13161 statement that can be moved to before the if.
13163 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13165 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13166 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13168 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13170 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13171 Create length attribute on define_insn_and_split. Only split for cases which we
13173 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13174 fix length computation.
13175 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13177 2021-06-08 Richard Biener <rguenther@suse.de>
13179 PR tree-optimization/100923
13180 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13181 the operand vector to be valueized.
13182 (valueize_refs): Likewise.
13183 (valueize_shared_reference_ops_from_ref): Adjust.
13184 (valueize_shared_reference_ops_from_call): Likewise.
13185 (vn_reference_lookup_3): Likewise.
13186 (vn_reference_lookup_pieces): Likewise. Re-valueize
13187 with honoring availability when we are about to create
13188 the ao_ref and valueized before.
13189 (vn_reference_lookup): Likewise.
13190 (vn_reference_insert_pieces): Adjust.
13192 2021-06-08 Richard Biener <rguenther@suse.de>
13194 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13195 (_slp_instance::root_stmts): ... a vector.
13196 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13197 (SLP_INSTANCE_ROOT_STMTS): ... this.
13198 (slp_root::root): Change to...
13199 (slp_root::roots): ... a vector.
13200 (slp_root::slp_root): Adjust.
13201 * tree-vect-slp.c (_slp_instance::location): Adjust.
13202 (vect_free_slp_instance): Release the root stmt vector.
13203 (vect_build_slp_instance): Adjust.
13204 (vect_analyze_slp): Likewise.
13205 (_bb_vec_info::~_bb_vec_info): Likewise.
13206 (vect_slp_analyze_operations): Likewise.
13207 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13208 costs for the root stmt.
13209 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13211 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13213 (vectorize_slp_instance_root_stmt): Adjust.
13214 (vect_schedule_slp): Likewise.
13216 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13218 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13219 (ssa_equiv_stack::ssa_equiv_stack): New.
13220 (ssa_equiv_stack::~ssa_equiv_stack): New.
13221 (ssa_equiv_stack::enter): New.
13222 (ssa_equiv_stack::leave): New.
13223 (ssa_equiv_stack::push_replacement): New.
13224 (ssa_equiv_stack::get_replacement): New.
13225 (is_pointer_ssa): New.
13226 (class pointer_equiv_analyzer): New.
13227 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13228 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13229 (pointer_equiv_analyzer::set_global_equiv): New.
13230 (pointer_equiv_analyzer::set_cond_equiv): New.
13231 (pointer_equiv_analyzer::get_equiv): New.
13232 (pointer_equiv_analyzer::enter): New.
13233 (pointer_equiv_analyzer::leave): New.
13234 (pointer_equiv_analyzer::get_equiv_expr): New.
13235 (pta_valueize): New.
13236 (pointer_equiv_analyzer::visit_stmt): New.
13237 (pointer_equiv_analyzer::visit_edge): New.
13238 (hybrid_folder::value_of_expr): Call PTA.
13239 (hybrid_folder::value_on_edge): Same.
13240 (hybrid_folder::pre_fold_bb): New.
13241 (hybrid_folder::post_fold_bb): New.
13242 (hybrid_folder::pre_fold_stmt): New.
13243 (rvrp_folder::pre_fold_bb): New.
13244 (rvrp_folder::post_fold_bb): New.
13245 (rvrp_folder::pre_fold_stmt): New.
13246 (rvrp_folder::value_of_expr): Call PTA.
13247 (rvrp_folder::value_on_edge): Same.
13249 2021-06-08 Jakub Jelinek <jakub@redhat.com>
13252 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
13253 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
13255 2021-06-08 Richard Biener <rguenther@suse.de>
13257 PR middle-end/100951
13258 * tree-vect-generic.c (expand_vector_piecewise): Build a
13259 VECTOR_CST if all elements are constant.
13260 (expand_vector_condition): Likewise.
13261 (lower_vec_perm): Likewise.
13262 (expand_vector_conversion): Likewise.
13264 2021-06-08 Martin Liska <mliska@suse.cz>
13266 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13268 2021-06-08 Martin Liska <mliska@suse.cz>
13270 * genautomata.c (create_automata): Fix typo.
13272 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13274 PR tree-optimization/100794
13275 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13276 allow_unroll_p and only allow unrolling when it's true.
13277 (tree_predictive_commoning): Add parameter allow_unroll_p and
13279 (run_tree_predictive_commoning): Likewise.
13280 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13281 global_options_set.x_flag_predictive_commoning.
13282 (pass_predcom::execute): Adjust for allow_unroll_p.
13284 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13286 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13287 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13288 lambda function cleanup, remove scev_reset call, and adjust return
13290 (tree_predictive_commoning): Adjust for different changed values,
13291 only set flag TODO_update_ssa_only_virtuals if changed.
13292 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13293 from todo_flags_finish.
13295 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13297 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13298 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13299 (sbr_sparse_bitmap::bitmap_set_quad): New.
13300 (sbr_sparse_bitmap::bitmap_get_quad): New.
13301 (sbr_sparse_bitmap::set_bb_range): New.
13302 (sbr_sparse_bitmap::get_bb_range): New.
13303 (sbr_sparse_bitmap::bb_range_p): New.
13304 (block_range_cache::block_range_cache): initialize bitmap obstack.
13305 (block_range_cache::~block_range_cache): Destruct obstack.
13306 (block_range_cache::set_bb_range): Decide when to utilze the
13307 sparse on entry cache.
13308 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13309 * params.opt (-param=evrp-sparse-threshold): New.
13311 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13313 * bitmap.c (bitmap_set_aligned_chunk): New.
13314 (bitmap_get_aligned_chunk): New.
13315 (test_aligned_chunk): New.
13316 (bitmap_c_tests): Call test_aligned_chunk.
13317 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13319 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13322 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13324 (ix86_expand_vector_init_one_nonzero): Ditto.
13325 (ix86_expand_vector_init_one_var): Ditto.
13326 (ix86_expand_vector_init_general): Ditto.
13327 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13329 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13331 * config/h8300/movepush.md: Change most _clobber_flags
13332 patterns to instead use <cczn> subst.
13333 (movsi_cczn): New pattern with usable CC cases split out.
13334 (movsi_h8sx_cczn): Likewise.
13336 2021-06-07 Martin Liska <mliska@suse.cz>
13338 * common/common-target.def: Split long lines and replace them
13340 * target.def: Likewise.
13341 * doc/tm.texi: Re-generated.
13343 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13346 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13347 read from a CONSTRUCTOR with vector type elements.
13349 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13351 PR middle-end/100898
13352 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13353 should copy any arguments. Don't call gimple_call_num_args
13354 on id->call_stmt or call_stmt more than once.
13356 2021-06-07 liuhongt <hongtao.liu@intel.com>
13359 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13361 (<insn>v4siv4di2): Delete constraints for define_expand.
13363 2021-06-07 liuhongt <hongtao.liu@intel.com>
13366 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13367 assignment of cfun->machine->has_explicit_vzeroupper.
13368 * config/i386/i386-features.c
13369 (ix86_add_reg_usage_to_vzerouppers): Delete.
13370 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13371 (rest_of_handle_insert_vzeroupper): Remove
13372 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13374 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13375 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13377 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13378 (ix86_initialize_callee_abi): Ditto.
13379 (ix86_expand_avx_vzeroupper): Ditto.
13380 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13382 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13383 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13385 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13386 has_explicit_vzeroupper.
13387 * config/i386/i386.md (enum unspec): New member
13389 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13390 define_constants for insn callee abi index.
13391 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13392 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13393 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13394 (*avx_vzeroupper): Rename to ..
13395 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13396 call_insn which has a special vzeroupper ABI.
13397 (*avx_vzeroupper_1): Deleted.
13399 2021-06-07 liuhongt <hongtao.liu@intel.com>
13402 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13403 it won't use stack pointer reg.
13404 * final.c (leaf_function_p): When call_insn is a fake call, it
13405 won't affect caller as a leaf function.
13406 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13407 (subst_stack_regs): When call_insn doesn't clobber any stack
13408 reg, don't clear the arguments.
13409 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13411 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13412 frame for a fake call.
13413 * rtl.h (FAKE_CALL_P): New macro.
13415 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13417 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13419 (sparc_order_regs_for_local_alloc): ...this.
13420 (sparc_leaf_reg_remap): Declare.
13421 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13422 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13423 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13424 (order_regs_for_local_alloc): Rename to...
13425 (sparc_order_regs_for_local_alloc): ...this.
13426 (sparc_leaf_reg_remap): New function.
13427 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13429 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13431 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13432 Use assemble_name to output BSS section name.
13434 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13436 * config/i386/constraints.md (Bs):
13437 Remove boolean operators from match_test RTX.
13440 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13443 2021-06-06 Martin Liska <mliska@suse.cz>
13445 * doc/extend.texi: Add missing @headitem.
13446 * doc/invoke.texi: Likewise.
13447 * doc/objc.texi: Likewise.
13449 2021-06-06 Martin Liska <mliska@suse.cz>
13451 * genhooks.c (emit_findices): Remove unused function.
13452 (emit_documentation): Do not call emit_findices
13453 and do not search for @Fcode directives.
13455 2021-06-06 Martin Liska <mliska@suse.cz>
13457 * doc/invoke.texi: Remove extra character.
13459 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13461 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13463 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13465 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13466 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13467 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13468 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13470 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13472 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13474 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13476 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13479 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13481 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13482 *zero_extendsidi2): Fix empty split condition.
13484 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13486 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13488 * config/h8300/bitfield.md: Likewise.
13489 * config/h8300/combiner.md: Likewise.
13490 * config/h8300/divmod.md: Likewise.
13491 * config/h8300/extensions.md: Likewise.
13492 * config/h8300/jumpcall.md: Likewise.
13493 * config/h8300/movepush.md: Likewise.
13494 * config/h8300/multiply.md: Likewise.
13495 * config/h8300/other.md: Likewise.
13496 * config/h8300/shiftrotate.md: Likewise.
13497 * config/h8300/logical.md: Likewise. Fix split pattern to use
13498 code iterator that somehow slipped through.
13500 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13502 PR middle-end/100905
13503 * tree-nested.c (convert_nonlocal_omp_clauses,
13504 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13506 2021-06-04 Martin Sebor <msebor@redhat.com>
13508 PR middle-end/100732
13509 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13510 with either source or destination argument of invalid type.
13511 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13512 calls with arguments of invalid type.
13514 2021-06-04 Martin Sebor <msebor@redhat.com>
13516 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13518 (attr_access::vla_bounds): Also handle VLA bounds.
13520 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13522 * config/i386/predicates.md (GOT_memory_operand):
13523 Implement using match_code RTXes.
13524 (GOT32_symbol_operand): Ditto.
13526 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13529 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13531 (ix86_expand_vector_init_general): Ditto.
13532 Use SImode instead of word_mode for logic operations
13533 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13534 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13535 implemented by expand_vec_perm_1.
13536 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13537 should be implemented using standard shuffle patterns.
13538 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13539 V2HI modes to modes, implementable with shuffle for one operand.
13540 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13541 (*pshufw_1): New insn pattern.
13542 (*vec_dupv2hi): Ditto.
13543 (vec_initv2hihi): New expander.
13545 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13547 * config/arm/vfp.md (no_literal_pool_df_immediate,
13548 no_literal_pool_sf_immediate): Fix empty split condition.
13550 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13552 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13553 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13554 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13555 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13556 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13558 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13561 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13562 concatenation from half-sized modes with TImode elements.
13564 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13566 * config/arc/arc.c (arc_override_options): Disable millicode
13567 thunks when RF16 is on.
13569 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13571 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13573 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13575 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13576 Replace PROMOTE_MODE marco with its content.
13578 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13580 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13582 2021-06-03 Jim Wilson <jimw@sifive.com>
13584 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13585 turn it on for all riscv targets.
13587 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13590 * config/i386/i386-expand.c (ix86_expand_vector_set):
13591 Handle V2HI and V4QI modes.
13592 (ix86_expand_vector_extract): Ditto.
13593 * config/i386/mmx.md (*pinsrw): New insn pattern.
13596 (*pextrw_zext): Ditto.
13598 (*pextrb_zext): Ditto.
13599 (vec_setv2hi): New expander.
13600 (vec_extractv2hihi): Ditto.
13601 (vec_setv4qi): Ditto.
13602 (vec_extractv4qiqi): Ditto.
13603 (vec_setv8qi): Enable only for TARGET_SSE4_1.
13604 (vec_extractv8qiqi): Ditto.
13606 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
13608 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
13609 order to subf instruction.
13610 * config/rs6000/fusion.md: Regenerate.
13612 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13614 * calls.c (get_size_range): Use range_of_expr instead of
13615 determine_value_range.
13616 * tree-affine.c (expr_to_aff_combination): Same.
13617 * tree-data-ref.c (split_constant_offset): Same.
13618 * tree-vrp.c (determine_value_range_1): Remove.
13619 (determine_value_range): Remove.
13620 * tree-vrp.h (determine_value_range): Remove.
13622 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13624 * function-tests.c (test_ranges): Call gimple_range_tests.
13625 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
13627 * gimple-range.cc (fur_source::get_operand): Do not call
13628 get_tree_range or gimple_range_global.
13630 (get_tree_range): Move to value-query.cc.
13631 Call get_arith_expr_range.
13632 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
13633 Include gimple-range-tests.cc.
13634 * gimple-range.h (fold_range): Add argument.
13635 (get_tree_range): Remove.
13636 * selftest.h (gimple_range_tests): New.
13637 * value-query.cc (global_range_query::range_of_expr): Add
13639 (range_query::get_tree_range): Move from gimple-range.cc.
13640 * value-query.h (class range_query): Add get_tree_range and
13641 get_arith_expr_range. Make fur_source a friend.
13642 * vr-values.c (vr_values::range_of_expr): Pass stmt to
13644 * gimple-range-tests.cc: New file.
13646 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13648 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
13649 update_global_range.
13650 * value-query.cc (update_global_range): New.
13651 * value-query.h (update_global_range): New.
13653 2021-06-03 David Malcolm <dmalcolm@redhat.com>
13655 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
13656 printing the same location twice if there are fix-it hints,
13657 multiple locations, or a label.
13659 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
13661 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
13662 thresholds to narrow the upper bound on epilogue iterations.
13664 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
13666 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
13667 (mve_vabsq_s<mode>): Likewise.
13668 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
13669 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
13670 * config/arm/vec-common.md (neg<mode>2): Rename to
13671 <absneg_str><mode>2.
13673 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
13675 * common/config/arc/arc-common.c (arc_option_optimization_table):
13676 Remove malign-call.
13677 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
13678 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
13679 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
13680 * config/arc/arc.md (abssi2_mixed): Remove pattern.
13681 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
13682 (malign-call): Likewise.
13683 (mmixed-code): Likewise.
13684 * doc/invoke.texi (ARC): Update doc.
13686 2021-06-03 Martin Liska <mliska@suse.cz>
13688 * common.opt: Use proper Enum values.
13689 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
13690 (parse_sanitizer_options): Handle only sanitizer_opts.
13691 (common_handle_option): Just assign value.
13693 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13696 * tree-inline.c (inline_forbidden_p): Remove test on return type.
13698 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13700 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
13701 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
13702 (type_byte_size): Inline into...
13703 (add_byte_size_attribute): ...this and call add_scalar_info.
13705 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13707 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
13708 (typed_binop_from_tree): New function.
13709 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
13710 turn a divide by a power of 2 into a shift.
13711 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
13712 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
13713 typed divide by calling typed_binop_from_tree.
13715 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13717 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
13718 (is_handled_procedure_type): Likewise.
13719 (struct loc_descr_context): Add strict_signedness field.
13720 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
13721 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
13722 (resolve_args_picking): Minor tweak.
13723 (function_to_dwarf_procedure): Initialize strict_signedness field.
13724 (type_byte_size): Likewise.
13725 (field_byte_offset): Likewise.
13726 (gen_descr_array_type_die): Likewise.
13727 (gen_variant_part): Likewise.
13728 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
13729 to true when a context is present before evaluating the arguments.
13730 <COND_EXPR>: Do not generate a useless comparison with zero.
13731 When dereferencing an address, if strict_signedness is true and the
13732 type is small and signed, use DW_OP_deref_type to do the dereference
13733 and then DW_OP_convert to convert back to the generic type.
13735 2021-06-03 Jakub Jelinek <jakub@redhat.com>
13738 * tree-inline.c (copy_tree_body_r): Handle iterators on
13739 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
13741 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13743 * config/arc/arc.md (*bbit_di): Remove.
13745 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
13747 PR rtl-optimization/100264
13748 * ree.c (get_sub_rtx): Ignore SET expressions without register
13749 destinations and remove assertion, as it is not valid anymore
13750 with this new behaviour.
13751 (merge_def_and_ext): Eliminate destination check for register
13752 as such SET expressions can't occur anymore.
13753 (combine_reaching_defs): Likewise.
13755 2021-06-02 Jakub Jelinek <jakub@redhat.com>
13758 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
13759 -Wtype-limits warnings.
13760 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
13761 in operands to avoid -Wsign-compare warnings.
13763 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
13765 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
13768 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
13770 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
13772 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
13774 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
13776 * config/s390/subst.md(cconly_subst): Use a single constraint
13777 in (match_scratch).
13779 2021-06-02 Martin Liska <mliska@suse.cz>
13781 * ipa-icf.h: Use auto_vec for memory_access_types.
13783 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
13785 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
13786 argument from prototype.
13787 (output_logical_op): Add rtx_code argument.
13788 (compute_logical_op_length): Likewise.
13789 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
13790 to compute_a_shift_length.
13791 (output_logical_op); New argument with the rtx code rather than
13792 extracting it from an operand. Handle QImode too.
13793 (compute_logical_op_length): Similary.
13794 (compute_a_shift_length): Drop unused argument.
13795 * config/h8300/h8300.md (logicals): New code iterator.
13796 * config/h8300/logical.md (<code><mode>3 expander): Combine
13797 the "and" expander with the "ior"/"xor" expander.
13798 (bclr<mode>msx): Combine the QI/HI mode patterns.
13799 (<logical><mode>3 insns): Use code iterator rather than match_operator.
13800 Handle QImode as well. Update call to output_logical_op and
13801 compute_logical_op_length to pass in rtx_code
13802 Fix split condition on all define_insn_and_split patterns.
13803 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
13804 the flags and setting ZN via existing define_subst.
13805 * config/h8300/shiftrotate.md: Drop unused argument from
13806 calls to compute_a_shift_length.
13807 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
13809 2021-06-01 Andrew Pinski <apinski@marvell.com>
13811 PR tree-optimization/25290
13812 * tree-ssa-phiopt.c (match_simplify_replacement):
13814 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
13815 (two_value_replacement): Change the comment about
13816 conditional_replacement.
13817 (conditional_replacement): Delete.
13819 2021-06-01 Andrew Pinski <apinski@marvell.com>
13821 PR tree-optimization/95481
13822 * tree-tailcall.c (find_tail_calls): Handle empty typed
13825 2021-06-01 Andrew Pinski <apinski@marvell.com>
13827 * gimplify.c (zero_sized_field_decl): Delete
13828 (zero_sized_type): Delete
13829 (gimplify_init_ctor_eval): Use is_empty_type instead
13830 of zero_sized_field_decl.
13831 (gimplify_modify_expr): Use is_empty_type instead of
13834 2021-06-01 Jason Merrill <jason@redhat.com>
13837 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
13839 2021-06-01 Jason Merrill <jason@redhat.com>
13842 * diagnostic.h (warning_enabled_at): Declare.
13843 * diagnostic.c (diagnostic_enabled): Factor out from...
13844 (diagnostic_report_diagnostic): ...here.
13845 (warning_enabled_at): New.
13847 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
13849 * gimple-ssa-evrp.c: Enable exporting of global ranges.
13851 2021-06-01 Martin Liska <mliska@suse.cz>
13854 * doc/invoke.texi: Mention that -fgcse-after-reload
13855 is enabled with -O3.
13857 2021-06-01 liuhongt <hongtao.liu@intel.com>
13859 PR tree-optimization/98365
13860 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
13861 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
13862 (convert_scalar_cond_reduction): Ditto.
13863 (predicate_scalar_phi): Ditto.
13865 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13867 PR tree-optimization/100781
13868 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
13869 value calculation by default.
13870 (ranger_cache::enable_new_values): New.
13871 (ranger_cache::disable_new_values): New.
13872 (ranger_cache::push_poor_value): Check if new values are allowed.
13873 * gimple-range-cache.h (class ranger_cache): New member/methods.
13874 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
13875 statement, and disable/renable new value calculation.
13877 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13879 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
13880 (ranger_cache::range_of_def): New.
13881 (ranger_cache::entry_range): New.
13882 (ranger_cache::exit_range): New.
13883 (ranger_cache::range_of_expr): Adjust.
13884 (ranger_cache::range_on_edge): Adjust.
13885 (ranger_cache::propagate_cache): Call exit_range directly.
13886 * gimple-range-cache.h (class ranger_cache): Adjust.
13888 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13890 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
13891 gori_compute being a member rather than base class.
13892 dervied call to member call.
13893 (ranger_cache::dump): No longer dump gori_map.
13894 (ranger_cache::dump_bb): New.
13895 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
13896 being a member rather than base class.
13897 (ranger_cache::set_global_range): Ditto.
13898 (ranger_cache::ssa_range_in_bb): Ditto.
13899 (ranger_cache::range_of_expr): New.
13900 (ranger_cache::range_on_edge): New.
13901 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
13902 (ranger_cache::propagate_cache): Adjust debugging output.
13903 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
13905 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
13906 member, and inherit from range_query instead.
13907 (ranger_cache::dump_bb): New. split from dump.
13908 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
13909 (gori_compute::expr_range_at_stmt): Delete.
13910 (gori_compute::compute_name_range_op): Delete.
13911 (gori_compute::compute_operand_range_switch): Add fur_source.
13912 (gori_compute::compute_operand_range): Add fur_source param, inline
13913 old compute_name_range_op and optimize_logical_operands.
13914 (struct tf_range): Delete.
13915 (gori_compute::logical_combine): Adjust
13916 (gori_compute::optimize_logical_operands): Delete.
13917 (gori_compute::compute_logical_operands_in_chain): Delete.
13918 (gori_compute::compute_logical_operands): Adjust.
13919 (gori_compute::compute_operand1_range): Adjust to fur_source.
13920 (gori_compute::compute_operand2_range): Ditto.
13921 (gori_compute::compute_operand1_and_operand2_range): Ditto.
13922 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
13923 and adjust to fur_source.
13924 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
13925 range_query and fur_source.
13926 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
13927 from the ranger_cache..
13928 (gimple_ranger::fold_range_internal): Adjust to base class change of
13930 (gimple_ranger::dump_bb): Adjust dump.
13931 * gimple-range.h (gimple_ranger):export gori computes object.
13933 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13935 PR tree-optimization/100774
13936 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
13937 Constant values are also not stale.
13938 (ranger_cache::set_global_range): Range invariant values should also
13939 have the correct timestamp.
13941 2021-05-31 Martin Liska <mliska@suse.cz>
13943 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
13944 Unpack FUNCTION_DECL_DECL_TYPE.
13945 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
13946 Stream FUNCTION_DECL_DECL_TYPE instead of
13947 DECL_IS_OPERATOR_NEW_P.
13948 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
13950 (DECL_IS_OPERATOR_NEW_P): Likewise.
13951 (DECL_IS_OPERATOR_DELETE_P): Likewise.
13952 (DECL_LAMBDA_FUNCTION_P): Likewise.
13954 2021-05-31 Richard Biener <rguenther@suse.de>
13957 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
13958 * internal-fn.def (SHUFFLEVECTOR): New.
13959 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
13960 * doc/extend.texi: Document __builtin_shufflevector.
13962 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
13965 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
13966 indexed form addresses.
13968 2021-05-29 Jeff Law <jlaw@tachyum.com>
13970 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
13971 parameter. Call callers fixed.
13973 (output_plussi): Add FALLTHRU markers.
13974 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
13976 2021-05-29 Jakub Jelinek <jakub@redhat.com>
13978 PR middle-end/99928
13979 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
13980 combined with parallel, make sure to add shared clause to
13981 parallel for explicit linear clause.
13983 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
13985 PR tree-optimization/100787
13986 * gimple-ssa-evrp.c: Disable exporting of global ranges.
13988 2021-05-28 Jason Merrill <jason@redhat.com>
13990 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
13991 operator--, operator*, operator==, and operator!=.
13992 (class tsi_range): New.
13994 2021-05-28 Richard Biener <rguenther@suse.de>
13996 PR tree-optimization/100778
13997 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
13998 trapping ops in different BBs.
14000 2021-05-28 Richard Biener <rguenther@suse.de>
14003 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
14004 copy fntype from original call.
14006 2021-05-28 Martin Liska <mliska@suse.cz>
14008 PR gcov-profile/100751
14009 * doc/gcov.texi: Revert partially a hunk that was wrong.
14011 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
14013 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
14015 (HAVE_sync_compare_and_swaphi): Likewise.
14016 (HAVE_sync_compare_and_swapsi): Likewise.
14018 2021-05-28 Jakub Jelinek <jakub@redhat.com>
14020 PR middle-end/99928
14021 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
14023 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
14025 * gimplify.c (gimplify_omp_affinity): New.
14026 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
14027 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
14028 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
14029 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
14030 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
14032 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
14033 Richard Biener <rguenther@suse.de>
14035 * match.pd <popcount & / + pattern matching>:
14036 When generating popcount directly fails, try doing it in two halves.
14038 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14040 * Makefile.in (generated_files): Add gimple-match.c and
14043 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14045 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
14047 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14049 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
14051 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
14053 PR tree-optimization/99398
14054 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
14055 where the fed operands are CTOR/CST and propagated through
14056 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
14057 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
14059 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
14062 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14064 * config/i386/mmx.md (addv2sf3): Do not call
14065 ix86_fixup_binary_operands_no_copy.
14068 (<smaxmin:code>v2sf3): Ditto.
14069 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
14070 (<plusminus:insn><VI_32:mode>3): Remove expander.
14071 (<plusminus:insn><VI_32:mode>3): Rename from
14072 "*<plusminus:insn><VI_32:mode>3".
14073 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
14074 (mulv2hi3): Remove expander.
14075 (mulv2hi3): Rename from *mulv2hi3.
14076 (<s>mulv2hi3_highpart): Remove expander.
14077 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
14078 (<smaxmin:code><MMXMODE14:mode>3): Rename from
14079 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
14080 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
14081 (SMAXMIN_MMXMODEI): Remove mode iterator.
14082 (<smaxmin:code>v4hi3): New expander.
14083 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
14084 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
14085 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14086 (SMAXMIN_VI_32): Remove mode iterator.
14087 (<umaxmin:code><MMXMODE24:mode>3): Rename from
14088 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
14089 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
14090 (UMAXMIN_MMXMODEI): Remove mode iterator.
14091 (<umaxmin:code>v8qi3): New expander.
14092 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
14093 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
14094 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14095 (UMAXMIN_VI_32): Remove mode iterator.
14096 (<any_shift:insn>v2hi3): Remove expander.
14097 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
14098 (<any_logic:code><MMXMODEI:mode>3): Do not call
14099 ix86_fixup_binary_operands_no_copy.
14100 (<any_logic:code><VI_32:mode>3): Remove expander.
14101 (<any_logic:code><VI_32:mode>3): Rename from
14102 "*<any_logic:code><VI_32:mode>3".
14103 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
14104 * config/i386/sse.md (div<VF2:mode>3): Do not call
14105 ix86_fixup_binary_operands_no_copy.
14106 (div<VF1:mode>3): Ditto.
14107 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
14108 (smulhrsv4hi3): Ditto.
14109 (smulhrsv2hi3): Ditto.
14111 2021-05-27 Martin Sebor <msebor@redhat.com>
14113 * ggc.h (gt_ggc_mx): Add overloads for all integers.
14115 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14117 (hash_map::operator==): New function.
14119 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14122 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14123 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14124 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14125 (*xop_maskcmp<VI_32:mode>3): Ditto.
14126 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14127 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14129 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14132 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14133 opts_set, directly check opts parameters for being non-null.
14134 (arm_option_restore): Update call to arm_configure_build_target.
14135 (arm_option_override): Likewise.
14136 (arm_can_inline_p): Likewise.
14137 (arm_valid_target_attribute_tree): Likewise.
14138 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14139 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14142 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14144 * vr-values.c (simplify_conversion_using_ranges): Use
14145 get_range_query instead of get_global_range_query.
14147 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14149 * gimple-range.cc (get_range_global): Move to value-query.cc.
14150 (gimple_range_global): Same.
14151 (get_global_range_query): Same.
14152 (global_range_query::range_of_expr): Same.
14153 * gimple-range.h (class global_range_query): Move to
14155 (gimple_range_global): Same.
14156 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14157 (get_ptr_nonnull): Same.
14158 * tree-ssanames.h (get_range_info): Remove.
14159 (get_ptr_nonnull): Remove.
14160 * value-query.cc (get_ssa_name_range_info): Move from
14162 (get_ssa_name_ptr_info_nonnull): Same.
14163 (get_range_global): Move from gimple-range.cc.
14164 (gimple_range_global): Same.
14165 (get_global_range_query): Same.
14166 (global_range_query::range_of_expr): Same.
14167 * value-query.h (class global_range_query): Move from
14169 (gimple_range_global): Same.
14171 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14174 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14175 (uavgv2hi3_ceil): Ditto.
14177 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14180 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14182 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14184 * tree-ssanames.c (get_range_info): Merge both copies of
14185 get_range_info into one that works with irange.
14186 * tree-ssanames.h (get_range_info): Remove version that works on
14189 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14191 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14192 (expand_builtin_strnlen): Same.
14193 (determine_block_size): Same.
14194 * fold-const.c (expr_not_equal_to): Same.
14195 * gimple-fold.c (size_must_be_zero_p): Same.
14196 * gimple-match-head.c: Include gimple-range.h.
14197 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14198 * gimple-ssa-warn-restrict.c
14199 (builtin_memref::extend_offset_range): Same.
14200 * graphite-sese-to-poly.c (add_param_constraints): Same.
14201 * internal-fn.c (get_min_precision): Same.
14202 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14203 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14205 * tree-data-ref.c (split_constant_offset): Same.
14206 (dr_step_indicator): Same.
14207 * tree-dfa.c (get_ref_base_and_extent): Same.
14208 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14209 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14210 (determine_value_range): Same.
14211 (record_nonwrapping_iv): Same.
14212 (infer_loop_bounds_from_signedness): Same.
14213 (scev_var_range_cant_overflow): Same.
14214 * tree-ssa-phiopt.c (two_value_replacement): Same.
14215 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14216 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14217 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14219 (dump_strlen_info): Same.
14220 (set_strlen_range): Same.
14221 (maybe_diag_stxncpy_trunc): Same.
14222 (get_len_or_size): Same.
14223 (handle_integral_assign): Same.
14224 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14225 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14226 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14227 * tree-vect-patterns.c (vect_get_range_info): Same.
14228 (vect_recog_divmod_pattern): Same.
14229 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14230 (register_edge_assert_for_2): Same.
14231 (determine_value_range_1): Same.
14232 * tree.c (get_range_pos_neg): Same.
14233 * vr-values.c (vr_values::get_lattice_entry): Same.
14234 (vr_values::update_value_range): Same.
14235 (simplify_conversion_using_ranges): Same.
14237 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14239 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
14240 get_range_query instead of query argument.
14241 (pass_walloca::execute): Enable and disable global ranger.
14243 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14245 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
14247 (rvrp_folder::~rvrp_folder): Call disable_ranger.
14248 (hybrid_folder::hybrid_folder): Call enable_ranger.
14249 (hybrid_folder::~hybrid_folder): Call disable_ranger.
14251 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14253 * function.c (allocate_struct_function): Set cfun->x_range_query.
14254 * function.h (struct function): Declare x_range_query.
14255 (get_range_query): New.
14256 (get_global_range_query): New.
14257 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
14258 Remove call to safe_grow_cleared.
14259 * gimple-range.cc (get_range_global): New.
14260 (gimple_range_global): Move from gimple-range.h.
14261 (get_global_range_query): New.
14262 (global_range_query::range_of_expr): New.
14263 (enable_ranger): New.
14264 (disable_ranger): New.
14265 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
14266 (class global_range_query): New.
14267 (enable_ranger): New.
14268 (disable_ranger): New.
14269 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14270 dump_all_value_ranges to dump.
14271 * tree-vrp.c (vrp_prop::finalize): Same.
14272 * value-query.cc (range_query::dump): New.
14273 * value-query.h (range_query::dump): New.
14274 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14275 (vr_values::dump): ...this.
14276 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14277 dump and make virtual.
14279 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14281 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14282 Add V4QImode and V16QImode for TARGET_SSE2.
14283 * doc/sourcebuild.texi (Vector-specific attributes):
14284 Add vect64 and vect32 description.
14286 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14288 * gimple-range-gori.cc (range_def_chain::register_dependency):
14289 Resize m_def_chain when needed.
14291 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14293 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14294 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14295 * config/arm/vec-common.md: .. here. Add support for MVE.
14297 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14299 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14301 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14303 microblaze_option_override, print_operand): Likewise.
14304 * config/microblaze/microblaze.md (call_internal_plt,
14305 call_value_intern_plt, call_value_intern): Likewise.
14306 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14307 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14308 call_value_multiple_internal1): Likewise.
14309 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14311 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14313 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14314 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14315 arc_raw_symbolic_reference_mentioned_p): Remove register
14318 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14321 * omp-low.c: Include omp-offload.h.
14322 (create_omp_child_function): If current_function_decl has
14323 "omp declare target" attribute and is_gimple_omp_offloaded,
14324 remove that attribute from the copy of attribute list and
14325 add "omp target entrypoint" attribute instead.
14326 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14327 variables for offloading if in omp_maybe_offloaded_ctx.
14328 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14329 argument to GOMP_target_data_ext in offloaded code.
14331 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14333 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14334 For csky, HF/SF mode use the low bits of VREGS.
14336 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14338 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14339 DECL which is a reference for OMP.
14341 2021-05-26 Martin Liska <mliska@suse.cz>
14343 PR gcov-profile/100751
14344 * doc/gcov.texi: Document that __gcov_dump can be called just
14345 once and that __gcov_reset resets run-time counters.
14347 2021-05-26 Martin Liska <mliska@suse.cz>
14349 * doc/install.texi: Port relevant part from install-old.texi
14350 and re-generate list of CPUs and systems.
14352 2021-05-26 Martin Liska <mliska@suse.cz>
14354 * Makefile.in: Remove it.
14355 * doc/include/fdl.texi: Update next/previous chapters.
14356 * doc/install.texi: Likewise.
14357 * doc/install-old.texi: Removed.
14359 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14361 * config/csky/csky.c (ck810_legitimate_index_p): Support
14362 "base + index" with DF mode.
14363 * config/csky/constraints.md ("Y"): New constraint for memory operands
14364 without index register.
14365 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14366 when mov between memory and general registers, and lower their priority.
14367 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14369 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14371 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14373 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14375 * config/csky/csky.md (untyped_call): Emit clobber for return
14376 registers to mark them used.
14378 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14380 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14382 2021-05-26 Andrew Pinski <apinski@marvell.com>
14384 * match.pd (x < 0 ? ~y : y): New patterns.
14386 2021-05-26 Andrew Pinski <apinski@marvell.com>
14388 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14389 A?POW2:0 and A?0:POW2.
14391 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14393 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14394 (logical_stmt_cache::logical_stmt_cache ): Delete.
14395 (logical_stmt_cache::~logical_stmt_cache): Delete.
14396 (logical_stmt_cache::cache_entry::dump): Delete.
14397 (logical_stmt_cache::get_range): Delete.
14398 (logical_stmt_cache::cached_name ): Delete.
14399 (logical_stmt_cache::same_cached_name): Delete.
14400 (logical_stmt_cache::cacheable_p): Delete.
14401 (logical_stmt_cache::slot_diagnostics ): Delete.
14402 (logical_stmt_cache::dump): Delete.
14403 (gori_compute_cache::gori_compute_cache): Delete.
14404 (gori_compute_cache::~gori_compute_cache): Delete.
14405 (gori_compute_cache::compute_operand_range): Delete.
14406 (gori_compute_cache::cache_stmt): Delete.
14407 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14409 (class gori_compute_cache): Delete.
14411 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14413 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14415 (fold_using_range::range_of_address): Adjust.
14416 (fold_using_range::range_of_phi): Adjust.
14417 * gimple-range.h (class fur_source): Adjust.
14418 (fur_source::fur_source): Adjust.
14420 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14422 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14423 from expr_range_in_bb and adjust.
14424 (gori_compute::compute_name_range_op): Adjust.
14425 (gori_compute::optimize_logical_operands): Adjust.
14426 (gori_compute::compute_logical_operands_in_chain): Adjust.
14427 (gori_compute::compute_operand1_range): Adjust.
14428 (gori_compute::compute_operand2_range): Adjust.
14429 (ori_compute_cache::cache_stmt): Adjust.
14430 * gimple-range-gori.h (gori_compute): Rename prototype.
14432 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14434 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14435 checked only after range_of_stmt, not range_on_entry.
14436 (gimple_ranger::range_on_entry): Check for non-null in any
14437 predecessor block, if it is not already non-null.
14438 (gimple_ranger::range_on_exit): DOnt check for non-null after
14439 range on entry call.
14440 (gimple_ranger::dump_bb): New. Split from dump.
14441 (gimple_ranger::dump): Adjust.
14442 * gimple-range.h (class gimple_ranger): Adjust.
14444 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14446 * gimple-range-cache.cc (struct range_timestamp): Delete.
14447 (class temporal_cache): Adjust.
14448 (temporal_cache::get_timestamp): Delete.
14449 (temporal_cache::set_dependency): Delete.
14450 (temporal_cache::temporal_value): Adjust.
14451 (temporal_cache::current_p): Take dependencies as params.
14452 (temporal_cache::set_timestamp): Adjust.
14453 (temporal_cache::set_always_current): Adjust.
14454 (ranger_cache::get_non_stale_global_range): Adjust.
14455 (ranger_cache::register_dependency): Delete.
14456 * gimple-range-cache.h (class range_cache): Adjust.
14458 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14460 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14462 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14463 each individual bitmap.
14464 (range_def_chain::set_import): New.
14465 (range_def_chain::get_imports): New.
14466 (range_def_chain::chain_import_p): New.
14467 (range_def_chain::register_dependency): Rename from build_def_chain
14469 (range_def_chain::def_chain_in_bitmap_p): New.
14470 (range_def_chain::add_def_chain_to_bitmap): New.
14471 (range_def_chain::has_def_chain): Just check first depenedence.
14472 (range_def_chain::get_def_chain): Process imports, use generic
14473 register_dependency routine.
14474 (range_def_chain::dump): New.
14475 (gori_map::gori_map): Allocate import list.
14476 (gori_map::~gori_map): Release imports.
14477 (gori_map::exports): Check for past allocated block size.
14478 (gori_map::imports): New.
14479 (gori_map::def_chain_in_export_p): Delete.
14480 (gori_map::is_import_p): New.
14481 (gori_map::maybe_add_gori): Handle imports.
14482 (gori_map::dump): Adjust output, add imports.
14483 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14484 (gori_export_iterator::gori_export_iterator): New.
14485 (gori_export_iterator::next): New.
14486 (gori_export_iterator::get_name): New.
14487 * gimple-range-gori.h (range_def_chain): Add imports and direct
14488 dependecies via struct rdc.
14489 (range_def_chain::depend1): New.
14490 (range_def_chain::depend2): New.
14491 (class gori_map): Adjust.
14492 (FOR_EACH_GORI_IMPORT_NAME): New.
14493 (FOR_EACH_GORI_EXPORT_NAME): New.
14494 (class gori_export_iterator): New.
14496 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14498 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14499 export cache filling to here.
14500 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14502 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14504 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14505 (gori_map): Move to gimple-range-gori.h.
14506 (gori_compute::gori_compute): Adjust.
14507 (gori_compute::~gori_compute): Delete.
14508 (gori_compute::compute_operand_range_switch): Adjust.
14509 (gori_compute::compute_operand_range): Adjust.
14510 (gori_compute::compute_logical_operands): Adjust.
14511 (gori_compute::has_edge_range_p ): Adjust.
14512 (gori_compute::set_range_invariant): Delete.
14513 (gori_compute::dump): Adjust.
14514 (gori_compute::outgoing_edge_range_p): Adjust.
14515 * gimple-range-gori.h (class range_def_chain): Relocate here.
14516 (class gori_map): Relocate here.
14517 (class gori_compute): Inherit from gori_map, and adjust.
14519 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14521 * value-range.cc (range_tests_legacy): Use
14522 build_nonstandard_integer_type instead of int and short.
14524 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14526 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14527 when really creating an initialization statement for it.
14529 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14531 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14533 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14535 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14537 2021-05-25 Martin Liska <mliska@suse.cz>
14539 PR tree-optimization/92860
14541 * optc-save-gen.awk: Remove exceptions.
14543 2021-05-25 Martin Liska <mliska@suse.cz>
14545 * asan.h (sanitize_coverage_p): New function.
14546 * doc/extend.texi: Document it.
14547 * fold-const.c (fold_range_test): Use sanitize_flags_p
14548 instead of flag_sanitize_coverage.
14549 (fold_truth_andor): Likewise.
14550 * sancov.c: Likewise.
14551 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14552 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14553 -fsanitize-coverage when inlining.
14555 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14557 * config/csky/csky-modes.def : Fix copyright.
14559 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14561 * config/csky/csky-modes.def : Amend copyright.
14562 * config/csky/csky_insn_fpuv2.md : Likewise.
14563 * config/csky/csky_insn_fpuv3.md : Likewise.
14565 2021-05-25 Richard Biener <rguenther@suse.de>
14567 PR middle-end/100727
14568 * calls.c (initialize_argument_information): Explicitely test
14569 for WITH_SIZE_EXPR.
14570 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14572 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14574 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14575 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14576 the signle definition. The signle definition may not work well
14577 at simplify_subreg_regno().
14578 (HARD_FRAME_POINTER_REGNUM): New.
14579 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14580 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14581 csky_initial_elimination_offset, csky_expand_prologue,
14582 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14584 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14586 * config/csky/csky.c (csky_option_override):
14587 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14588 and TARGET_DIV can be set well.
14590 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14592 * config/csky/constraints.md ("l", "h"): Delete.
14593 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14594 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14595 * config/csky/csky.c (regno_reg_classm,
14596 csky_secondary_reload, csky_register_move_cost):
14597 Use HILO_REGS instead of LO_REGS and HI_REGS.
14599 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14601 * config/csky/constraints.md ("W"): New constriant for mem operand
14602 with base reg, index register.
14603 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
14604 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
14606 ("Dv"): New constraint for const double value that can be used at
14608 * config/csky/csky-modes.def (HFmode): New mode.
14609 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
14610 to "csky_valid_mem_constraint_operand" and support new constraint
14612 (csky_get_movedouble_length): New.
14613 (fpuv3_output_move): New.
14614 (fpuv3_const_double): New.
14615 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
14616 (decompose_csky_address): Refine.
14617 (csky_print_operand): New "CONST_DOUBLE" operand.
14618 (csky_output_move): Support fpv3 instructions.
14619 (csky_get_movedouble_length): New.
14620 (fpuv3_output_move): New.
14621 (fpuv3_const_double): New.
14622 (csky_emit_compare): Cover float comparsion.
14623 (csky_emit_compare_float): Refine.
14624 (csky_vaild_fpuv2_mem_operand): Rename to
14625 "csky_valid_mem_constraint_operand" and support new constraint "W".
14626 (ck860_rtx_costs): New.
14627 (csky_rtx_costs): Add the cost calculation of CK860.
14628 (regno_reg_class): New vregs for fpuv3.
14629 (csky_dbx_regno): Likewise.
14630 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
14631 (csky_conditional_register_usage): Suporrot fpuv3.
14632 (csky_dwarf_register_span): Suporrot fpuv3.
14633 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
14634 (ck810_legitimate_index_p): Support fp16.
14635 * config/csky/csky.h (TARGET_TLS): ADD CK860.
14636 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
14637 (TARGET_SINGLE_FPU): Support fpuv3.
14638 (TARGET_SUPPORT_FPV3): New.
14639 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
14640 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
14641 REG_CLASS_CONTENTS): Support fpuv3.
14642 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
14643 (csky_movsf_fpv2): Likewise.
14644 (ck801_movsf): Likewise.
14645 (csky_movsf): Likewise.
14647 (csky_movdf_fpv2): Likewise.
14648 (ck801_movdf): Likewise.
14649 (csky_movdf): Likewise.
14650 (movsicc): Refine. Use "comparison_operatior" instead of
14651 "ordered_comparison_operatior".
14652 (addsicc): Likewise.
14653 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
14654 (call_value_internal_vh): New.
14655 * config/csky/csky_cores.def (CK860): New arch and cpu.
14660 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
14661 into emit-patterns and match-patterns, remain the emit-patterns here,
14662 and move the match-patterns to csky_insn_fpuv2.md or
14663 csky_insn_fpuv3.md.
14664 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
14665 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
14667 * config/csky/csky_isa.def (fcr): New.
14672 (CK860): New definition for ck860.
14673 * config/csky/csky_tables.opt (ck860): New processors ck860,
14674 ck860f. And new arch ck860.
14679 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
14680 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
14681 * config/csky/t-csky-elf: Support 860.
14682 * config/csky/t-csky-linux: Likewise.
14683 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
14685 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
14687 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
14688 add generation of logical-add and add-logical fusion pairs.
14689 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
14691 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
14692 logical-add and add-logical fusion by default.
14693 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
14694 -mpower10-fusion-add-logical options.
14695 * config/rs6000/fusion.md: Regenerate file.
14697 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
14699 * value-range.cc (irange::legacy_equal_p): Check type when
14700 comparing VR_VARYING types.
14701 (range_tests_legacy): Test comparing VARYING ranges of different
14704 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14706 * config/aarch64/aarch64.c (neoversen1_tunings):
14707 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
14709 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14711 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
14712 extern weak symbols. Limit symbol offsets for non-GOT symbols with
14715 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14717 * config/arm/neon.md (vec_load_lanesxi<mode>)
14718 (vec_store_lanexoi<mode>): Move ...
14719 * config/arm/vec-common.md: here.
14721 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14723 * config/arm/neon.md (vec_load_lanesoi<mode>)
14724 (vec_store_lanesoi<mode>): Move ...
14725 * config/arm/vec-common.md: here.
14727 2021-05-24 liuhongt <hongtao.liu@intel.com>
14730 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
14731 stmt with GIMPLE_NOP when lhs doesn't exist.
14733 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
14736 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
14737 New instruction pattern.
14738 (*push<VI_32:mode>2): Ditto.
14739 (push splitter for SSE registers): New splitter.
14741 2021-05-23 Andrew Pinski <apinski@marvell.com>
14743 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
14745 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
14747 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
14748 * config/rs6000/fusion.md: Regenerate file.
14750 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14752 * config/rs6000/genfusion.pl (gen_addadd): New function.
14753 * config/rs6000/fusion.md: Regenerate file.
14754 * config/rs6000/rs6000-cpus.def: Add
14755 OPTION_MASK_P10_FUSION_2ADD to masks.
14756 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14757 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
14758 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
14760 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14762 PR middle-end/99928
14763 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
14764 * gimplify.c (enum gimplify_omp_var_data): Fix up
14765 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
14766 (omp_lastprivate_for_combined_outer_constructs): If combined target
14767 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
14768 GOVD_MAP | GOVD_SEEN.
14769 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
14770 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
14771 (gimplify_adjust_omp_clauses): For firstprivate clauses with
14772 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
14773 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
14774 let it be replaced by implicit map clause.
14776 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14778 PR middle-end/99928
14779 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
14781 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
14782 (gimplify_omp_for): Likewise.
14784 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14786 PR middle-end/90115
14787 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
14788 'external' in blocks.
14790 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14792 PR middle-end/90115
14793 * flag-types.h (enum openacc_privatization): New.
14794 * params.opt (-param=openacc-privatization): New.
14795 * doc/invoke.texi (openacc-privatization): Document it.
14796 * omp-general.h (get_openacc_privatization_dump_flags): New
14798 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
14799 * omp-offload.c (execute_oacc_device_lower)
14800 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
14801 * target.def (goacc.adjust_private_decl): Add 'location_t'
14803 * doc/tm.texi: Regenerate.
14804 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
14805 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
14806 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
14807 Likewise. Preserve it for...
14808 (nvptx_goacc_expand_var_decl): ... use here.
14810 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14812 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
14815 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14817 PR middle-end/90115
14818 * omp-low.c (oacc_privatization_candidate_p): New function.
14819 (oacc_privatization_scan_clause_chain)
14820 (oacc_privatization_scan_decl_chain): Use it. Also
14821 'gcc_checking_assert' that we're not seeing duplicates.
14823 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14825 PR middle-end/90115
14826 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
14829 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14831 PR middle-end/90115
14832 * omp-offload.c (execute_oacc_device_lower): Explain.
14834 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14836 PR middle-end/90115
14837 * omp-offload.c (execute_oacc_device_lower)
14838 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
14840 * internal-fn.c (expand_UNIQUE): Don't expect
14841 'IFN_UNIQUE_OACC_PRIVATE'.
14843 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14845 PR middle-end/90115
14846 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
14848 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14850 PR middle-end/90115
14851 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
14852 (nvptx_goacc_expand_var_decl): Tighten.
14854 2021-05-21 Julian Brown <julian@codesourcery.com>
14855 Chung-Lin Tang <cltang@codesourcery.com>
14856 Thomas Schwinge <thomas@codesourcery.com>
14858 PR middle-end/90115
14859 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
14860 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
14861 * doc/tm.texi: Regenerate.
14862 * expr.c (expand_expr_real_1): Expand decls using the
14863 expand_var_decl OpenACC hook if defined.
14864 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
14865 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
14866 * omp-low.c (omp_context): Add oacc_privatization_candidates
14868 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
14870 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
14871 private marker's gimple call arguments, and pass it to
14872 lower_oacc_reductions.
14873 (oacc_privatization_scan_clause_chain)
14874 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
14876 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
14877 * omp-offload.c (convert.h): Include.
14878 (oacc_loop_xform_head_tail): Treat private-variable markers like
14879 fork/join when transforming head/tail sequences.
14880 (struct var_decl_rewrite_info): Add struct.
14881 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
14882 (execute_oacc_device_lower): Support rewriting gang-private
14883 variables using target hook, and fix up addr_expr and var_decl
14885 * target.def (adjust_private_decl, expand_var_decl): New hooks.
14886 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
14888 (gcn_goacc_adjust_private_decl): ...this.
14889 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
14891 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
14892 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
14893 definition using gcn_goacc_adjust_gangprivate_decl...
14894 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
14895 gcn_goacc_adjust_private_decl.
14896 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
14897 (gang_private_shared_size): New global variable.
14898 (gang_private_shared_align): Likewise.
14899 (gang_private_shared_sym): Likewise.
14900 (gang_private_shared_hmap): Likewise.
14901 (nvptx_option_override): Initialize these.
14902 (nvptx_file_end): Output gang_private_shared_sym.
14903 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
14905 (nvptx_set_current_function): Clear gang_private_shared_hmap.
14906 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
14907 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
14909 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
14911 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
14913 2021-05-21 Richard Biener <rguenther@suse.de>
14914 H.J. Lu <hjl.tools@gmail.com>
14916 PR middle-end/90773
14917 * expr.c (expand_constructor): Elide expand_constructor if
14918 move by pieces is preferred.
14920 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14922 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
14923 Take a flag and mode value as arguments.
14924 (aarch64_modifies_global_state_p): Likewise.
14925 (aarch64_reads_global_state_p): Likewise.
14926 (aarch64_could_trap_p): Likewise.
14927 (aarch64_get_attributes): Likewise.
14928 (aarch64_init_simd_builtins): Adjust callsite of above.
14929 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
14930 function attributes to apply to builtins.
14931 (aarch64_init_crc32_builtins): Likewise.
14932 (aarch64_init_builtin_rsqrt): Likewise.
14934 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14936 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
14937 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
14938 (gen_2logical): Use new fusion types.
14939 * config/rs6000/fusion.md: Regenerate.
14941 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
14944 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
14945 Handle V4QI and V2HI modes.
14946 (ix86_expand_sse_movcc): Ditto.
14947 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
14948 New instruction pattern.
14949 (*eq<VI_32:mode>3): Ditto.
14950 (*gt<VI_32:mode>3): Ditto.
14951 (*xop_pcmov_<VI_32:mode>): Ditto.
14952 (mmx_pblendvb32): Ditto.
14953 (mmx_pblendvb64): Rename from mmx_pblendvb.
14954 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
14955 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
14956 (vcond<VI_32:mode><VI_32:mode>): Ditto.
14957 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
14958 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
14960 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14962 PR tree-optimization/94589
14963 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
14964 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
14965 not the same as cmp3) treat <= the same as < and >= the same as >.
14966 Don't require that cond2_phi_edge is true edge, instead take
14967 false/true edges into account based on cmp1/cmp2 comparison kinds.
14969 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
14972 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
14973 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
14974 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
14975 using SMAXMIN_MMXMODEI mode iterator.
14976 (*<smaxmin:code>v4qi3): New insn pattern.
14977 (*<smaxmin:code>v2hi3): Ditto.
14978 (SMAXMIN_VI_32): New mode iterator.
14979 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
14980 (UMAXMIN_MMXMODEI): New mode iterator.
14981 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
14982 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
14983 using UMAXMIN_MMXMODEI mode iterator.
14984 (*<umaxmin:code>v4qi3): New insn pattern.
14985 (*<umaxmin:code>v2hi3): Ditto.
14986 (UMAXMIN_VI_32): New mode iterator.
14987 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
14988 (abs<VI_32:mode>2): New insn pattern.
14989 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
14990 * config/i386/sse.md: ... here.
14992 2021-05-20 Clement Chigot <clement.chigot@atos.net>
14993 David Edelsohn <dje.gcc@gmail.com>
14995 * collect2.c (scan_prog_file): Issue non-fatal warning for
14998 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
15000 * doc/invoke.texi (-Wno-c++11-extensions)
15001 (-Wno-c++14-extensions, -Wno-c++17-extensions)
15002 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
15005 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15007 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
15008 * config/darwin.c (darwin_override_options): Likewise.
15009 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
15010 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
15011 (DWARF2_FRAME_REG_OUT): Likewise.
15012 * config/mips/mips.c (mips_output_filename): Likewise.
15013 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
15015 (rs6000_dbx_register_number): Likewise.
15016 * dbxout.c: Include flags.h.
15017 * dwarf2cfi.c (cfi_label_required_p): Likewise.
15018 (dwarf2out_do_frame): Likewise.
15019 * except.c: Include flags.h.
15020 * final.c (dwarf2_debug_info_emitted_p): Likewise.
15021 (final_scan_insn_1): Likewise.
15022 * flags.h (dwarf_debuginfo_p): New function declaration.
15023 * opts.c (dwarf_debuginfo_p): New function definition.
15024 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
15025 * toplev.c (process_options): Likewise.
15027 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15029 * common.opt: Change type to support bitmasks.
15030 * flag-types.h (enum debug_info_type): Rename enumerator constants.
15031 (NO_DEBUG): New bitmask.
15032 (DBX_DEBUG): Likewise.
15033 (DWARF2_DEBUG): Likewise.
15034 (XCOFF_DEBUG): Likewise.
15035 (VMS_DEBUG): Likewise.
15036 (VMS_AND_DWARF2_DEBUG): Likewise.
15037 * flags.h (debug_set_to_format): New function declaration.
15038 (debug_set_count): Likewise.
15039 (debug_set_names): Likewise.
15040 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
15041 (debug_set_to_format): New function definition.
15042 (debug_set_count): Likewise.
15043 (debug_set_names): Likewise.
15044 (set_debug_level): Update access to debug_type_names.
15045 * toplev.c: Likewise.
15047 2021-05-20 Martin Sebor <msebor@redhat.com>
15049 PR middle-end/100684
15050 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
15052 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15055 * config/i386/i386.md (isa): Remove x64_bmi.
15056 (enabled): Remove x64_bmi.
15057 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15058 Remove general register alternative.
15059 (*andnot<VI_32:mode>3): Ditto.
15060 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15061 (*<any_logic:code><VI_32:mode>3): Ditto.
15063 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15065 * config/arm/arm.c: Include head files tree-vectorizer.h and
15068 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15071 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
15072 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
15073 (<s>mulv4hi3_highpart): New expander.
15074 (*<s>mulv2hi3_highpart): New insn pattern.
15075 (<s>mulv2hi3_higpart): New expander.
15076 (*<any_shift:insn>v2hi3): New insn pattern.
15077 (<any_shift:insn>v2hi3): New expander.
15078 * config/i386/sse.md (smulhrsv2hi3): New expander.
15079 (*smulhrsv2hi3): New insn pattern.
15081 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15083 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
15085 * params.opt (vect-inner-loop-cost-factor): New.
15086 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
15087 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
15088 tree-vectorizer.h and its required ones.
15089 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
15090 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
15091 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
15092 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15093 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
15094 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15096 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
15097 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
15098 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
15100 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
15101 Torbjörn Svensson <torbjorn.svensson@st.com>
15104 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
15106 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15108 PR middle-end/99928
15109 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
15110 explicit linear clause when combined with target, make it map(tofrom:)
15111 instead of no clause or firstprivate.
15113 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15115 PR tree-optimization/94589
15116 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15117 of integral conversions.
15119 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15121 * gimple-range.cc (fur_source::get_operand): New.
15122 (gimple_range_fold): Delete.
15123 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15124 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15125 (fold_using_range::range_of_address): Ditto.
15126 (fold_using_range::range_of_phi): Ditto.
15127 (fold_using_range::range_of_call): Ditto.
15128 (fold_using_range::range_of_builtin_ubsan_call): Move from
15129 range_of_builtin_ubsan_call.
15130 (fold_using_range::range_of_builtin_call): Move from
15131 range_of_builtin_call.
15132 (gimple_ranger::range_of_builtin_call): Delete.
15133 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15134 (gimple_ranger::fold_range_internal): New.
15135 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15136 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15137 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15138 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15139 (class fur_source): New.
15140 (class fold_using_range): New.
15141 (fur_source::fur_source): New.
15143 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15144 instead of range_of_builtin_call.
15146 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15148 * doc/cpp.texi (Common Predefined Macros): Update documentation
15149 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15151 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15154 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15155 callee's address is in a register.
15157 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15159 * common/config/riscv/riscv-common.c
15160 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15162 (riscv_subset_list::parse_std_ext,
15163 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15164 in riscv_subset_list::parsing_subset_version.
15166 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15168 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15169 type attribute in patterns generating XTN(2).
15171 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15173 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15174 Remove as duplicate of...
15175 (aarch64_xtn<mode>): This.
15176 (aarch64_xtn2<mode>_le): Move position in file.
15177 (aarch64_xtn2<mode>_be): Move position in file.
15178 (aarch64_xtn2<mode>): Move position in file.
15179 (vec_pack_trunc_<mode>): Define as an expander.
15181 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15183 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15184 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15185 separate scalar and vector generators.
15186 * config/aarch64/aarch64-simd.md
15187 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15189 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15190 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15191 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15193 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15195 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15197 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15199 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15201 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15202 Implement as an expand emitting a big/little endian
15203 instruction pattern.
15204 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15205 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15207 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15209 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15210 Implement as an expand emitting a big/little endian
15211 instruction pattern.
15212 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15213 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15214 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15215 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15217 2021-05-19 Richard Biener <rguenther@suse.de>
15219 PR middle-end/100672
15220 * fold-const.c (fold_negate_expr_1): Use element_precision.
15221 (negate_expr_p): Likewise.
15223 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15225 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15226 (pred_load): New int attribute.
15227 * config/aarch64/aarch64-sve.md
15228 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15229 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15230 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15231 code_for_aarch64_load.
15233 2021-05-19 Richard Biener <rguenther@suse.de>
15235 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
15236 sure TARGET_MEM_REF bases are expanded as memory.
15237 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
15238 Do not mark TARGET_MEM_REF bases addressable.
15239 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
15240 TARGET_MEM_REF bases as never rewritable.
15241 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
15242 walk TARGET_MEM_REF bases as address-takens.
15243 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
15245 2021-05-19 Richard Biener <rguenther@suse.de>
15247 * builtins.c (get_object_alignment_1): Strip outer
15249 * tree-dfa.c (get_ref_base_and_extent): Handle outer
15250 WITH_SIZE_EXPR for size processing and process the
15252 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
15253 outer WITH_SIZE_EXPR.
15254 (ao_ref_base_alias_ptr_type): Likewise.
15255 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
15256 and handle that accordingly, stripping it for the
15257 core alias workers.
15258 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
15259 looking through it instead of returning NULL.
15261 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15263 PR middle-end/100576
15264 * builtins.c (check_read_access): Convert bound to size_type_node if
15267 2021-05-19 Richard Biener <rguenther@suse.de>
15269 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15270 (verify_types_in_gimple_reference): ... here. Sanitize.
15271 (verify_gimple_call): Verify references in LHS and arguments.
15272 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15274 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15276 * config/i386/i386.h (VALID_INT_MODE_P):
15277 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15278 * config/i386/i386.md (isa): Add x64_bmi.
15279 (enabled): Handle x64_bmi.
15280 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15281 Add alternative using 64bit general registers.
15282 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15284 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15286 PR middle-end/99928
15287 * tree.h (OMP_MASTER_COMBINED): Define.
15288 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15289 handling for outer combined/composite constructs to a loop.
15290 Handle lastprivate on combined target.
15291 (gimplify_expr): Formatting fix.
15293 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15295 * passes.def: Add sink_code pass before store_merging.
15296 * tree-ssa-sink.c (pass_sink_code:clone): New.
15298 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15300 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15301 rs6000_special_adjust_field_align_p.
15302 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15303 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15304 Remove ABI warning.
15305 (rs6000_function_arg): Likewise.
15306 * config/rs6000/rs6000-protos.h
15307 (rs6000_special_adjust_field_align_p): Remove prototype.
15308 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15310 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15311 rs6000_special_adjust_field_align_p.
15313 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15316 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15317 Add V4QI and V2HI modes.
15318 (VALID_INT_MODE_P): Ditto.
15319 * config/i386/mmx.md (VI_32): New mode iterator.
15320 (mmxvecsize): Handle V4QI and V2HI.
15322 (mov<VI_32:mode>): New expander.
15323 (*mov<mode>_internal): New insn pattern.
15324 (movmisalign<VI_32:mode>): New expander.
15325 (neg<VI_32:mode>): New expander.
15326 (<plusminus:insn><VI_32:mode>3): New expander.
15327 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15328 (mulv2hi3): New expander.
15329 (*mulv2hi3): New insn pattern.
15330 (one_cmpl<VI_32:mode>2): New expander.
15331 (*andnot<VI_32:mode>3): New insn pattern.
15332 (<any_logic:code><VI_32:mode>3): New expander.
15333 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15335 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15337 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15338 Fix a mode mismatch with operand 1.
15340 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15343 * config/i386/i386-expand.c (split_double_mode): Return
15344 temporary register when simplify_gen_subreg fails with
15345 the high half od the paradoxical subreg.
15347 2021-05-18 Richard Biener <rguenther@suse.de>
15349 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15350 and honor it when expanding.
15351 (expand_used_vars_for_block): Pass through forced_stack_var.
15352 (expand_used_vars): Likewise.
15353 (discover_nonconstant_array_refs_r): Set bits in
15354 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15355 (avoid_type_punning_on_regs): Likewise.
15356 (discover_nonconstant_array_refs): Likewise.
15357 (pass_expand::execute): Create and pass down forced_stack_var
15358 bitmap. For parameters and returns temporarily set
15359 TREE_ADDRESSABLE when expand_function_start.
15361 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15363 * doc/sourcebuild.texi: Document 'dg-note'.
15365 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15368 * configure: Regenerate.
15369 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15371 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15373 * gimple.h (is_gimple_omp_oacc): Tighten.
15374 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15376 2021-05-18 Richard Biener <rguenther@suse.de>
15378 * tree-ssa-operands.c (mark_address_taken): Simplify.
15380 2021-05-18 Martin Liska <mliska@suse.cz>
15382 * config/gcn/mkoffload.c (STR): Redefine.
15383 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15384 * config/nvptx/mkoffload.c (STR): Likewise.
15386 2021-05-18 Martin Liska <mliska@suse.cz>
15388 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15389 Use startswith function instead of strncmp.
15390 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15391 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15392 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15393 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15394 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15395 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15396 * config/arm/arm.c (arm_file_start): Likewise.
15397 (arm_valid_target_attribute_rec): Likewise.
15398 (thumb1_md_asm_adjust): Likewise.
15399 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15400 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15401 (avr_set_current_function): Likewise.
15402 (avr_handle_addr_attribute): Likewise.
15403 (avr_asm_output_aligned_decl_common): Likewise.
15404 (avr_asm_named_section): Likewise.
15405 (avr_section_type_flags): Likewise.
15406 (avr_asm_select_section): Likewise.
15407 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15408 (c6x_section_type_flags): Likewise.
15409 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15410 (darwin_objc_declare_unresolved_class_reference): Likewise.
15411 (darwin_objc_declare_class_definition): Likewise.
15412 * config/darwin.c (indirect_data): Likewise.
15413 (darwin_encode_section_info): Likewise.
15414 (darwin_objc2_section): Likewise.
15415 (darwin_objc1_section): Likewise.
15416 (machopic_select_section): Likewise.
15417 (darwin_globalize_label): Likewise.
15418 (darwin_label_is_anonymous_local_objc_name): Likewise.
15419 (darwin_asm_named_section): Likewise.
15420 (darwin_asm_output_dwarf_offset): Likewise.
15421 * config/frv/frv.c (frv_string_begins_with): Likewise.
15422 (frv_in_small_data_p): Likewise.
15423 * config/gcn/mkoffload.c (STR): Likewise.
15425 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15426 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15427 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15428 (ix86_md_asm_adjust): Likewise.
15429 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15430 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15431 (i386_pe_file_end): Likewise.
15432 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15433 (ia64_section_type_flags): Likewise.
15434 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15435 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15436 (mips16_stub_function_p): Likewise.
15437 (mips_function_rodata_section): Likewise.
15438 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15439 (msp430_function_section): Likewise.
15440 (msp430_section_type_flags): Likewise.
15441 (msp430_expand_helper): Likewise.
15442 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15443 (nios2_valid_target_attribute_rec): Likewise.
15444 * config/nvptx/mkoffload.c (process): Likewise.
15446 * config/pa/som.h: Likewise.
15447 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15448 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15449 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15450 (rs6000_inner_target_options): Likewise.
15451 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15452 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15453 * config/vax/vax.c (vax_output_int_move): Likewise.
15454 * config/vms/vms-ld.c (startswith): Likewise.
15455 (process_args): Likewise.
15457 * config/vms/vms.c: Likewise.
15459 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15461 PR rtl-optimization/100590
15462 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15463 they are NONJUMP_INSN_P.
15465 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15468 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15469 DECL_ASSEMBLER_NAME on the fn_decl.
15471 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15473 PR tree-optimization/94589
15474 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15475 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15476 res == 2 would be UB with -ffinite-math-only.
15478 2021-05-18 Martin Liska <mliska@suse.cz>
15480 * Makefile.in: genversion.o should depend on DATESTAMP.
15482 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15484 * config/arc/simdext.md (negv2si2): Remove round bracket.
15486 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15488 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15489 _Bool as macro expanding to _Bool.
15491 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15494 * tree.c (build_reference_type_for_mode)
15495 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15497 (build_reference_type, build_pointer_type): Invoke
15498 build_*_type_for_mode with VOIDmode.
15500 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15502 PR tree-optimization/100512
15503 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15504 and non-zero pointer ranges as invariant.
15505 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15506 processing from here.
15508 2021-05-17 Tom de Vries <tdevries@suse.de>
15511 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15512 * config/nvptx/nvptx.c (nvptx_output_barrier)
15513 (nvptx_output_atomic_insn): New function.
15514 (nvptx_print_operand): Add support for 'B'.
15515 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15518 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15520 PR tree-optimization/100349
15521 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15524 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15526 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15527 (host_detect_local_cpu): Use it.
15529 2021-05-17 Martin Liska <mliska@suse.cz>
15531 * doc/invoke.texi: Add 2 missing dots.
15533 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15535 PR bootstrap/100552
15536 * configure.ac: Replace pattern substitution with call to sed.
15537 * configure: Regenerate.
15539 2021-05-17 Richard Biener <rguenther@suse.de>
15541 PR middle-end/100582
15542 * tree.c (array_at_struct_end_p): Get to the base of the
15543 reference before looking for the underlying decl.
15545 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15547 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15548 alternatives" message more specific, and remove assumption on where
15551 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15553 * config/arm/iterators.md (V16): New iterator.
15554 (VH_cvtto): New iterator.
15555 (v_cmp_result): Added V4HF and V8HF support.
15556 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15557 (vcond<mode><mode>): Likewise.
15558 (vcond_mask_<mode><v_cmp_result>): Likewise.
15559 (vcond<VH_cvtto><mode>): New expander.
15561 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15563 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15565 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15567 (arm_expand_vcond): Likewise.
15568 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15569 VCMPEQQ_N_S, VCMPNEQ_N_S.
15570 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15571 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15572 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15573 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15574 (@mve_vpselq_<supf><mode>): Likewise.
15575 (@mve_vpselq_f<mode>"): Likewise.
15576 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15577 and move to vec-common.md.
15578 (vec_cmpu<mode><mode>): Likewise.
15579 (vcond<mode><mode>): Likewise.
15580 (vcond<V_cvtto><mode>): Likewise.
15581 (vcondu<mode><v_cmp_result>): Likewise.
15582 (vcond_mask_<mode><v_cmp_result>): Likewise.
15583 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15584 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15585 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15586 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15587 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15588 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15590 (vec_cmpu<mode><mode>): Likewise.
15591 (vcond<mode><mode>): Likewise.
15592 (vcond<V_cvtto><mode>): Likewise.
15593 (vcondu<mode><v_cmp_result>): Likewise.
15594 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15597 2021-05-17 liuhongt <hongtao.liu@intel.com>
15600 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
15601 gsi_insert_seq_before instead.
15603 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15605 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
15606 (arm_sat_ok): ...this.
15608 2021-05-17 Martin Liska <mliska@suse.cz>
15610 * lto-wrapper.c (merge_flto_options): Factor out a new function.
15611 (merge_and_complain): Use it.
15612 (run_gcc): Merge also linker command line -flto=foo argument
15615 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
15617 * config/arm/arm.h (CPP_SPEC): Remove error message about
15618 -mlittle-endian/-mbig-endian conflict.
15620 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15622 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
15623 __ROP_PROTECT__ if -mrop-protect is selected.
15625 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15627 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
15628 rop_hash_save_offset and rop_hash_size.
15629 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
15630 rop_hash_size and rop_hash_save_offset.
15631 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
15632 (rs6000_emit_prologue): Emit hashst[p] in prologue.
15633 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
15634 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
15636 (hashst): New define_insn.
15637 (hashchk): Likewise.
15639 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15641 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15642 Disable shrink wrap when inserting ROP-protect instructions.
15643 * config/rs6000/rs6000.opt (mrop-protect): New option.
15644 (mprivileged): Likewise.
15645 * doc/invoke.texi: Document mrop-protect and mprivileged.
15647 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
15649 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
15652 2021-05-15 Martin Jambor <mjambor@suse.cz>
15655 2021-05-13 Martin Jambor <mjambor@suse.cz>
15657 PR tree-optimization/100453
15658 * tree-sra.c (sra_modify_assign): All const base accesses do not
15659 need refreshing, not just those from decl_pool.
15660 (sra_modify_assign): Do not refresh into a const base decl.
15662 2021-05-15 Jakub Jelinek <jakub@redhat.com>
15664 PR rtl-optimization/100342
15665 * regcprop.c (copy_value): When copying a source reg in a wider
15666 mode than it has recorded for the value, adjust recorded destination
15667 mode too or punt if !REG_CAN_CHANGE_MODE_P.
15669 2021-05-14 Jason Merrill <jason@redhat.com>
15671 * intl.h: Add comments.
15673 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15675 * config/aarch64/aarch64-simd.md
15676 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
15677 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
15678 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
15679 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
15680 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
15681 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
15682 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
15683 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
15684 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
15686 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15689 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
15690 boolean logic equivalent.
15691 (vtst_s16): Likewise.
15692 (vtst_s32): Likewise.
15693 (vtst_u8): Likewise.
15694 (vtst_u16): Likewise.
15695 (vtst_u32): Likewise.
15696 (vtst_p8): Likewise.
15697 (vtst_p16): Likewise.
15698 (vtstq_s8): Likewise.
15699 (vtstq_s16): Likewise.
15700 (vtstq_s32): Likewise.
15701 (vtstq_u8): Likewise.
15702 (vtstq_u16): Likewise.
15703 (vtstq_u32): Likewise.
15704 (vtstq_p8): Likewise.
15705 (vtstq_p16): Likewise.
15706 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
15707 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
15709 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15711 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
15712 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
15713 (aarch64_sqdmlsl2<mode>): Delete.
15714 (aarch64_sqdmlal2_lane<mode>): Merge this...
15715 (aarch64_sqdmlsl2_lane<mode>): ... And this...
15716 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
15717 (aarch64_sqdmlal2_laneq<mode>): Merge this...
15718 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
15719 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
15720 (aarch64_sqdmlal2_n<mode>): Merge this...
15721 (aarch64_sqdmlsl2_n<mode>): ... And this...
15722 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
15724 2021-05-13 Martin Sebor <msebor@redhat.com>
15726 PR middle-end/100574
15727 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
15728 all null arguments.
15730 2021-05-13 Martin Sebor <msebor@redhat.com>
15732 PR tree-optimization/93100
15733 PR middle-end/98583
15734 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
15735 don't modify referenced objects.
15737 2021-05-13 Martin Jambor <mjambor@suse.cz>
15739 PR tree-optimization/100453
15740 * tree-sra.c (sra_modify_assign): All const base accesses do not
15741 need refreshing, not just those from decl_pool.
15742 (sra_modify_assign): Do not refresh into a const base decl.
15744 2021-05-13 Martin Liska <mliska@suse.cz>
15746 * tree-ssa-dom.c: Remove m_simplifier.
15748 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
15751 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
15752 canonicalize DImode inequality comparisons against the
15753 maximum integral value.
15755 2021-05-13 Jakub Jelinek <jakub@redhat.com>
15757 PR tree-optimization/98856
15758 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
15759 Expect V2DI and V4DI arithmetic right shifts to be emulated.
15760 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
15762 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
15763 expand_vec_perm_2perm_pblendv): New functions.
15764 (ix86_expand_vec_perm_const_1): Use them.
15765 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
15766 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
15767 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
15768 (ashrv4di3): New define_expand.
15769 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
15770 and !TARGET_AVX512VL expansion.
15772 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
15775 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
15776 sizes < 16 to a register when constructing vpcmov pattern.
15777 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
15779 2021-05-13 Martin Liska <mliska@suse.cz>
15781 * gcov-io.c (gcov_write_block): Remove.
15782 (gcov_write_words): Likewise.
15783 (gcov_read_words): Re-implement using gcov_read_bytes.
15784 (gcov_allocate): Remove.
15785 (GCOV_BLOCK_SIZE): Likewise.
15786 (struct gcov_var): Remove most of the fields.
15787 (gcov_position): Implement with ftell.
15788 (gcov_rewrite): Remove setting of start and offset fields.
15789 (from_file): Re-format.
15790 (gcov_open): Remove setbuf call. It should not be needed.
15791 (gcov_close): Remove internal buffer handling.
15792 (gcov_magic): Use __builtin_bswap32.
15793 (gcov_write_counter): Use directly gcov_write_unsigned.
15794 (gcov_write_string): Use direct fwrite and do not round
15796 (gcov_seek): Use directly fseek.
15797 (gcov_write_tag): Use gcov_write_unsigned directly.
15798 (gcov_write_length): Likewise.
15799 (gcov_write_tag_length): Likewise.
15800 (gcov_read_bytes): Use directly fread.
15801 (gcov_read_unsigned): Use gcov_read_words.
15802 (gcov_read_counter): Likewise.
15803 (gcov_read_string): Use gcov_read_bytes.
15804 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
15805 that size is not in bytes, but words (4B).
15806 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
15807 (GCOV_TAG_ARCS_LENGTH): Likewise.
15808 (GCOV_TAG_ARCS_NUM): Likewise.
15809 (GCOV_TAG_COUNTER_LENGTH): Likewise.
15810 (GCOV_TAG_COUNTER_NUM): Likewise.
15811 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
15813 2021-05-13 liuhongt <hongtao.liu@intel.com>
15816 * config/i386/sse.md (ssedoublevecmode): Add attribute for
15817 V64QI/V32HI/V16SI/V4DI.
15818 (ssehalfvecmode): Add attribute for V2DI/V2DF.
15819 (*vec_concatv4si_0): Extend to VI124_128.
15820 (*vec_concat<mode>_0): New pre-reload splitter.
15821 * config/i386/predicates.md (movq_parallel): New predicate.
15823 2021-05-13 Alexandre Oliva <oliva@adacore.com>
15825 * targhooks.c (default_zero_call_used_regs): Retry using
15826 successfully-zeroed registers as sources.
15828 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
15830 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
15832 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
15835 * gimple-range.cc (range_of_builtin_call): Skip out on
15836 processing __builtin_clz when varying.
15838 2021-05-12 Tom de Vries <tdevries@suse.de>
15841 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
15842 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
15843 to ptx_version_option.
15844 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
15845 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
15846 (define_insn "nvptx_vote_ballot"): Use sync variant for
15848 * config/nvptx/nvptx.opt (ptx_version): Add enum.
15849 (mptx): Add option.
15850 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
15852 2021-05-12 Richard Biener <rguenther@suse.de>
15854 PR tree-optimization/100566
15855 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
15856 allow_back for all edge queries.
15858 2021-05-12 liuhongt <hongtao.liu@intel.com>
15861 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
15862 splitters for pblendvb of NOT mask register.
15864 2021-05-12 Richard Biener <rguenther@suse.de>
15866 PR tree-optimization/100519
15867 * tree-ssa-reassoc.c (can_associate_p): Split into...
15868 (can_associate_op_p): ... this
15869 (can_associate_type_p): ... and this.
15870 (is_reassociable_op): Call can_associate_op_p.
15871 (break_up_subtract_bb): Call the appropriate predicates.
15872 (reassociate_bb): Likewise.
15874 2021-05-12 Martin Liska <mliska@suse.cz>
15876 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
15877 (run_gcc): Use -flto argument detection for merged
15880 2021-05-12 Martin Liska <mliska@suse.cz>
15882 * lto-wrapper.c (print_lto_docs_link): New function.
15883 (run_gcc): Print warning about missing job server detection
15884 after we know NR of partitions. Do the same for -flto{,=1}.
15885 * opts.c (get_option_html_page): Support -flto option.
15887 2021-05-12 Martin Liska <mliska@suse.cz>
15889 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
15891 (append_option): Remove.
15892 (find_option): Rework to use the vector type.
15893 (remove_option): Remove.
15894 (merge_and_complain): Use vectors for cl_decoded_option data
15896 (append_compiler_options): Likewise.
15897 (append_diag_options): Likewise.
15898 (append_linker_options): Likewise.
15899 (append_offload_options): Likewise.
15900 (compile_offload_image): Likewise.
15901 (compile_images_for_offload_targets): Likewise.
15902 (find_and_merge_options): Likewise.
15903 (run_gcc): Likewise.
15905 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
15908 * dwarf2out.c (dwarf2out_finish): Set
15909 have_multiple_function_sections with multi-range text_section.
15911 2021-05-12 Martin Liska <mliska@suse.cz>
15913 PR bootstrap/100560
15914 * Makefile.in: Remove version.h from linker command line.
15916 2021-05-12 Richard Biener <rguenther@suse.de>
15918 PR middle-end/100547
15919 * rtl.h (rtvec_alloc): Make argument size_t.
15920 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
15922 2021-05-12 Jakub Jelinek <jakub@redhat.com>
15924 PR middle-end/100508
15925 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
15926 type, don't reuse DECL_RTL if it has different mode, instead force
15927 creation of a new DEBUG_EXPR.
15929 2021-05-12 Jakub Jelinek <jakub@redhat.com>
15930 Marc Glisse <marc.glisse@inria.fr>
15932 PR tree-optimization/94589
15933 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
15934 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
15936 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
15939 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
15940 * config/i386/mmx.md (MMXMODE124): New mode iterator.
15942 (mmxintvecmode): New mode attribute.
15943 (mmxintvecmodelower): Ditto.
15944 (*mmx_maskcmpv2sf3_comm): New insn pattern.
15945 (*mmx_maskcmpv2sf3): Ditto.
15946 (vec_cmpv2sfv2si): New expander.
15947 (vcond<V2FI:mode>v2si): Ditto.
15948 (mmx_vlendvps): New insn pattern.
15949 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
15950 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
15951 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
15953 2021-05-11 Martin Sebor <msebor@redhat.com>
15955 PR middle-end/21433
15956 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
15958 2021-05-11 Richard Biener <rguenther@suse.de>
15960 * gimple-fold.c (gimple_fold_call): Do not call
15961 maybe_fold_reference on call arguments or the static chain.
15962 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
15965 2021-05-11 Martin Liska <mliska@suse.cz>
15967 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
15968 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
15969 (DEF_HSAIL_SAT_BUILTIN): Likewise.
15970 (DEF_HSAIL_INTR_BUILTIN): Likewise.
15971 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
15972 * doc/frontends.texi: Remove BRIG.
15973 * doc/install.texi: Likewise.
15974 * doc/invoke.texi: Likewise.
15975 * doc/standards.texi: Likewise.
15976 * brig-builtins.def: Removed.
15977 * brig/ChangeLog: Removed.
15978 * brig/Make-lang.in: Removed.
15979 * brig/brig-builtins.h: Removed.
15980 * brig/brig-c.h: Removed.
15981 * brig/brig-lang.c: Removed.
15982 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
15983 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
15984 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
15985 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
15986 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
15987 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
15988 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
15989 * brig/brigfrontend/brig-comment-handler.cc: Removed.
15990 * brig/brigfrontend/brig-control-handler.cc: Removed.
15991 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
15992 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
15993 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
15994 * brig/brigfrontend/brig-function-handler.cc: Removed.
15995 * brig/brigfrontend/brig-function.cc: Removed.
15996 * brig/brigfrontend/brig-function.h: Removed.
15997 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
15998 * brig/brigfrontend/brig-label-handler.cc: Removed.
15999 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
16000 * brig/brigfrontend/brig-machine.c: Removed.
16001 * brig/brigfrontend/brig-machine.h: Removed.
16002 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
16003 * brig/brigfrontend/brig-module-handler.cc: Removed.
16004 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
16005 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
16006 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
16007 * brig/brigfrontend/brig-to-generic.cc: Removed.
16008 * brig/brigfrontend/brig-to-generic.h: Removed.
16009 * brig/brigfrontend/brig-util.cc: Removed.
16010 * brig/brigfrontend/brig-util.h: Removed.
16011 * brig/brigfrontend/brig-variable-handler.cc: Removed.
16012 * brig/brigfrontend/hsa-brig-format.h: Removed.
16013 * brig/brigfrontend/phsa.h: Removed.
16014 * brig/brigspec.c: Removed.
16015 * brig/config-lang.in: Removed.
16016 * brig/gccbrig.texi: Removed.
16017 * brig/lang-specs.h: Removed.
16018 * brig/lang.opt: Removed.
16020 2021-05-11 Richard Biener <rguenther@suse.de>
16023 * ipa-param-manipulation.c
16024 (ipa_param_body_adjustments::modify_call_stmt): Avoid
16025 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
16026 via gimple_call_lhs_ptr.
16028 2021-05-11 Alex Coplan <alex.coplan@arm.com>
16031 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
16032 Avoid emitting CFA adjusts on the sp if we have the fp.
16034 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
16036 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
16037 (VMULD): New iterator.
16038 (VCOND): Handle V4HF and V8HF.
16039 (VCONQ): Fix entry for V2SF.
16040 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
16041 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
16042 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
16043 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
16044 Use a 128-bit vector mode for the indexed operand. Use stype for
16045 the scheduling type.
16047 2021-05-11 Richard Biener <rguenther@suse.de>
16049 * gimple-fold.c (maybe_fold_reference): Only return
16050 is_gimple_min_invariant values.
16052 2021-05-11 Richard Biener <rguenther@suse.de>
16054 PR middle-end/100509
16055 * gimple-fold.c (fold_gimple_assign): Only call
16056 get_symbol_constant_value on register type symbols.
16058 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16059 Joe Ramsay <joe.ramsay@arm.com>
16062 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
16063 (__arm_vcmpneq): Remove duplicate definition.
16064 (__arm_vstrwq_scatter_offset_p): Likewise.
16065 (__arm_vmaxq_x): Likewise.
16066 (__arm_vmlsdavaq): Likewise.
16067 (__arm_vmlsdavaxq): Likewise.
16068 (__arm_vmlsdavq_p): Likewise.
16069 (__arm_vmlsdavxq_p): Likewise.
16070 (__arm_vrmlaldavhaq): Likewise.
16071 (__arm_vstrbq_p): Likewise.
16072 (__arm_vstrbq_scatter_offset): Likewise.
16073 (__arm_vstrbq_scatter_offset_p): Likewise.
16074 (__arm_vstrdq_scatter_offset): Likewise.
16075 (__arm_vstrdq_scatter_offset_p): Likewise.
16076 (__arm_vstrdq_scatter_shifted_offset): Likewise.
16077 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
16079 2021-05-11 Jakub Jelinek <jakub@redhat.com>
16081 PR middle-end/100471
16082 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
16083 is 0, bypass the reduction loop including
16084 GOMP_taskgroup_reduction_unregister call.
16086 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16088 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
16089 costing_for_scalar.
16090 (rs6000_density_test): Early return if costing_for_scalar is true.
16091 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
16093 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16095 * doc/tm.texi: Regenerated.
16096 * target.def (init_cost): Add new parameter costing_for_scalar.
16097 * targhooks.c (default_init_cost): Adjust for new parameter.
16098 * targhooks.h (default_init_cost): Likewise.
16099 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
16100 (vect_compute_single_scalar_iteration_cost): Likewise.
16101 (vect_analyze_loop_2): Likewise.
16102 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
16103 (vect_bb_vectorization_profitable_p): Likewise.
16104 * tree-vectorizer.h (init_cost): Likewise.
16105 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
16106 * config/i386/i386.c (ix86_init_cost): Likewise.
16107 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
16109 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16111 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
16112 vect_nonmem and moved into...
16113 (struct rs6000_cost_data): ...here.
16114 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
16115 (rs6000_add_stmt_cost): Likewise.
16116 (rs6000_finish_cost): Likewise.
16118 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16120 * range-op.cc (get_bool_state): Adjust head comment.
16121 (operator_not_equal::op1_range): Fix comment.
16122 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16124 2021-05-10 Martin Sebor <msebor@redhat.com>
16126 PR middle-end/100425
16127 PR middle-end/100510
16128 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16129 (pass_walloca::xlimit_certain_p): ...to this.
16130 (pass_walloca::gate): Execute for any kind of handled warning.
16131 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16132 warnings when xlimit_certain_p is set.
16134 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16136 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16137 Return ALTIVEC_REGS if that is best_class.
16138 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16140 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16142 * config/arm/arm.h (CPP_SPEC): Remove error message about
16145 2021-05-10 Martin Jambor <mjambor@suse.cz>
16147 * ipa-prop.h (IPA_NODE_REF): Removed.
16148 (IPA_NODE_REF_GET_CREATE): Likewise.
16149 (IPA_EDGE_REF): Likewise.
16150 (IPA_EDGE_REF_GET_CREATE): Likewise.
16151 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16152 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16153 use of ipa_node_params_sum.
16154 (ipcp_versionable_function_p): Likewise.
16155 (push_node_to_stack): Likewise.
16156 (pop_node_from_stack): Likewise.
16157 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16158 direct use of ipa_node_params_sum.
16159 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16160 ipa_node_params_sum.
16161 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16163 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16164 use of ipa_node_params_sum.
16165 (self_recursively_generated_p): Likewise.
16166 (propagate_scalar_across_jump_function): Likewise.
16167 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16168 direct use of ipa_edge_args_sum, moved the lookup after the early
16169 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16170 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16171 direct uses of ipa_node_params_sum.
16172 (propagate_vr_across_jump_function): Likewise.
16173 (propagate_aggregate_lattice): Likewise.
16174 (propagate_aggs_across_jump_function): Likewise.
16175 (propagate_constants_across_call): Likewise, also replaced
16176 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16177 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16178 of ipa_node_params_sum.
16179 (estimate_local_effects): Likewise.
16180 (add_all_node_vals_to_toposort): Likewise.
16181 (propagate_constants_topo): Likewise.
16182 (ipcp_propagate_stage): Likewise.
16183 (ipcp_discover_new_direct_edges): Likewise.
16184 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16185 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16186 (get_info_about_necessary_edges): Likewise.
16187 (want_remove_some_param_p): Likewise.
16188 (create_specialized_node): Likewise.
16189 (self_recursive_pass_through_p): Likewise.
16190 (self_recursive_agg_pass_through_p): Likewise.
16191 (find_more_scalar_values_for_callers_subset): Likewise and also
16192 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16193 case replacing two of those with a single query.
16194 (find_more_contexts_for_caller_subset): Likewise for the
16195 ipa_polymorphic_call_context overload.
16196 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16197 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16198 ipa_node_params_sum.
16199 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16200 results of ipa_edge_args_sum->get.
16201 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16202 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16203 direct use of ipa_edge_args_sum.
16204 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16205 summary query after the early exit and reused the result later.
16206 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16207 ipa_node_params_sum.
16208 (decide_whether_version_node): Likewise. Removed re-querying for
16209 summaries after cloning.
16210 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16211 ipa_node_params_sum.
16212 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16214 (identify_dead_nodes): Likewise.
16215 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16216 ipa_node_params_sum.
16217 (ipcp_store_vr_results): Likewise.
16218 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16219 (ipa_fn_summary_t::duplicate): Likewise.
16220 (analyze_function_body): Likewise.
16221 (estimate_calls_size_and_time): Likewise.
16222 (ipa_cached_call_context::duplicate_from): Likewise.
16223 (ipa_call_context::equal_to): Likewise.
16224 (remap_edge_params): Likewise.
16225 (ipa_merge_fn_summary_after_inlining): Likewise.
16226 (inline_read_section): Likewise.
16227 * ipa-icf.c (sem_function::param_used_p): Likewise.
16228 * ipa-modref.c (compute_parm_map): Likewise.
16229 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16231 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16232 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16234 * ipa-profile.c (check_argument_count): Likewise.
16235 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
16236 with a direct use of ipa_node_params_sum.
16237 (ipa_initialize_node_params): Likewise.
16238 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
16239 direct use of ipa_edge_args_sum and reused the query result.
16240 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
16241 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16242 direct use of ipa_edge_args_sum.
16243 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
16244 ipa_node_params_sum and reused the result of the query.
16245 (ipa_analyze_node): Likewise.
16246 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
16247 of ipa_node_params_sum.
16248 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
16249 direct uses of ipa_edge_args_sum.
16250 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
16251 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16252 direct use of ipa_edge_args_sum. Removed superficial re-querying the
16254 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
16255 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16257 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
16258 direct use of ipa_edge_args_sum.
16259 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
16260 use of ipa_node_params_sum.
16261 (ipa_print_node_params): Likewise.
16262 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
16263 direct uses of ipa_edge_args_sum.
16264 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
16266 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16267 ipa_node_params_sum.
16268 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16269 scopes where it is used.
16271 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16273 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16274 <case E_V2SImode>: Force op_true to register.
16276 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16278 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16279 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16280 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16282 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16283 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16284 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16285 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16286 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16287 (mve_vcmpneq_n_f<mode>): Remove.
16288 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16289 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16290 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16292 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16294 * config/arm/iterators.md (MVE_COMPARISONS): New.
16296 (mve_cmp_type): New.
16297 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16299 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16300 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16301 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16302 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16303 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16304 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16305 (mve_vcmpneq_n_<mode>): Remove.
16307 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16309 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16310 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16311 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16314 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16316 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16317 (vcmpneq_n_u): Likewise.
16318 (vcmpeqq_u,): Likewise.
16319 (vcmpeqq_n_u): Likewise.
16320 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16321 VCMPEQQ_N_U and VCMPNEQ_N_U.
16322 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16323 (mve_vcmpeqq_n): Likewise.
16324 (mve_vcmpeqq): Likewise.
16325 (mve_vcmpneq_n): Likewise.
16327 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16329 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16330 the 's' version of the builtin.
16332 2021-05-10 Richard Biener <rguenther@suse.de>
16334 PR tree-optimization/100492
16335 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16336 Find nothing when the loop contains an irreducible region.
16338 2021-05-10 Richard Biener <rguenther@suse.de>
16340 PR middle-end/100464
16342 * gimple-fold.c (canonicalize_constructor_val): Do not set
16345 2021-05-10 Richard Biener <rguenther@suse.de>
16347 PR tree-optimization/100434
16348 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16350 (dse_optimize_stmt): Handle call LHS by dropping the
16351 LHS or the whole call if it doesn't have other
16353 (pass_dse::execute): Adjust.
16355 2021-05-10 Martin Liska <mliska@suse.cz>
16357 * Makefile.in: Add missing genversion rule.
16359 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16362 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16363 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16365 2021-05-10 Martin Liska <mliska@suse.cz>
16367 * builtins.c (is_builtin_name): Use startswith
16368 function instead of strncmp.
16369 * collect2.c (main): Likewise.
16370 (has_lto_section): Likewise.
16371 (scan_libraries): Likewise.
16372 * coverage.c (coverage_checksum_string): Likewise.
16373 (coverage_init): Likewise.
16374 * dwarf2out.c (is_cxx): Likewise.
16375 (gen_compile_unit_die): Likewise.
16376 * gcc-ar.c (main): Likewise.
16377 * gcc.c (init_spec): Likewise.
16378 (read_specs): Likewise.
16379 (execute): Likewise.
16380 (check_live_switch): Likewise.
16381 * genattrtab.c (write_attr_case): Likewise.
16382 (IS_ATTR_GROUP): Likewise.
16383 * gencfn-macros.c (main): Likewise.
16384 * gengtype.c (type_for_name): Likewise.
16385 (gen_rtx_next): Likewise.
16386 (get_file_langdir): Likewise.
16387 (write_local): Likewise.
16388 * genmatch.c (get_operator): Likewise.
16389 (get_operand_type): Likewise.
16390 (expr::gen_transform): Likewise.
16391 * genoutput.c (validate_optab_operands): Likewise.
16392 * incpath.c (add_sysroot_to_chain): Likewise.
16393 * langhooks.c (lang_GNU_C): Likewise.
16394 (lang_GNU_CXX): Likewise.
16395 (lang_GNU_Fortran): Likewise.
16396 (lang_GNU_OBJC): Likewise.
16397 * lto-wrapper.c (run_gcc): Likewise.
16398 * omp-general.c (omp_max_simt_vf): Likewise.
16399 * omp-low.c (omp_runtime_api_call): Likewise.
16400 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16401 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16402 * real.c (real_from_string): Likewise.
16403 * selftest.c (assert_str_startswith): Likewise.
16404 * timevar.c (timer::validate_phases): Likewise.
16405 * tree.c (get_file_function_name): Likewise.
16406 * ubsan.c (ubsan_use_new_style_p): Likewise.
16407 * varasm.c (default_function_rodata_section): Likewise.
16408 (incorporeal_function_p): Likewise.
16409 (default_section_type_flags): Likewise.
16410 * system.h (startswith): Define startswith.
16412 2021-05-10 Martin Liska <mliska@suse.cz>
16414 * bitmap.h (class auto_bitmap): Remove
16415 __cplusplus >= 201103.
16416 * config/aarch64/aarch64.c: Likewise.
16417 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16419 * sbitmap.h: Likewise.
16421 2021-05-10 Martin Liska <mliska@suse.cz>
16423 * Makefile.in: Rename gcov-iov to genversion and depend
16424 on version.h (instead of gcov-iov.h).
16425 * gcov-io.h: Include version.h instread of gcov-iov.h.
16426 * gengtype-state.c (read_state_version): Likewise.
16427 * gcov-iov.c: Moved to...
16428 * genversion.c: ...here.
16429 * lto-streamer.h (LTO_major_version): Define it with
16431 * version.c: Removed.
16432 * version.h: Removed.
16434 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16436 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16437 * config/arc/simdext.md (VCT): Add predicates for iterator
16440 (voptab): Likewise.
16441 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16442 (<voptab>v2si3): New patterns.
16444 (reduc_plus_scal_v4hi): Likewise.
16445 (reduc_plus_scal_v2si): Likewise.
16446 (vec_duplicatev2si): Likewise.
16447 (vec_duplicatev4hi): Likewise.
16449 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16451 * config/arc/simdext.md: Format and cleanup file.
16453 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16455 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16456 only when munaligned-access option is on.
16457 (movmisalign<mode>): Likewise.
16459 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16461 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16463 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16465 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16467 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16468 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16469 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16470 (norm_f): Likewise.
16473 (clzsi2): Use fls instruction when available.
16474 (arc_clzsi2): Likewise.
16476 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16478 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16480 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16482 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16484 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16486 PR middle-end/100467
16487 * toplev.c (compile_file): Call insn_locations_init before
16488 targetm.asm_out.code_end.
16490 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16493 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16495 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16497 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16498 Andrew Stubbs <amd@codesourcery.com>
16501 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16502 emit_move_insn operands.
16504 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16506 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16507 location for the outgoing edges of an empty block.
16508 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16509 type by means of the get_array_descr_info langhook, if it is set and
16510 returns true. Remove obsolete code dealing with unnamed subtypes.
16512 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16514 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16515 (sbr_vector): Renamed from ssa_block_cache.
16516 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16517 (ssa_block_ranges::~ssa_block_ranges): Remove.
16518 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16519 (ssa_block_ranges::set_bb_varying): Remove.
16520 (sbr_vector::get_bb_range): Adjust assert.
16521 (sbr_vector::bb_range_p): Adjust assert.
16522 (~block_range_cache): No freeing loop required.
16523 (block_range_cache::get_block_ranges): Remove.
16524 (block_range_cache::set_bb_range): Inline get_block_ranges.
16525 (block_range_cache::set_bb_varying): Remove.
16526 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16527 * value-range.h (irange_allocator::get_memory): New.
16529 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16531 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16532 dominator tree is available and requested.
16533 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16534 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16535 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16537 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16539 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16540 only PHI nodes better.
16542 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16544 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16546 (gcond_edge_range): Export prototype.
16547 * gimple-range-edge.cc (gcond_edge_range): New.
16548 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16549 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16551 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16553 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16554 default range into a temp and allocate only what is needed.
16556 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16558 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16560 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16562 * gimple-range.h (gimple_range_global): Pick up parameter initial
16563 values, and use-before defined locals are UNDEFINED.
16565 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16567 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16569 * tree.h (reverse_storage_order_for_component_p): Return false if
16570 the type is a pointer.
16572 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16574 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16576 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16579 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16580 Handle V8QI, V4HI and V2SI modes.
16581 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16582 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16583 * config/i386/i386.md (unspec): ... here.
16585 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16586 Tom de Vries <tdevries@suse.de>
16588 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16589 a truth_value_p reduction variable is nonintegral.
16591 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16594 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16595 Return false for mode sizes < 16.
16597 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16600 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
16602 2021-05-06 Martin Jambor <mjambor@suse.cz>
16604 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
16605 when there is no function summary.
16606 (ipa_sra_summarize_function): produce edge summaries even when
16609 2021-05-06 Tom Tromey <tom@tromey.com>
16611 * godump.c (string_hash_eq): Remove.
16612 (go_finish): Use htab_eq_string.
16614 2021-05-06 Tom Tromey <tom@tromey.com>
16616 * gengtype-state.c (read_state): Use htab_eq_string.
16617 (string_eq): Remove.
16619 2021-05-06 Tom Tromey <tom@tromey.com>
16621 * gensupport.c (htab_eq_string): Remove.
16623 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16626 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
16627 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
16628 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
16629 set_ignored_loc callbacks.
16630 * debug.c (do_nothing_debug_hooks): Likewise.
16631 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
16632 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
16633 (in_text_section_p, last_text_label, last_cold_label,
16634 switch_text_ranges, switch_cold_ranges): New data items.
16635 (dwarf2out_note_section_used): Remove.
16636 (dwarf2out_begin_prologue): Set fde->ignored_debug and
16638 (mark_ignored_debug_section): New helper function.
16639 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
16640 mark_ignored_debug_section.
16641 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
16642 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
16643 (size_of_aranges): Adjust formula for multi-part text ranges size.
16644 (output_aranges): Output multi-part text ranges.
16645 (dwarf2out_set_ignored_loc): New callback function.
16646 (dwarf2out_finish): Output multi-part text ranges.
16647 (dwarf2out_c_finalize): Clear new data items.
16648 * final.c (final_start_function_1): Call set_ignored_loc callback.
16649 (final_scan_insn_1): Likewise.
16650 * ggc-page.c (gt_ggc_mx): New helper function.
16651 * stringpool.c (gt_pch_nx): Likewise.
16653 2021-05-06 Richard Biener <rguenther@suse.de>
16655 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
16656 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
16657 (TV_TREE_INTO_SSA): New.
16658 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
16659 (rewrite_blocks): Likewise.
16660 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
16662 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16664 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
16665 abs_replacement, xor_replacement,
16666 cond_removal_in_popcount_clz_ctz_pattern,
16667 replace_phi_edge_with_variable): Change type of phi argument from
16668 gimple * to gphi *.
16670 2021-05-06 Richard Biener <rguenther@suse.de>
16672 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
16673 Output an opt-info message.
16674 (do_split_loop_on_cond): Likewise.
16675 (tree_ssa_split_loops): Update SSA form here.
16677 2021-05-06 Richard Biener <rguenther@suse.de>
16679 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
16680 return variable removal.
16682 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
16684 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
16685 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
16686 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
16688 * config/s390/s390.c (s390_const_operand_ok): Remove unused
16691 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16693 PR tree-optimization/94589
16694 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
16695 spaceship_replacement.
16696 (cond_only_block_p, spaceship_replacement): New functions.
16698 2021-05-06 Richard Biener <rguenther@suse.de>
16701 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
16702 dealing with a debug use and only query existing addresses
16704 (lower_emutls_1): Avoid splitting out addresses for debug
16705 stmts, reset the debug stmt when we fail to find existing
16707 (lower_emutls_phi_arg): Set wi.stmt.
16709 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
16712 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
16713 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
16714 (stack_protect_test): Use cbranch helper.
16716 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16719 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
16720 always return the establisher frame for __builtin_frame_address (0).
16722 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
16725 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
16726 (ix86_cpu_model_var): Likewise.
16727 (ix86_cpu_features2_type_node): Likewise.
16728 (ix86_cpu_features2_var): Likewise.
16729 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
16732 2021-05-05 Martin Sebor <msebor@redhat.com>
16734 * passes.def (pass_warn_printf): Run after SSA.
16736 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16738 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
16739 * config/arm/predicates.md (minus_one_operand): New predicate.
16741 2021-05-05 Jeff Law <jlaw@tachyum.com>
16743 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
16745 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
16747 PR rtl-optimization/100263
16748 * postreload.c (move2add_valid_value_p): Ensure register can
16751 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16753 PR rtl-optimization/100411
16754 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
16755 and beginning of function markers.
16757 2021-05-05 Jeff Law <jlaw@tachyum.com>
16759 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
16760 * config/cr16/cr16.c (notice_update_cc): Remove.
16761 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
16763 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
16766 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
16767 Handle V8QI, V4HI and V2SI modes.
16768 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
16769 (ix86_build_signbit_mask): Ditto.
16770 * config/i386/mmx.md (MMXMODE14): New mode iterator.
16771 (<smaxmin:code><MMXMODE14:mode>3): New expander.
16772 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
16773 (<umaxmin:code><MMXMODE24:mode>3): New expander.
16774 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
16775 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
16776 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16777 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16778 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16779 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16781 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16783 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
16784 not expand the VALUE_EXPR of variables put in the non-local frame.
16785 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
16786 to be ignored for debug info, ensure its variable offsets are not.
16788 2021-05-05 Richard Biener <rguenther@suse.de>
16790 PR tree-optimization/79333
16791 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
16792 Fold stmt following SSA edges.
16794 2021-05-05 Richard Biener <rguenther@suse.de>
16796 PR middle-end/100394
16797 * calls.c (expand_call): Preserve possibly throwing calls.
16798 * cfgexpand.c (expand_call_stmt): When a call can throw signal
16799 RTL expansion there are side-effects.
16800 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
16801 mark all possibly throwing stmts necessary unless we can elide
16803 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
16804 -fdelete-dead-exceptions.
16805 * tree.h (DECL_PURE_P): Add note about exceptions.
16807 2021-05-05 Alexandre Oliva <oliva@adacore.com>
16809 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
16812 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
16814 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
16815 get_fnname_from_decl for name of thunk.
16816 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
16817 and ASM_OUTPUT_LABEL.
16818 (rs6000_xcoff_declare_function_name): Use assemble_name and
16820 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
16821 (rs6000_xcoff_encode_section_info): Don't add mapping class
16822 for aliases. Always add [DS] mapping class to primary
16824 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
16826 2021-05-04 Martin Sebor <msebor@redhat.com>
16828 PR middle-end/100307
16829 * builtins.c (compute_objsize_r): Clear base0 for pointers.
16831 2021-05-04 Jeff Law <jlaw@tachyum.com>
16833 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
16835 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16837 * caller-save.c: Remove CC0.
16838 * cfgcleanup.c: Remove CC0.
16839 * cfgrtl.c: Remove CC0.
16840 * combine.c: Remove CC0.
16841 * compare-elim.c: Remove CC0.
16842 * conditions.h: Remove CC0.
16843 * config/h8300/h8300.h: Remove CC0.
16844 * config/h8300/h8300-protos.h: Remove CC0.
16845 * config/h8300/peepholes.md: Remove CC0.
16846 * config/i386/x86-tune-sched.c: Remove CC0.
16847 * config/m68k/m68k.c: Remove CC0.
16848 * config/rl78/rl78.c: Remove CC0.
16849 * config/sparc/sparc.c: Remove CC0.
16850 * config/xtensa/xtensa.c: Remove CC0.
16851 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
16852 RTL where that is used as a placeholder only.
16853 * cprop.c: Remove CC0.
16854 * cse.c: Remove CC0.
16855 * cselib.c: Remove CC0.
16856 * df-problems.c: Remove CC0.
16857 * df-scan.c: Remove CC0.
16858 * doc/md.texi: Remove CC0. Adjust an example.
16859 * doc/rtl.texi: Remove CC0. Adjust an example.
16860 * doc/tm.texi: Regenerate.
16861 * doc/tm.texi.in: Remove CC0.
16862 * emit-rtl.c: Remove CC0.
16863 * final.c: Remove CC0.
16864 * fwprop.c: Remove CC0.
16865 * gcse-common.c: Remove CC0.
16866 * gcse.c: Remove CC0.
16867 * genattrtab.c: Remove CC0.
16868 * genconfig.c: Remove CC0.
16869 * genemit.c: Remove CC0.
16870 * genextract.c: Remove CC0.
16871 * gengenrtl.c: Remove CC0.
16872 * genrecog.c: Remove CC0.
16873 * haifa-sched.c: Remove CC0.
16874 * ifcvt.c: Remove CC0.
16875 * ira-costs.c: Remove CC0.
16876 * ira.c: Remove CC0.
16877 * jump.c: Remove CC0.
16878 * loop-invariant.c: Remove CC0.
16879 * lra-constraints.c: Remove CC0.
16880 * lra-eliminations.c: Remove CC0.
16881 * optabs.c: Remove CC0.
16882 * postreload-gcse.c: Remove CC0.
16883 * postreload.c: Remove CC0.
16884 * print-rtl.c: Remove CC0.
16885 * read-rtl-function.c: Remove CC0.
16886 * reg-notes.def: Remove CC0.
16887 * reg-stack.c: Remove CC0.
16888 * reginfo.c: Remove CC0.
16889 * regrename.c: Remove CC0.
16890 * reload.c: Remove CC0.
16891 * reload1.c: Remove CC0.
16892 * reorg.c: Remove CC0.
16893 * resource.c: Remove CC0.
16894 * rtl.c: Remove CC0.
16895 * rtl.def: Remove CC0.
16896 * rtl.h: Remove CC0.
16897 * rtlanal.c: Remove CC0.
16898 * sched-deps.c: Remove CC0.
16899 * sched-rgn.c: Remove CC0.
16900 * shrink-wrap.c: Remove CC0.
16901 * simplify-rtx.c: Remove CC0.
16902 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
16903 CC_STATUS_MDEP, and CC_STATUS.
16904 * target.def: Remove CC0.
16905 * valtrack.c: Remove CC0.
16906 * var-tracking.c: Remove CC0.
16908 2021-05-04 Richard Biener <rguenther@suse.de>
16910 PR tree-optimization/100414
16911 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
16913 (tree_ssa_phiopt_worker): But unconditionally here.
16915 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
16917 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
16918 && and || with floating-point and complex arguments.
16920 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
16922 * tree-inline.c (insert_debug_decl_map): Delete.
16923 (copy_debug_stmt): Minor tweak.
16924 (setup_one_parameter): Do not use a variable if the value is either
16925 a read-only DECL or a non-addressable local variable in the caller.
16926 In this case, insert the debug-only variable in the map manually.
16927 (expand_call_inline): Do not generate a CLOBBER for these values.
16928 * tree-inline.h (debug_map): Minor tweak.
16930 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
16932 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
16933 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
16935 2021-05-04 Richard Biener <rguenther@suse.de>
16937 PR tree-optimization/100329
16938 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
16940 (insert_stmt_after): Assert we're not running into asm goto.
16942 2021-05-04 Richard Biener <rguenther@suse.de>
16944 PR tree-optimization/100398
16945 * tree-ssa-dse.c (pass_dse::execute): Preserve control
16948 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16950 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
16952 2021-05-04 Alexandre Oliva <oliva@adacore.com>
16954 * builtins.c (try_store_by_multiple_pieces): New.
16955 (expand_builtin_memset_args): Use it. If target_char_cast
16956 fails, proceed as for non-constant val. Pass len's ctz to...
16957 * expr.c (clear_storage_hints): ... this. Try store by
16958 multiple pieces after setmem.
16959 (clear_storage): Adjust.
16960 * expr.h (clear_storage_hints): Likewise.
16961 (try_store_by_multiple_pieces): Declare.
16962 * passes.def: Replace the last copy_prop with ccp.
16964 2021-05-03 Tom de Vries <tdevries@suse.de>
16967 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
16970 2021-05-03 Richard Biener <rguenther@suse.de>
16972 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
16974 2021-05-03 Richard Biener <rguenther@suse.de>
16976 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
16977 (dse_dom_walker): Remove.
16978 (dse_dom_walker::dse_optimize_stmt): Rename...
16979 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
16980 (dse_dom_walker::before_dom_children): Inline ...
16981 (pass_dse::execute): ... here. Perform a reverse program
16984 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
16987 * configure: Regenerated.
16989 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
16992 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
16993 (s390_md_asm_adjust): Handle hard registers.
16995 2021-05-03 Jakub Jelinek <jakub@redhat.com>
16997 PR tree-optimization/100382
16998 * tree-ssa-dse.c: Include tree-eh.h.
16999 (dse_dom_walker::before_dom_children): Don't remove stmts if
17000 stmt_unremovable_because_of_non_call_eh_p is true.
17002 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
17004 * varasm.c (compute_reloc_for_var): Split out from...
17005 (get_variable_section): Use it.
17006 * output.h (compute_reloc_for_var): Declare.
17007 * config/rs6000/rs6000-protos.h
17008 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
17010 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
17011 Don't append storage mapping class to symbol.
17012 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
17013 Don't convert TLS BSS to common.
17014 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
17015 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
17017 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
17019 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
17020 If align is 0 from TLS class, use the same rules as varasm.c
17021 If not common, switch to BSS section manually.
17022 If common, emit appropriate comm or lcomm directive.
17023 (rs6000_xcoff_encode_section_info): Add logic to append all
17024 storage mapping classes.
17025 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
17026 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17027 rs6000_xcoff_asm_output_aligned_decl_common.
17028 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17029 rs6000_xcoff_asm_output_aligned_decl_common.
17030 (ASM_OUTPUT_TLS_COMMON): Use
17031 rs6000_xcoff_asm_output_aligned_decl_common.
17033 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17036 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
17037 as first argument of pseudo_node_t constructors.
17039 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17042 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
17044 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
17046 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
17047 (gt_pch_nx (int_range<1> *&)): New.
17048 (gt_ggc_mx (int_range<1> *&)): New.
17049 * value-range.h (class irange): Add GTY support for
17052 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
17054 * doc/options.texi (Negative): Change either or to both and.
17056 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17058 * config/aarch64/aarch64-simd-builtins.def: Add
17059 float_ml[as][q]_laneq builtin generator macros.
17060 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
17061 (aarch64_float_mla_laneq<mode>): Define.
17062 (aarch64_float_mls_laneq<mode>): Define.
17063 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
17064 instead of GCC vector extensions.
17065 (vmlaq_laneq_f32): Likewise.
17066 (vmls_laneq_f32): Likewise.
17067 (vmlsq_laneq_f32): Likewise.
17069 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17071 * config/aarch64/aarch64-simd-builtins.def: Add
17072 float_ml[as]_lane builtin generator macros.
17073 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
17075 (mul_lane<mode>3): This, and re-order arguments.
17076 (aarch64_float_mla_lane<mode>): Define.
17077 (aarch64_float_mls_lane<mode>): Define.
17078 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
17079 instead of GCC vector extensions.
17080 (vmlaq_lane_f32): Likewise.
17081 (vmls_lane_f32): Likewise.
17082 (vmlsq_lane_f32): Likewise.
17084 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17086 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
17087 builtin generator macros.
17088 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
17090 (aarch64_float_mls<mode>): Define.
17091 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
17092 instead of relying on GCC vector extensions.
17093 (vmla_f64): Likewise.
17094 (vmlaq_f32): Likewise.
17095 (vmlaq_f64): Likewise.
17096 (vmls_f32): Likewise.
17097 (vmls_f64): Likewise.
17098 (vmlsq_f32): Likewise.
17099 (vmlsq_f64): Likewise.
17100 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
17102 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17104 * config/aarch64/aarch64-simd-builtins.def: Add
17105 float_ml[as]_n_builtin generator macros.
17106 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
17108 (mul_n<mode>3): This, and re-order arguments.
17109 (aarch64_float_mla_n<mode>): Define.
17110 (aarch64_float_mls_n<mode>): Define.
17111 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
17112 instead of inline asm.
17113 (vmlaq_n_f32): Likewise.
17114 (vmls_n_f32): Likewise.
17115 (vmlsq_n_f32): Likewise.
17117 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17119 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17120 builtin generator macros.
17121 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17122 (aarch64_pmull_hiv16qi_insn): Define.
17123 (aarch64_pmull_hiv16qi): Define.
17124 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17125 instead of inline asm.
17126 (vmull_p8): Likewise.
17128 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17130 * config/avr/avr.md: Adjust peepholes to match and
17131 generate parallels with clobber of REG_CC.
17132 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17133 (*mov<mode>_insn): Rename to mov<mode>_insn.
17135 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17137 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17138 for VAR_DECLs if -fdata-sections enabled.
17140 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17142 PR bootstrap/100327
17143 * config/rs6000/rs6000.c
17144 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17145 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17147 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17149 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17151 (class back_threader_registry): ...to here...
17152 (class back_threader_profitability): ...and here...
17153 (thread_jumps::thread_through_all_blocks): Remove argument.
17154 (back_threader_registry::back_threader_registry): New.
17155 (back_threader_registry::~back_threader_registry): New.
17156 (back_threader_registry::thread_through_all_blocks): New.
17157 (thread_jumps::profitable_jump_thread_path): Move from here...
17158 (back_threader_profitability::profitable_path_p): ...to here.
17159 (thread_jumps::find_taken_edge): New.
17160 (thread_jumps::convert_and_register_current_path): Move...
17161 (back_threader_registry::register_path): ...to here.
17162 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17163 (thread_jumps::maybe_register_path): ...to here.
17164 (thread_jumps::handle_phi): Call find_taken_edge and
17165 maybe_register_path.
17166 (thread_jumps::handle_assignment): Same.
17167 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17168 tree argument to handle_phi and handle_assignment.
17169 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17170 set of m_speed_p and m_max_threaded_paths.
17171 (pass_thread_jumps::execute): Remove second argument from
17172 find_jump_threads_backwards.
17173 (pass_early_thread_jumps::execute): Same.
17175 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17177 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17178 (class dom_opt_dom_walker): Initialize some class variables.
17179 (pass_dominator::execute): Pass evrp_range_analyzer and
17180 dom_jump_threader_simplifier to dom_opt_dom_walker.
17181 Adjust for some functions moving into classes.
17182 (simplify_stmt_for_jump_threading): Adjust and move to...
17183 (jump_threader_simplifier::simplify): ...here.
17184 (dom_opt_dom_walker::before_dom_children): Adjust for
17185 m_evrp_range_analyzer.
17186 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17187 (test_for_singularity): Place in dom_opt_dom_walker class.
17188 (dom_opt_dom_walker::optimize_stmt): The argument
17189 evrp_range_analyzer is now a class field.
17190 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17191 (thread_jumps::thread_through_all_blocks): New.
17192 (thread_jumps::convert_and_register_current_path): Use m_registry.
17193 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17194 being in the threader class.
17195 (pass_early_thread_jumps::execute): Same.
17196 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17197 (jump_threader::jump_threader): ...here.
17198 (threadedge_finalize_values): Move...
17199 (jump_threader::~jump_threader): ...here.
17200 (jump_threader::remove_jump_threads_including): New.
17201 (jump_threader::thread_through_all_blocks): New.
17202 (record_temporary_equivalences_from_phis): Move...
17203 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17204 (record_temporary_equivalences_from_stmts_at_dest): Move...
17205 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17207 (simplify_control_stmt_condition_1): Move to jump_threader class.
17208 (simplify_control_stmt_condition): Move...
17209 (jump_threader::simplify_control_stmt_condition): ...here.
17210 (thread_around_empty_blocks): Move...
17211 (jump_threader::thread_around_empty_blocks): ...here.
17212 (thread_through_normal_block): Move...
17213 (jump_threader::thread_through_normal_block): ...here.
17214 (thread_across_edge): Move...
17215 (jump_threader::thread_across_edge): ...here.
17216 (thread_outgoing_edges): Move...
17217 (jump_threader::thread_outgoing_edges): ...here.
17218 * tree-ssa-threadedge.h: Move externally facing functings...
17219 (class jump_threader): ...here...
17220 (class jump_threader_simplifier): ...and here.
17221 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17222 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17223 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17224 (jump_thread_path_allocator::allocate_thread_edge): New.
17225 (jump_thread_path_allocator::allocate_thread_path): New.
17226 (jump_thread_path_registry::jump_thread_path_registry): New.
17227 (jump_thread_path_registry::~jump_thread_path_registry): New.
17228 (jump_thread_path_registry::allocate_thread_edge): New.
17229 (jump_thread_path_registry::allocate_thread_path): New.
17230 (dump_jump_thread_path): Make extern.
17231 (debug (const vec<jump_thread_edge *> &path)): New.
17232 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17233 (struct thread_stats_d): Remove.
17234 (remove_ctrl_stmt_and_useless_edges): Make static.
17235 (lookup_redirection_data): Move...
17236 (jump_thread_path_registry::lookup_redirection_data): ...here.
17237 (ssa_redirect_edges): Make static.
17238 (thread_block_1): Move...
17239 (jump_thread_path_registry::thread_block_1): ...here.
17240 (thread_block): Move...
17241 (jump_thread_path_registry::thread_block): ...here.
17242 (thread_through_loop_header): Move...
17243 (jump_thread_path_registry::thread_through_loop_header): ...here.
17244 (mark_threaded_blocks): Move...
17245 (jump_thread_path_registry::mark_threaded_blocks): ...here.
17246 (debug_path): Move...
17247 (jump_thread_path_registry::debug_path): ...here.
17248 (debug_all_paths): Move...
17249 (jump_thread_path_registry::dump): ..here.
17250 (rewire_first_differing_edge): Move...
17251 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
17252 (adjust_paths_after_duplication): Move...
17253 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
17254 (duplicate_thread_path): Move...
17255 (jump_thread_path_registry::duplicate_thread_path): ..here.
17256 (remove_jump_threads_including): Move...
17257 (jump_thread_path_registry::remove_jump_threads_including): ...here.
17258 (thread_through_all_blocks): Move to...
17259 (jump_thread_path_registry::thread_through_all_blocks): ...here.
17260 (delete_jump_thread_path): Remove.
17261 (register_jump_thread): Move...
17262 (jump_thread_path_registry::register_jump_thread): ...here.
17263 * tree-ssa-threadupdate.h: Move externally facing functions...
17264 (class jump_thread_path_allocator): ...here...
17265 (class jump_thread_path_registry): ...and here.
17266 (thread_through_all_blocks): Remove.
17267 (struct removed_edges): New.
17268 (register_jump_thread): Remove.
17269 (remove_jump_threads_including): Remove.
17270 (delete_jump_thread_path): Remove.
17271 (remove_ctrl_stmt_and_useless_edges): Remove.
17272 (free_dom_edge_info): New prototype.
17273 * tree-vrp.c: Remove x_vr_values hack.
17274 (class vrp_jump_threader_simplifier): New.
17275 (vrp_jump_threader_simplifier::simplify): New.
17276 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17277 Remove m_dummy_cond.
17278 Instantiate m_simplifier and m_threader.
17279 (vrp_jump_threader::thread_through_all_blocks): New.
17280 (vrp_jump_threader::simplify_stmt): Remove.
17281 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17282 Remove x_vr_values hack.
17283 (execute_vrp): Adjust for thread_through_all_blocks being in a
17286 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17288 * genflags.c (gen_insn): Print failed expansion string.
17290 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17292 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17293 without limit to MAX_FIXED_MODE_SIZE.
17295 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17297 PR middle-end/90773
17298 * builtins.c (builtin_memset_gen_str): Don't use return from
17299 simplify_gen_subreg.
17301 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17304 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17305 (*addsi3_carry_zext_0r): Ditto.
17306 (*sub<mode>3_carry_0): Ditto.
17307 (*subsi3_carry_zext_0r): Ditto.
17308 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17310 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17311 Also consider ix86_carry_flag_unset_operator to calculate
17312 the cost of adc/sbb insn.
17314 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17316 PR rtl-optimization/100225
17317 PR rtl-optimization/84878
17318 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17319 where we have an instruction which touches (writes) any hard
17320 register from df->regular_block_artificial_uses set.
17321 Allow not-single-set instruction only right before basic block
17324 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17326 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17328 2021-04-30 LevyHsu <admin@levyhsu.com>
17330 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17331 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17332 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17333 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17335 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17337 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17339 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17341 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17342 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17343 * config/i386/i386.c: Adjust.
17344 * config/i386/i386.md: Adjust.
17345 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17346 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17347 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17348 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17349 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17350 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17351 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17352 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17353 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17354 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17355 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17357 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17359 * config/i386/i386-expand.c (ix86_expand_int_compare):
17360 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17361 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17362 predicate to allow more combine opportunities with memory operands.
17363 (*sub<mode>3_carry_0): Ditto.
17365 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17367 PR rtl-optimization/100303
17368 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17369 boolean that indicates whether the use will only be used in
17370 debug instructions. Treat it in the same way that existing
17371 cross-EBB debug references would be handled if so.
17372 (function_info::make_uses_available): Likewise.
17373 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17374 prototype accordingly.
17375 (function_info::make_uses_available): Likewise.
17376 * fwprop.c (try_fwprop_subst): Update call accordingly.
17378 2021-04-29 Jeff Law <jlaw@tachyum.com>
17380 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17383 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17384 Richard Biener <rguenther@suse.de>
17387 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17388 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17389 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17390 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17391 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17392 to PURE_ARGS category.
17393 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17394 Handle PURE_ARGS category.
17395 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17397 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17399 * configure.ac: Check for the presence of sys/locking.h header and
17400 for whether _LK_LOCK is supported by _locking.
17401 * configure: Regenerate.
17402 * config.in: Likewise.
17403 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17404 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17405 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17407 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17409 * config/i386/predicates.md (fcmov_comparison_operator):
17410 Do not check for trivial FP comparison operator.
17411 <case GEU, case LTU>: Allow CCGZmode.
17412 <case GTU, case LEU>: Do not allow CCCmode.
17413 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17414 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17415 Do not check for trivial FP comparison operator. Allow CCGZmode.
17417 2021-04-29 Tom de Vries <tdevries@suse.de>
17419 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17420 fd->loop.step by either step or orig_step.
17422 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17424 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17425 (load_got_register): Do the PIC dance here.
17426 (sparc_legitimize_tls_address): Simplify.
17427 (sparc_emit_probe_stack_range): Likewise.
17428 (sparc32_initialize_trampoline): Likewise.
17429 (sparc64_initialize_trampoline): Likewise.
17430 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17431 (probe_stack_range<P:mode>): Likewise.
17432 (flush<P:mode>): Likewise.
17433 (tgd_hi22<P:mode>): Likewise.
17434 (tgd_lo10<P:mode>): Likewise.
17435 (tgd_add<P:mode>): Likewise.
17436 (tgd_call<P:mode>): Likewise.
17437 (tldm_hi22<P:mode>): Likewise.
17438 (tldm_lo10<P:mode>): Likewise.
17439 (tldm_add<P:mode>): Likewise.
17440 (tldm_call<P:mode>): Likewise.
17441 (tldo_hix22<P:mode>): Likewise.
17442 (tldo_lox10<P:mode>): Likewise.
17443 (tldo_add<P:mode>): Likewise.
17444 (tie_hi22<P:mode>): Likewise.
17445 (tie_lo10<P:mode>): Likewise.
17446 (tie_add<P:mode>): Likewise.
17447 (tle_hix22<P:mode>): Likewise.
17448 (tle_lox10<P:mode>): Likewise.
17449 (stack_protect_setsi): Rename to...
17450 (stack_protect_set32): ...this.
17451 (stack_protect_setdi): Rename to...
17452 (stack_protect_set64): ...this.
17453 (stack_protect_set): Adjust calls to above.
17454 (stack_protect_testsi): Rename to...
17455 (stack_protect_test32): ...this.
17456 (stack_protect_testdi): Rename to...
17457 (stack_protect_test64): ...this.
17458 (stack_protect_test): Adjust calls to above.
17460 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17462 PR middle-end/90773
17463 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17464 (builtin_strncpy_read_str): Likewise.
17465 (builtin_memset_read_str): Add an argument for the previous RTL
17466 information and generate the new RTL from the previous RTL info.
17467 (builtin_memset_gen_str): Likewise.
17468 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17469 (builtin_memset_read_str): Likewise.
17470 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17471 returns true, round up size and alignment to the widest integer
17472 mode for maximum size.
17473 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17474 and pass it to m_constfn.
17475 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17476 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17477 initialize m_push. Initialize m_overlap_op_by_pieces with
17478 targetm.overlap_op_by_pieces_p ().
17479 (op_by_pieces_d::run): Pass the previous RTL information to
17480 pieces_addr::adjust and generate overlapping operations if
17481 m_overlap_op_by_pieces is true.
17483 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17485 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17487 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17488 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17489 for op_by_pieces_d change.
17490 (clear_by_pieces_1): Add a dummy argument.
17491 (clear_by_pieces): Updated for op_by_pieces_d change.
17492 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17493 (string_cst_read_str): Add a dummy argument.
17494 * expr.h (by_pieces_constfn): Add a dummy argument.
17495 (by_pieces_prev): New.
17496 * target.def (overlap_op_by_pieces_p): New target hook.
17497 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17498 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17499 * doc/tm.texi: Regenerated.
17501 2021-04-29 Richard Biener <rguenther@suse.de>
17503 PR tree-optimization/100253
17504 * tree-vect-stmts.c (vectorizable_load): Do not assume
17505 element alignment when DR_MISALIGNMENT is -1.
17506 (vectorizable_store): Likewise.
17508 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17511 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17512 absu_hwi instead of abs_hwi.
17514 2021-04-29 Richard Biener <rguenther@suse.de>
17516 PR middle-end/38474
17517 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17518 forwarding when indirect forwarding through ESCAPED
17521 2021-04-29 Tom de Vries <tdevries@suse.de>
17524 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17525 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17526 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17527 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17529 2021-04-29 Richard Biener <rguenther@suse.de>
17531 PR tree-optimization/99912
17532 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17533 (dse_dom_walker::todo): Likewise.
17534 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17536 (dse_dom_walker::before_dom_children): Remove trivially
17537 dead SSA defs and schedule CFG cleanup if we removed all
17539 (pass_dse::execute): Get TODO as computed by the DOM walker
17540 and return it. Wipe dominator info earlier.
17542 2021-04-29 Richard Biener <rguenther@suse.de>
17545 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17546 Track blocks to cleanup EH in new m_need_eh_cleanup.
17547 (ipcp_modif_dom_walker::cleanup_eh): New.
17548 (ipcp_transform_function): Release dominator info before
17551 2021-04-29 Martin Sebor <msebor@redhat.com>
17553 PR middle-end/100250
17554 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17555 a pointer when it's null.
17557 2021-04-29 Martin Sebor <msebor@redhat.com>
17559 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17560 * ipa-free-lang-data.cc: New file.
17561 * tree.c: Move pass free_lang_data to file above.
17562 (build_array_type_1): Declare extern.
17563 * tree.h (build_array_type_1): Declare.
17565 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17567 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17568 make consistent with updated RTL pattern.
17569 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17570 Implement using ss_truncate and us_truncate rather than
17572 * config/aarch64/iterators.md: Remove redundant unspecs and
17573 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17575 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17577 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17578 attributes consistent with those defined in arm_neon.h.
17580 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17582 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17583 attributes consistent with those defined in arm_neon.h.
17585 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17587 * config/aarch64/aarch64-simd-builtins.def: Add
17588 float_trunc_rodd builtin generator macros.
17589 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17591 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17592 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17593 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17594 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17595 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17596 instead of inline asm.
17597 (vcvtx_high_f32_f64): Likewise.
17598 (vcvtxd_f32_f64): Likewise.
17599 * config/aarch64/iterators.md: Add FCVTXN unspec.
17601 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17603 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
17605 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
17607 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
17608 instead of inline asm.
17609 (vqtbx1_u8): Likewise.
17610 (vqtbx1_p8): Likewise.
17611 (vqtbx1q_s8): Likewise.
17612 (vqtbx1q_u8): Likewise.
17613 (vqtbx1q_p8): Likewise.
17614 (vtbx2_s8): Likewise.
17615 (vtbx2_u8): Likewise.
17616 (vtbx2_p8): Likewise.
17618 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17620 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
17622 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
17623 instead of inline asm.
17624 (vqtbl1_s8): Likewise.
17625 (vqtbl1_u8): Likewise.
17626 (vqtbl1q_p8): Likewise.
17627 (vqtbl1q_s8): Likewise.
17628 (vqtbl1q_u8): Likewise.
17629 (vtbl1_s8): Likewise.
17630 (vtbl1_u8): Likewise.
17631 (vtbl1_p8): Likewise.
17632 (vtbl2_s8): Likewise.
17633 (vtbl2_u8): Likewise.
17634 (vtbl2_p8): Likewise.
17636 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17638 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
17639 ssri_n buitin generator macro.
17640 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
17641 instead of inline asm.
17642 (vsri_n_p16): Likewise.
17643 (vsri_n_p64): Likewise.
17644 (vsriq_n_p8): Likewise.
17645 (vsriq_n_p16): Likewise.
17646 (vsriq_n_p64): Likewise.
17648 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17650 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
17651 iterator for polynomial ssli_n builtin generator macro.
17652 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
17653 instead of inline asm.
17654 (vsli_n_p16): Likewise.
17655 (vsliq_n_p8): Likewise.
17656 (vsliq_n_p16): Likewise.
17657 * config/aarch64/iterators.md: Define VALLP mode iterator.
17659 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17661 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
17662 iterator to generate [su]adalp RTL builtins.
17663 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
17664 [su]adalp RTL pattern.
17665 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
17666 instead of inline asm.
17667 (vpadal_u32): Likewise.
17669 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17671 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
17672 builtin generator macros.
17673 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
17675 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
17676 instead of inline asm.
17677 (vpaddl_s16): Likewise.
17678 (vpaddl_s32): Likewise.
17679 (vpaddl_u8): Likewise.
17680 (vpaddl_u16): Likewise.
17681 (vpaddl_u32): Likewise.
17682 (vpaddlq_s8): Likewise.
17683 (vpaddlq_s16): Likewise.
17684 (vpaddlq_s32): Likewise.
17685 (vpaddlq_u8): Likewise.
17686 (vpaddlq_u16): Likewise.
17687 (vpaddlq_u32): Liwewise.
17688 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
17689 appropriate attributes.
17691 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17693 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
17694 for aarch64_addp<mode> builtin macro generator.
17695 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
17696 aarch64_addp<mode> RTL pattern.
17697 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
17698 instead of inline asm.
17699 (vpaddq_s16): Likewise.
17700 (vpaddq_s32): Likewise.
17701 (vpaddq_s64): Likewise.
17702 (vpaddq_u8): Likewise.
17703 (vpaddq_u16): Likewise.
17704 (vpaddq_u32): Likewise.
17705 (vpaddq_u64): Likewise.
17707 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17709 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
17710 builtin generator macros.
17711 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
17713 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
17714 instead of inline asm.
17715 (vqdmulh_n_s32): Likewise.
17716 (vqdmulhq_n_s16): Likewise.
17717 (vqdmulhq_n_s32): Likewise.
17718 (vqrdmulh_n_s16): Likewise.
17719 (vqrdmulh_n_s32): Likewise.
17720 (vqrdmulhq_n_s16): Likewise.
17721 (vqrdmulhq_n_s32): Likewise.
17723 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
17725 * doc/install.texi (--enable-offload-defaulted): Document.
17727 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17729 * config/avr/avr-dimode.md: Turn existing patterns into
17730 define_insn_and_split style patterns where the splitter
17731 adds a clobber of the condition code register. Drop "cc"
17732 attribute. Add new patterns to match output of
17734 * config/avr/avr-fixed.md: Likewise.
17735 * config/avr/avr.c (cc_reg_rtx): New.
17736 (avr_parallel_insn_from_insns): Adjust insn count
17737 for removal of set of cc0.
17738 (avr_is_casesi_sequence): Likewise.
17739 (avr_casei_sequence_check_operands): Likewise.
17740 (avr_optimize_casesi): Likewise. Also insert
17741 new insns after jump_insn.
17742 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
17743 for removal of set of cc0.
17744 (avr_init_expanders): Initialize cc_reg_rtx.
17745 (avr_regno_reg_class): Handle REG_CC.
17746 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
17747 (avr_notice_update_cc): Remove function.
17748 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
17749 (compare_condition): Adjust for PARALLEL with
17751 (out_shift_with_cnt): Likewise.
17752 (ashlhi3_out): Likewise.
17753 (ashrhi3_out): Likewise.
17754 (lshrhi3_out): Likewise.
17755 (avr_class_max_nregs): Return single reg for REG_CC.
17756 (avr_compare_pattern): Check for REG_CC instead
17758 (avr_reorg_remove_redundant_compare): Likewise.
17759 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
17760 (avr_hard_regno_nregs): Return single reg for REG_CC.
17761 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
17762 (avr_md_asm_adjust): Clobber REG_CC.
17763 (TARGET_HARD_REGNO_NREGS): Define.
17764 (TARGET_CLASS_MAX_NREGS): Define.
17765 (TARGET_MD_ASM_ADJUST): Define.
17766 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
17768 (enum reg_class): Add CC_REG class.
17769 (NOTICE_UPDATE_CC): Remove.
17770 (CC_OVERFLOW_UNUSABLE): Remove.
17771 (CC_NO_CARRY): Remove.
17772 * config/avr/avr.md: Turn existing patterns into
17773 define_insn_and_split style patterns where the splitter
17774 adds a clobber of the condition code register. Drop "cc"
17775 attribute. Add new patterns to match output of
17777 (sez): Remove unused pattern.
17779 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
17782 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
17785 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
17788 * config/aarch64/constraints.md (Utq): Require the address to
17789 be valid for both the element mode and for V2DImode.
17791 2021-04-28 Jakub Jelinek <jakub@redhat.com>
17792 Tobias Burnus <tobias@codesourcery.com>
17794 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
17795 * gcc.c (process_command): New variable.
17796 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
17797 set it if -foffload is defaulted.
17798 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
17799 (compile_offload_image): If OFFLOAD_DEFAULTED and
17800 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
17801 if corresponding mkoffload can't be found.
17802 (compile_images_for_offload_targets): Likewise. Free and clear
17803 offload_names if no valid offload is found.
17804 * config.in: Regenerate.
17805 * configure: Regenerate.
17807 2021-04-28 Richard Biener <rguenther@suse.de>
17809 PR tree-optimization/100292
17810 * tree-vect-generic.c (expand_vector_condition): Do not fold
17813 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17815 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
17816 * config/rs6000/aix64.opt (m64): New.
17819 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17821 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
17822 (index_term_p): Handle ASHIFT too.
17824 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17826 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
17827 (sync_lock_test_and_set<mode>): Adjust accordingly.
17828 (sync_lock_release<mode>): Likewise.
17830 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17832 * config/vax/vax-protos.h (adjacent_operands_p): Remove
17834 * config/vax/vax.c (adjacent_operands_p): Remove.
17836 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
17838 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
17839 through to the non-conditional execution case if getting the
17840 condition for conditional execution has failed.
17842 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17844 PR middle-end/100284
17845 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
17846 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
17847 than asserting on it.
17849 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17851 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
17852 with TARGET_AIX_OS.
17854 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17857 * calls.c (precompute_register_parameters): Additionally test
17858 targetm.precompute_tls_p to pre-compute argument.
17859 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
17860 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
17861 * target.def (precompute_tls_p): New.
17862 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
17863 * doc/tm.texi: Regenerated.
17865 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17868 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
17869 back to HOST_WIDE_INT.
17871 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17874 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
17875 memory alignment for the outer mode.
17877 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
17879 PR middle-end/90773
17880 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
17881 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
17883 2021-04-27 Alex Coplan <alex.coplan@arm.com>
17886 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
17887 with negative immediates: ensure we expand cbranchsi4_scratch
17888 correctly and ensure we satisfy its constraints.
17889 * config/arm/sync.md
17890 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
17891 attempt to tie two output operands together with constraints;
17892 collapse two alternatives.
17893 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
17894 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
17896 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17899 * config/aarch64/predicates.md (aarch64_sub_immediate,
17900 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
17901 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
17902 * config/aarch64/aarch64.c (aarch64_print_operand,
17903 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
17905 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17907 PR tree-optimization/100239
17908 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
17909 permutations with all indices from the first zero element as vec_shl.
17911 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17913 PR rtl-optimization/100254
17914 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
17915 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
17917 2021-04-27 Richard Biener <rguenther@suse.de>
17919 PR tree-optimization/99912
17920 * passes.def: Add comment about new TODO_remove_unused_locals.
17921 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
17924 2021-04-27 Richard Biener <rguenther@suse.de>
17926 PR tree-optimization/99912
17927 * passes.def (pass_all_optimizations): Add pass_dse before
17928 the first pass_dce, move the first pass_dse before the
17929 pass_dce following pass_pre.
17931 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17933 PR tree-optimization/95527
17934 * generic-match-head.c: Include tm.h.
17935 * gimple-match-head.c: Include tm.h.
17936 * match.pd (CLZ == INTEGER_CST): Don't use
17937 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
17938 if clz == CFN_CLZ. Add missing val declaration.
17939 (CTZ cmp CST): New simplifications.
17941 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17943 PR tree-optimization/96696
17944 * expr.c (expand_expr_divmod): New function.
17945 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
17946 divisions. Formatting fixes.
17947 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
17950 2021-04-27 Martin Jambor <mjambor@suse.cz>
17953 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
17954 If removing a call statement LHS SSA name, release it.
17956 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
17959 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
17960 is valid before including it in the mask.
17962 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17965 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
17968 2021-04-27 Richard Biener <rguenther@suse.de>
17970 PR tree-optimization/100051
17971 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
17972 disambiguator based on access size vs. decl size.
17974 2021-04-27 Richard Biener <rguenther@suse.de>
17976 PR tree-optimization/100278
17977 * tree-ssa-pre.c (compute_avail): Give up when we cannot
17978 adjust TBAA beacuse of mismatching bases.
17980 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17983 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
17984 For any_rotate define_insn_split and following splitters, use
17985 SWI iterator instead of SWI48.
17987 2021-04-27 Richard Biener <rguenther@suse.de>
17989 PR tree-optimization/99776
17990 * match.pd (bit_field_ref (ctor)): Relax element extract
17991 type compatibility checks.
17993 2021-04-27 Cui,Lili <lili.cui@intel.com>
17995 * common/config/i386/i386-common.c (processor_names):
17996 Sync processor_names with processor_type.
17997 * config/i386/i386-options.c (processor_cost_table):
17998 Sync processor_cost_table with processor_type.
18000 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18002 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
18003 (irange::set): Call irange_set_1bit_anti_range for handling all
18004 1-bit ranges. Fall through on ~[MIN,MAX].
18006 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18008 * value-range.cc (irange::legacy_num_pairs): Remove.
18009 (irange::invert): Change gcc_assert to gcc_checking_assert.
18010 * value-range.h (irange::num_pairs): Adjust for a cached
18011 num_pairs(). Also, rename all gcc_assert's to
18012 gcc_checking_assert's.
18014 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18016 * value-range.cc (irange::operator=): Set m_kind.
18017 (irange::copy_to_legacy): Handle varying and undefined sources
18018 as a legacy copy since they can be easily copied.
18019 (irange::irange_set): Set m_kind.
18020 (irange::irange_set_anti_range): Same.
18021 (irange::set): Rename normalize_min_max to normalize_kind.
18022 (irange::verify_range): Adjust for multi-ranges having the
18024 (irange::irange_union): Set m_kind.
18025 (irange::irange_intersect): Same.
18026 (irange::invert): Same.
18027 * value-range.h (irange::kind): Always return m_kind.
18028 (irange::varying_p): Rename to...
18029 (irange::varying_comptaible_p): ...this.
18030 (irange::undefined_p): Only look at m_kind.
18031 (irange::irange): Always set VR_UNDEFINED if applicable.
18032 (irange::set_undefined): Always set VR_UNDEFINED.
18033 (irange::set_varying): Always set m_kind to VR_VARYING.
18034 (irange::normalize_min_max): Rename to...
18035 (irange::normalize_kind): ...this.
18037 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18039 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
18040 Adjust for constant_p including varying_p.
18041 * tree-vrp.c (vrp_prop::finalize): Same.
18042 (determine_value_range): Same.
18043 * vr-values.c (vr_values::range_of_expr): Same.
18044 * value-range.cc (irange::symbolic_p): Do not check varying_p.
18045 (irange::constant_p): Same.
18047 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18049 * value-range.cc (irange::legacy_lower_bound): Replace
18050 !undefined_p check with num_ranges > 0.
18051 (irange::legacy_upper_bound): Same.
18052 * value-range.h (irange::type): Same.
18053 (irange::lower_bound): Same.
18054 (irange::upper_bound): Same.
18056 2021-04-26 Richard Biener <rguenther@suse.de>
18058 PR tree-optimization/99956
18059 * gimple-loop-interchange.cc (compute_access_stride):
18060 Try instantiating the access in a shallower loop nest
18061 if instantiating failed.
18062 (compute_access_strides): Pass adjustable loop_nest
18063 to compute_access_stride.
18065 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
18067 * doc/sourcebuild.texi (arm_cmse_hw): Document.
18069 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18071 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
18073 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
18074 Nathan Sidwell <nathan@codesourcery.com>
18075 Tom de Vries <vries@codesourcery.com>
18076 Julian Brown <julian@codesourcery.com>
18077 Kwok Cheung Yeung <kcy@codesourcery.com>
18079 * omp-offload.c (oacc_validate_dims): Implement
18080 '-Wopenacc-parallelism'.
18081 * doc/invoke.texi (-Wopenacc-parallelism): Document.
18083 2021-04-26 Richard Biener <rguenther@suse.de>
18085 * tree-cfg.h (gimplify_build1): Remove.
18086 (gimplify_build2): Likewise.
18087 (gimplify_build3): Likewise.
18088 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
18089 (gimplify_build2): Likewise.
18090 (gimplify_build3): Likewise.
18091 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
18093 (gimplify_build2): Likewise.
18094 (gimplify_build3): Likewise.
18095 (tree_vec_extract): Use resimplify with following SSA edges.
18096 (expand_vector_parallel): Avoid passing NULL size/bitpos
18097 to tree_vec_extract.
18098 * expr.c (store_constructor): Deal with zero-element CTORs.
18099 * match.pd (bit_field_ref <vector CTOR>): Make sure to
18100 produce vector constants when possible.
18102 2021-04-26 Richard Biener <rguenther@suse.de>
18104 * tree-complex.c: Include gimple-fold.h.
18105 (expand_complex_addition): Use gimple_build.
18106 (expand_complex_multiplication_components): Likewise.
18107 (expand_complex_multiplication): Likewise.
18108 (expand_complex_div_straight): Likewise.
18109 (expand_complex_div_wide): Likewise.
18110 (expand_complex_division): Likewise.
18111 (expand_complex_conjugate): Likewise.
18112 (expand_complex_comparison): Likewise.
18114 2021-04-26 Richard Biener <rguenther@suse.de>
18116 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18117 of legacy gimplify_buildN API.
18119 2021-04-26 Richard Biener <rguenther@suse.de>
18121 PR tree-optimization/99473
18122 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18125 2021-04-26 Richard Biener <rguenther@suse.de>
18127 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18128 Use replace_call_with_value.
18130 2021-04-26 Richard Biener <rguenther@suse.de>
18132 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18133 (update_gimple_call): Likewise.
18134 (update_call_from_tree): Likewise.
18135 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18136 (valid_gimple_call_p): Likewise.
18137 (move_ssa_defining_stmt_for_defs): Likewise.
18138 (finish_update_gimple_call): Likewise.
18139 (update_gimple_call): Likewise.
18140 (update_call_from_tree): Likewise.
18141 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18142 * gimple-fold.h (update_gimple_call): Declare.
18143 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18144 tree-ssa-propagate.c.
18145 (update_gimple_call): Likewise.
18146 (valid_gimple_call_p): Likewise.
18147 (finish_update_gimple_call): Likewise, and simplify.
18148 (gimplify_and_update_call_from_tree): Implement
18149 update_call_from_tree functionality, avoid excessive
18150 push/pop_gimplify_context.
18151 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18152 (gimple_fold_call): Likewise.
18153 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18154 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18155 (pass_fold_builtins::execute): Likewise.
18156 (optimize_stack_restore): Use replace_call_with_value.
18157 * tree-cfg.c (fold_loop_internal_call): Likewise.
18158 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18159 only gimplify_and_update_call_from_tree.
18160 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18161 (handle_builtin_strchr): Likewise.
18162 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18164 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18167 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18168 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18171 2021-04-25 liuhongt <hongtao.liu@intel.com>
18174 * config/i386/i386-builtin.def (BDESC): Change the icode of
18175 the following builtins to CODE_FOR_nothing.
18176 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18177 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18178 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18179 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18180 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18181 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18182 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18183 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18184 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18185 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18186 (sse2_eq<mode>3): Ditto.
18187 (sse4_1_eqv2di3): Ditto.
18188 (sse2_gt<mode>3): Rename to ..
18189 (*sse2_gt<mode>3): .. this.
18191 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18194 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18197 * config/darwin.c (darwin_binds_local_p): Assume that any
18198 public symbol might be interposed for PIC code. Update function
18199 header comment to reflect current Darwin capability.
18201 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18204 * config/darwin.c (darwin_binds_local_p): Assume that any
18205 public symbol might be interposed for PIC code. Update function
18206 header comment to reflect current Darwin capability.
18208 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18210 * doc/sourcebuild.texi: Document no-opts and any-opts target
18213 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18215 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18217 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18219 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18221 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18224 * config/i386/i386-options.c (ix86_option_override_internal):
18225 Error out when -m96bit-long-double is used with 64bit targets.
18226 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18228 2021-04-23 Martin Liska <mliska@suse.cz>
18230 * lto-wrapper.c: Remove FIXME about usage of
18231 hardware_concurrency. The function is not on par with
18234 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18237 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
18238 Copy operand 3 to operand 4. Use sse_reg_operand
18239 as operand 3 predicate.
18240 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
18241 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
18242 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
18243 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
18244 Copy operand 1 to operand 0.
18245 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
18246 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
18247 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
18249 2021-04-23 Alex Coplan <alex.coplan@arm.com>
18251 PR rtl-optimization/100230
18252 * early-remat.c (early_remat::sort_candidates): Use delete[]
18253 instead of delete for array allocated with new[].
18255 2021-04-23 Richard Biener <rguenther@suse.de>
18257 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
18258 (capture_info::capture_info): Likewise.
18259 (capture_info::walk_match): Likewise.
18260 (expr::gen_transform): Likewise.
18261 (dt_simplify::gen_1): Likewise.
18262 * gimple-match-head.c (maybe_resimplify_conditional_op):
18263 Remove VEC_COND_EXPR special-casing.
18264 (gimple_simplify): Likewise.
18265 * gimple.c (gimple_could_trap_p_1): Adjust.
18266 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18267 to participate in PRE.
18269 2021-04-23 Richard Biener <rguenther@suse.de>
18271 * cfganal.c (connect_infinite_loops_to_exit): First call
18272 add_noreturn_fake_exit_edges.
18273 * ipa-sra.c (process_scan_results): Do not call the now redundant
18274 add_noreturn_fake_exit_edges.
18275 * predict.c (tree_estimate_probability): Likewise.
18276 (rebuild_frequencies): Likewise.
18277 * store-motion.c (one_store_motion_pass): Likewise.
18279 2021-04-23 Richard Biener <rguenther@suse.de>
18281 PR tree-optimization/100222
18282 * predict.c (pass_profile::execute): Remove redundant call to
18283 mark_irreducible_loops.
18284 (report_predictor_hitrates): Likewise.
18286 2021-04-23 Richard Biener <rguenther@suse.de>
18288 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18289 valid_gimple_rhs_p by instead gimplifying to one.
18291 2021-04-23 Richard Biener <rguenther@suse.de>
18293 PR tree-optimization/99971
18294 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18295 Always use TBAA for loads.
18297 2021-04-23 liuhongt <hongtao.liu@intel.com>
18300 * config/i386/i386-options.c (ix86_option_override_internal):
18301 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18302 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18303 by target attribute.
18305 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18307 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18309 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18311 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18313 * config.gcc (powerpc-ibm-aix6.*): Remove.
18314 * config/rs6000/aix61.h: Delete.
18316 2021-04-22 Martin Liska <mliska@suse.cz>
18318 PR testsuite/100159
18319 PR testsuite/100192
18320 * builtins.c (expand_builtin): Fix typos and missing comments.
18321 * dwarf2out.c (gen_subprogram_die): Likewise.
18322 (gen_struct_or_union_type_die): Likewise.
18324 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18327 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18328 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18330 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18332 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18333 based on TARGET_ISA_64BIT.
18334 (TARGET_64BIT_P): Remove definition based on
18335 TARGET_ISA_64BIT_P().
18337 2021-04-21 Martin Liska <mliska@suse.cz>
18340 2021-04-21 Martin Liska <mliska@suse.cz>
18342 * lto-wrapper.c (cpuset_popcount): Remove.
18343 (init_num_threads): Remove and use hardware_concurrency.
18345 2021-04-21 Martin Liska <mliska@suse.cz>
18348 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18349 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18350 when incremental LTO linking happens.
18352 2021-04-21 Martin Liska <mliska@suse.cz>
18354 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18355 makeserver cannot be detected, then use -flto=N fallback.
18357 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18359 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18360 default to yes for aarch64-linux-gnu.
18361 * configure: Regenerate.
18363 2021-04-21 Martin Liska <mliska@suse.cz>
18365 * lto-wrapper.c (cpuset_popcount): Remove.
18366 (init_num_threads): Remove and use hardware_concurrency.
18368 2021-04-21 Martin Liska <mliska@suse.cz>
18370 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18371 which remains to be '(... || 0)' and clang complains about it.
18372 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18373 (add_AT_vms_delta): Likewise.
18374 * tree.c (fld_simplified_type): Use rather more common pattern
18375 for disabling of something (#if 0).
18376 (get_tree_code_name): Likewise.
18377 (verify_type_variant): Likewise.
18379 2021-04-21 Martin Liska <mliska@suse.cz>
18381 * config/i386/i386-expand.c (decide_alignment): Use newly named
18382 macro TARGET_CPU_P.
18383 * config/i386/i386.c (ix86_decompose_address): Likewise.
18384 (ix86_address_cost): Likewise.
18385 (ix86_lea_outperforms): Likewise.
18386 (ix86_avoid_lea_for_addr): Likewise.
18387 (ix86_add_stmt_cost): Likewise.
18388 * config/i386/i386.h (TARGET_*): Remove.
18389 (TARGET_CPU_P): New macro.
18390 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18391 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18392 (swap_top_of_ready_list): Likewise.
18393 (ix86_atom_sched_reorder): Likewise.
18394 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18395 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18397 2021-04-21 Martin Liska <mliska@suse.cz>
18399 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18401 (SET_TARGET_NO_SAHF): Likewise.
18402 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18403 (SET_TARGET_PREFETCH_SSE): Likewise.
18404 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18405 (SET_TARGET_NO_TUNE): Likewise.
18406 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18407 (SET_TARGET_NO_80387): Likewise.
18409 * config/i386/i386.h (TARGET_*): Remove.
18410 * opth-gen.awk: Generate new used macros.
18412 2021-04-21 Martin Liska <mliska@suse.cz>
18414 * config/i386/i386.h (PTA_*): Remove.
18415 (enum pta_flag): New.
18416 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18417 * config/i386/i386-isa.def: New file.
18419 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18422 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18423 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18424 jump table targets.
18426 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18428 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18429 x86_64-*-* targets.
18430 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18432 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18433 (ix86_handle_option): Handle -mmwait.
18434 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18435 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18436 __builtin_ia32_monitor and __builtin_ia32_mwait.
18437 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18438 (ix86_valid_target_attribute_inner_p): Likewise.
18439 (ix86_option_override_internal): Enable mwait/monitor
18440 instructions for -msse3.
18441 * config/i386/i386.h (TARGET_MWAIT): New.
18442 (TARGET_MWAIT_P): Likewise.
18443 * config/i386/i386.opt: Add -mmwait.
18444 * config/i386/mwaitintrin.h: New file.
18445 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18446 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18448 (@sse3_monitor_<mode>): Likewise.
18449 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18450 * doc/extend.texi: Document mwait target attribute.
18451 * doc/invoke.texi: Document -mmwait.
18453 2021-04-21 Martin Liska <mliska@suse.cz>
18455 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18456 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18457 * config/i386/stringop.def (DEF_ENUM): Likewise.
18459 2021-04-21 Martin Liska <mliska@suse.cz>
18461 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18462 of print_generic_expr.
18464 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18466 PR rtl-optimization/100148
18467 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18470 2021-04-21 Martin Liska <mliska@suse.cz>
18473 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18474 free_dominance_info calls.
18476 2021-04-21 Richard Biener <rguenther@suse.de>
18478 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18479 parameter (and assume it to be false).
18480 (fold_gimple_assign): Adjust, remove all callers of
18481 maybe_fold_reference calling it with is_lhs true.
18482 (gimple_fold_call): Likewise.
18483 (fold_stmt_1): Likewise.
18485 2021-04-21 Richard Biener <rguenther@suse.de>
18487 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18488 (fold_binary_loc): Adjust.
18489 (fold_ternary_loc): Likewise.
18491 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18493 PR middle-end/100130
18494 * varasm.c (get_block_for_decl): Make sure that any use of the
18495 retain attribute matches the section's retain flag.
18496 (switch_to_section): Check for retain mismatches even when
18497 changing sections, but do not warn if the given decl is the
18498 section's named.decl.
18499 (output_object_block): Pass the first decl in the block (if any)
18500 to switch_to_section.
18502 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18504 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18505 __CRC32__ for -mcrc32.
18506 * config/i386/i386-options.c (ix86_option_override_internal):
18507 Enable crc32 instruction for -msse4.2.
18508 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18510 (sse4_2_crc32di): Likewise.
18511 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18514 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18517 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18520 2021-04-20 Martin Liska <mliska@suse.cz>
18522 * doc/invoke.texi: Fix typo.
18523 * params.opt: Likewise.
18525 2021-04-20 Martin Liska <mliska@suse.cz>
18527 * doc/invoke.texi: Document new param.
18529 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18531 PR tree-optimization/100081
18532 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18533 rather than gori_compute_cache.
18534 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18535 (range_def_chain::m_logical_depth): New member.
18536 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18537 (range_def_chain::get_def_chain): Don't build defchains through more
18538 than LOGICAL_LIMIT logical expressions.
18539 * params.opt (param_ranger_logical_depth): New.
18541 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18544 * config/arm/arm.c (arm_configure_build_target): Do not strip
18545 extended FPU/SIMD feature bits from the target ISA when -mfpu
18546 is specified (partial revert of r11-8168).
18548 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18550 * params.opt (-param=openacc-kernels=): Add.
18551 * omp-oacc-kernels-decompose.cc
18552 (pass_omp_oacc_kernels_decompose::gate): Use it.
18553 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18554 (--param): ... here, 'openacc-kernels'.
18556 2021-04-19 Martin Liska <mliska@suse.cz>
18559 * gengtype.c (finish_root_table): Align function arguments
18560 in between declaration and definition.
18562 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18564 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18565 frames larger than the SEH maximum frame size.
18567 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18569 PR rtl-optimization/99927
18570 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18571 is dead, just drop it.
18573 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18576 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18577 * doc/tm.texi: Regenerate.
18578 * doc/tm.texi.in (D language and ABI): Add @hook for
18579 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18581 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18583 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18585 (darwin_d_register_target_info): New function.
18586 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18587 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18589 (dragonfly_d_register_target_info): New function.
18590 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18591 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18593 (freebsd_d_register_target_info): New function.
18594 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18595 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18597 (glibc_d_register_target_info): New function.
18598 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18599 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
18601 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
18602 as handler for objectFormat key.
18603 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
18605 (winnt_d_register_target_info): New function.
18606 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18607 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
18609 (netbsd_d_register_target_info): New function.
18610 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18611 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
18613 (openbsd_d_register_target_info): New function.
18614 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18615 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
18616 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
18617 handler for objectFormat key.
18618 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
18620 (rs6000_d_register_target_info): Add
18621 rs6000_d_handle_target_object_format as handler for objectFormat key.
18622 * config/sol2-d.c (solaris_d_handle_target_object_format): New
18624 (solaris_d_register_target_info): New function.
18625 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18627 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18630 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
18631 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
18633 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
18634 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
18635 the old and new alignment after applying MIN/MAX to it is different.
18637 2021-04-16 Tamar Christina <tamar.christina@arm.com>
18640 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
18641 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
18643 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
18645 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
18647 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
18648 this section and its subsections.
18650 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18653 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
18654 define_insn patterns.
18656 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18658 PR rtl-optimization/98689
18659 * reg-notes.def (UNTYPED_CALL): New note.
18660 * combine.c (distribute_notes): Handle it.
18661 * emit-rtl.c (try_split): Likewise.
18662 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
18663 that calls with the note implicitly set all return value registers.
18664 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
18667 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18669 PR rtl-optimization/99596
18670 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
18671 register accesses for const calls. Assume that pure functions
18672 can only read from global registers. Ignore cases in which
18673 the stack pointer has been marked global.
18675 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18678 * tree-vect-loop.c (vect_transform_loop): Don't remove just
18679 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
18680 them by their last argument.
18682 2021-04-15 Martin Liska <mliska@suse.cz>
18684 * doc/invoke.texi: Other params don't use it, remove it.
18686 2021-04-15 Richard Biener <rguenther@suse.de>
18688 * gimple-builder.h: Add deprecation note.
18690 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18693 * attribs.h (restrict_type_identity_attributes_to): Declare.
18694 * attribs.c (restrict_type_identity_attributes_to): New function.
18696 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18699 * attribs.h (affects_type_identity_attributes): Declare.
18700 * attribs.c (remove_attributes_matching): New function.
18701 (affects_type_identity_attributes): Likewise.
18703 2021-04-15 Jakub Jelinek <jakub@redhat.com>
18706 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
18707 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
18708 ZERO_EXTEND, SIGN_EXTEND or AND.
18710 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18712 PR rtl-optimization/99929
18713 * rtl.h (same_vector_encodings_p): New function.
18714 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
18715 * cselib.c (rtx_equal_for_cselib_1): Likewise.
18716 * jump.c (rtx_renumbered_equal_p): Likewise.
18717 * lra-constraints.c (operands_match_p): Likewise.
18718 * reload.c (operands_match_p): Likewise.
18719 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
18721 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18723 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
18724 more information about variable-length CONST_VECTORs.
18726 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
18728 PR rtl-optimization/100066
18729 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
18730 ordered modes when choosing splitting mode for hard reg.
18732 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18735 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
18737 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
18739 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
18741 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
18742 for mask operand types.
18743 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18744 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
18746 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
18748 * config/s390/s390.c (s390_const_operand_ok): Check the new
18749 operand types and generate a list of valid values.
18751 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18753 * doc/tm.texi: Regenerate.
18754 * doc/tm.texi.in (D language and ABI): Add @hook for
18755 TARGET_D_REGISTER_OS_TARGET_INFO.
18757 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18759 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
18761 (aarch64_d_register_target_info): New function.
18762 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
18764 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
18766 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
18767 (arm_d_register_target_info): New function.
18768 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
18769 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18770 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
18771 (ix86_d_register_target_info): New function.
18772 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
18773 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18774 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
18775 (mips_d_register_target_info): New function.
18776 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
18777 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18778 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
18779 (pa_d_register_target_info): New function.
18780 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
18781 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18782 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
18784 (riscv_d_register_target_info): New function.
18785 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
18786 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18787 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
18789 (rs6000_d_register_target_info): New function.
18790 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
18792 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18793 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
18794 (s390_d_register_target_info): New function.
18795 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
18796 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18797 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
18799 (sparc_d_register_target_info): New function.
18800 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
18801 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18802 * doc/tm.texi: Regenerate.
18803 * doc/tm.texi.in (D language and ABI): Add @hook for
18804 TARGET_D_REGISTER_CPU_TARGET_INFO.
18806 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18808 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
18809 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
18810 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
18811 * doc/tm.texi: Regenerate.
18812 * doc/tm.texi.in (D language and ABI): Add @hook for
18813 TARGET_D_HAS_STDCALL_CONVENTION.
18815 2021-04-14 Richard Biener <rguenther@suse.de>
18817 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
18818 VEC_COND_EXPRs have a gimple_val condition.
18819 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
18820 can no longer have a GENERIC condition.
18822 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
18825 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
18826 from the isa_delta when -mfpu has been used.
18827 (arm_options_perform_arch_sanity_checks): It's the architecture that
18828 lacks an FPU not the processor.
18830 2021-04-13 Richard Biener <rguenther@suse.de>
18832 PR tree-optimization/100053
18833 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
18834 not use optimistic dominance queries for backedges to validate
18836 (dominated_by_p_w_unex): Add parameter to ignore executable
18837 state on backedges.
18838 (rpo_elim::eliminate_avail): Adjust.
18840 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18843 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
18844 *aarch64_bfxilsi_extrdi): New define_insn patterns.
18846 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18849 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
18850 outermode, return NULL if the result doesn't encode back to the
18851 original byte sequence.
18852 (simplify_gen_subreg): Don't create SUBREGs from constants to
18853 MODE_COMPOSITE_P outermode.
18855 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18857 PR rtl-optimization/99905
18858 * combine.c (expand_compound_operation): If pos + len > modewidth,
18859 perform the right shift by pos in inner_mode and then convert to mode,
18860 instead of trying to simplify a shift of rtx with inner_mode by pos
18861 as if it was a shift in mode.
18863 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18866 * combine.c (simplify_and_const_int_1): Don't optimize varop
18867 away if it has side-effects.
18869 2021-04-12 Martin Liska <mliska@suse.cz>
18871 * doc/extend.texi: Escape @smallexample content.
18873 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
18875 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
18876 alternative in order to load a DFP zero.
18878 2021-04-12 Martin Liska <mliska@suse.cz>
18880 * doc/extend.texi: Be more precise in documentation
18881 of symver attribute.
18883 2021-04-12 Martin Liska <mliska@suse.cz>
18886 * gimplify.c (gimplify_expr): Right now, we unpoison all
18887 variables before a goto <dest>. We should not do it if we are
18890 2021-04-12 Cui,Lili <lili.cui@intel.com>
18892 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
18894 * common/config/i386/i386-common.c (processor_names): Add
18896 (processor_alias_table): Add rocketlake.
18897 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
18898 INTEL_COREI7_ROCKETLAKE.
18899 * config.gcc: Add -march=rocketlake.
18900 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18902 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
18903 (processor_cost_table): Add rocketlake cost.
18904 * config/i386/i386.h (ix86_size_cost) : Define
18906 (processor_type) : Add PROCESSOR_ROCKETLAKE.
18907 (PTA_ROCKETLAKE): Ditto.
18908 * doc/extend.texi: Add rocketlake.
18909 * doc/invoke.texi: Add rocketlake.
18911 2021-04-12 Cui,Lili <lili.cui@intel.com>
18913 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
18914 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
18915 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
18916 * doc/invoke.texi: Change alderlake ISA list.
18918 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
18920 PR middle-end/98088
18921 * omp-expand.c (expand_oacc_collapse_init): Update condition in
18924 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
18927 * config/i386/serializeintrin.h (_serialize): Defined as macro.
18929 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18932 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
18933 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
18935 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18937 PR middle-end/99989
18938 * gimple-ssa-warn-alloca.c
18939 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
18940 0 with integer precision unconditionally.
18942 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18944 PR rtl-optimization/98601
18945 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
18946 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
18947 unaligned_mems handle VOIDmode like BLKmode.
18949 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
18952 * tree.c (free_lang_data_in_decl): Do not release body of
18953 declare_variant_alt.
18955 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
18957 * config/aarch64/aarch64.c (aarch64_option_restore): If the
18958 architecture was specified explicitly and the tuning wasn't,
18959 tune for the architecture rather than the configured default CPU.
18961 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
18963 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
18964 as the temporary register.
18966 2021-04-09 Martin Liska <mliska@suse.cz>
18968 * doc/extend.texi: Move non-target attributes on the top level.
18970 2021-04-09 Martin Liska <mliska@suse.cz>
18972 * doc/invoke.texi: Document minimum and maximum value of the
18973 argument for both supported compression algorithms.
18975 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
18977 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
18978 TLS BSS before TLS data.
18979 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
18981 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
18983 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
18985 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
18987 * match.pd: Extend vec_cond folds to handle shifts.
18989 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
18991 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
18994 2021-04-08 Alex Coplan <alex.coplan@arm.com>
18997 * config/arm/iterators.md (MVE_vecs): New.
18998 (V_elem): Also handle V2DF.
18999 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
19000 (*mve_vdup<mode>): ... this. Remove second alternative since
19001 vec_duplicate of const_int is not canonical RTL, and we don't
19002 want to match symbol_refs.
19003 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
19005 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
19007 * fold-const.c (fold_single_bit_test): Fix typo.
19008 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
19011 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19013 PR tree-optimization/97513
19014 * tree-vect-slp.c (vect_add_slp_permutation): New function,
19016 (vectorizable_slp_permutation): ...here. Detect cases in which
19017 all VEC_PERM_EXPRs are guaranteed to have the same stepped
19018 permute vector and only generate one permute vector for that case.
19019 Extend that case to handle variable-length vectors.
19021 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19023 PR tree-optimization/99873
19024 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
19025 (vect_build_slp_instance): Don't split store groups that could
19026 use IFN_STORE_LANES.
19028 2021-04-07 Jakub Jelinek <jakub@redhat.com>
19031 * varasm.c (output_constant_pool_contents): Don't strip name encoding
19032 from XSTR (desc->sym, 0) or from label before passing those to
19035 2021-04-07 Richard Biener <rguenther@suse.de>
19037 PR tree-optimization/99954
19038 * tree-loop-distribution.c: Include tree-affine.h.
19039 (generate_memcpy_builtin): Try using tree-affine to prove
19041 (loop_distribution::classify_builtin_ldst): Always classify
19044 2021-04-07 Richard Biener <rguenther@suse.de>
19046 PR tree-optimization/99947
19047 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
19048 steps vector to avoid pushing elements from the reallocated
19051 2021-04-07 Richard Biener <rguenther@suse.de>
19053 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
19054 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
19056 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
19058 (debug_vn_reference_ops): New.
19060 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
19062 PR tree-optimization/98736
19063 * tree-loop-distribution.c
19064 * (loop_distribution::bb_top_order_init):
19065 Compute RPO with programing order preserved by calling function
19066 rev_post_order_and_mark_dfs_back_seme.
19068 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
19071 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
19072 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
19074 (process_bb_lives): Don't update biggest mode of hard reg for
19075 implicit in multi-register group. Use the new functions for
19076 updating dead_set and unused_set by register notes.
19078 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
19080 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
19083 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
19085 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
19086 (skylake_memset): Likewise.
19087 (skylake_cost): Change CLEAR_RATIO to 17.
19088 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19089 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
19090 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
19092 2021-04-06 Richard Biener <rguenther@suse.de>
19094 PR tree-optimization/99880
19095 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
19096 set vectorized defs of relevant PHIs.
19098 2021-04-06 Richard Biener <rguenther@suse.de>
19100 PR tree-optimization/99924
19101 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
19102 nodes w/o scalar stmts as visited.
19104 2021-04-06 Alex Coplan <alex.coplan@arm.com>
19107 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
19108 PCS for [su]fix_optab.
19110 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19112 * config/darwin.c (machopic_legitimize_pic_address): Check
19113 that the current pic register is one of the hard reg set
19114 before setting liveness.
19116 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19118 * config/darwin.c (machopic_legitimize_pic_address): Fix
19119 whitespace, remove unused code.
19121 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19123 PR tree-optimization/99882
19124 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19127 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19129 PR rtl-optimization/99863
19130 * dse.c (replace_read): Drop regs_live argument. Instead of
19131 regs_live, use store_insn->fixed_regs_live if non-NULL,
19132 otherwise punt if insns sequence clobbers or sets any hard
19135 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19138 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19139 * targhooks.c (default_print_patchable_function_entry_1): New function,
19140 copied from default_print_patchable_function_entry with an added flags
19142 (default_print_patchable_function_entry): Rewritten into a small
19143 wrapper around default_print_patchable_function_entry_1.
19144 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19146 (rs6000_print_patchable_function_entry): New function.
19148 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19150 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19152 2021-04-01 Jason Merrill <jason@redhat.com>
19155 * common.opt: Document v15 and v16.
19157 2021-04-01 Richard Biener <rguenther@suse.de>
19159 PR tree-optimization/99863
19160 * gimplify.c (gimplify_init_constructor): Recompute vector
19163 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19165 * doc/extend.texi (symver attribute): Fix up syntax errors
19168 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19170 PR tree-optimization/96573
19171 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19172 also pointer types.
19174 2021-04-01 Richard Biener <rguenther@suse.de>
19176 PR tree-optimization/99856
19177 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19178 precision to vector element precision.
19180 2021-04-01 Martin Jambor <mjambor@suse.cz>
19182 PR tree-optimization/97009
19183 * tree-sra.c (access_or_its_child_written): New function.
19184 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19187 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19190 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19192 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19195 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19196 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19197 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19198 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19199 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19200 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19202 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19203 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19204 (define_attr "prefixed"): Update initializer.
19206 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19209 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19210 (DW_RANGES_IDX_SKELETON): Define.
19211 (struct dw_ranges): Add begin_entry and end_entry members.
19212 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19213 (add_ranges_num): Adjust r initializer for addition of *_entry
19215 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19216 set idx to DW_RANGES_IDX_SKELETON.
19217 (use_distinct_base_address_for_range): New function.
19218 (index_rnglists): Don't set r->idx if it is equal to
19219 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19220 r->end_entry for -gsplit-dwarf if those will be needed by
19222 (output_rnglists): Add DWO argument. If true, switch to
19223 debug_ranges_dwo_section rather than debug_ranges_section.
19224 Adjust l1/l2 label indexes. Only output the offset table when
19225 dwo is true and don't include in there the skeleton range
19226 entry if present. For -gsplit-dwarf, skip ranges that belong
19227 to the other rnglists section. Change return type from void
19228 to bool and return true if there are any range entries for
19229 the other section. For dwarf_split_debug_info use
19230 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19231 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19232 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19233 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19234 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19235 and range_base_label indexes.
19236 (dwarf2out_finish): Call index_rnglists earlier before finalizing
19237 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
19238 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
19239 with different dwo arguments.
19240 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
19242 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19244 PR tree-optimization/98268
19245 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
19246 recompute_tree_invariant_for_addr_expr after successfully
19247 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
19249 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19251 PR tree-optimization/99726
19252 * tree-data-ref.c (create_intersect_range_checks_index): Bail
19253 out if there is more than one access function SCEV for the loop
19256 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19258 PR rtl-optimization/97141
19259 PR rtl-optimization/98726
19260 * emit-rtl.c (valid_for_const_vector_p): Return true for
19262 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
19263 poly_wide_int instead of a wide_int.
19264 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
19266 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19267 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19268 false for CONST_VECTORs that cannot be forced to memory.
19269 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19270 is too complex to force to memory, build it up from individual
19273 2021-03-31 Jan Hubicka <jh@suse.cz>
19276 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19278 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19281 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19284 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19286 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19287 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19289 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19290 "rep movsb/stosb" only for known sizes.
19291 * config/i386/i386-options.c (processor_cost_table): Use Ice
19292 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19293 Rapids and Alder Lake.
19294 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19295 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19296 (icelake_memset): Likewise.
19297 (icelake_cost): Likewise.
19298 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19301 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19304 * config/aarch64/aarch64.c
19305 (aarch64_vectorize_preferred_vector_alignment): Query the size
19306 of the provided SVE vector; do not assume that all SVE vectors
19307 have the same size.
19309 2021-03-31 Jan Hubicka <jh@suse.cz>
19312 * cgraph.c (cgraph_node::release_body): Remove all callers and
19314 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19315 * cgraphunit.c (cgraph_node::expand): And here.
19317 2021-03-31 Martin Liska <mliska@suse.cz>
19319 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19320 and one negated condition.
19322 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19323 Richard Sandiford <richard.sandiford@arm.com>
19326 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19327 constraints on operands[2] and similarly 0 and rk constraints
19328 on operands[1] corresponding to that.
19330 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19333 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19334 linker doesn't support DWARF sections new in DWARF5.
19335 * config/i386/i386-options.c (ix86_option_override_internal): Default
19336 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19338 * config.in: Regenerated.
19339 * configure: Regenerated.
19341 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19344 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19345 available issue_info before using it.
19347 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19350 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19353 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19356 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19357 (altivec_lvsl_reg_<mode>): ... this.
19358 (altivec_lvsr_reg): Change to ...
19359 (altivec_lvsr_reg_<mode>): ... this.
19360 * config/rs6000/predicates.md (vec_set_index_operand): New.
19361 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19362 Enable 32bit variable vec_insert for all TARGET_VSX.
19363 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19364 Enable 32bit variable vec_insert for p9 and above.
19365 (rs6000_expand_vector_set_var_p8): Rename to ...
19366 (rs6000_expand_vector_set_var_p7): ... this.
19367 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19369 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19370 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19371 gen_altivec_lvsr_reg_di.
19373 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19376 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19377 (__rdtscp): Likewise.
19379 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19381 PR tree-optimization/99825
19382 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19383 Reject non-mult 2 lanes.
19385 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19388 * config/arm/arm.c (arm_file_start): Fix emission of
19389 Tag_ABI_VFP_args attribute.
19391 2021-03-30 Richard Biener <rguenther@suse.de>
19393 PR tree-optimization/99824
19394 * stor-layout.c (set_min_and_max_values_for_integral_type):
19395 Assert the precision is within the bounds of
19396 WIDE_INT_MAX_PRECISION.
19397 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19398 the outermost component ref only to lower the access size
19399 and initialize that from the access type.
19401 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19404 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19405 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19407 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19409 * config/aarch64/aarch64.md
19410 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19411 attribute to disambiguate between SIMD and FP variants of the
19414 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19416 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19417 (analyze_ssa_name_flags): Fix typo in comment.
19419 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19422 * config/aarch64/aarch64-sve-builtins.cc
19423 (function_builder::add_function): Add placeholder_p argument, use
19424 placeholder decls if this is set.
19425 (function_builder::add_unique_function): Instead of conditionally adding
19426 direct overloads, unconditionally add either a direct overload or a
19428 (function_builder::add_overloaded_function): Set placeholder_p if we're
19429 using C++ overloads. Use the obstack for string storage instead
19430 of relying on the tree nodes.
19431 (function_builder::add_overloaded_functions): Don't return early for
19432 m_direct_overloads: we need to add placeholders.
19433 * config/aarch64/aarch64-sve-builtins.h
19434 (function_builder::add_function): Add placeholder_p argument.
19436 2021-03-29 Richard Biener <rguenther@suse.de>
19438 PR tree-optimization/99807
19439 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19440 assert below VEC_PERM handling.
19442 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19445 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19446 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19447 matching const_int 0.
19448 (move_lo_quad_internal_be_<mode>): Likewise.
19449 (move_lo_quad_<mode>): Update for the above.
19450 * config/aarch64/iterators.md (VQ_2E): Delete.
19452 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19454 PR tree-optimization/99777
19455 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19456 types other than scalar integral types.
19458 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19460 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19461 XCOFF TLS reloc decorations.
19463 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19465 * doc/analyzer.texi (Analyzer Internals): Update link to
19466 "A Memory Model for Static Analysis of C Programs".
19468 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19470 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19471 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19473 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19474 (rs6000_special_round_type_align): Recursively check innermost first
19477 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19480 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19481 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19482 assignment with drap_reg active, queue reg save for hfp with offset 0
19483 and flush queued reg saves. When handling a push with rule18,
19484 defer queueing reg save for hfp and just assert the offset is 0.
19485 (scan_trace): Assert that fde->rule18 is false.
19487 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19490 * ira-costs.c (record_reg_classes): Put case with
19491 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19492 * ira.c (ira_setup_alts): Ditto.
19493 * lra-constraints.c (process_alt_operands): Ditto.
19494 * recog.c (asm_operand_ok): Ditto.
19495 * reload.c (find_reloads): Ditto.
19497 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19499 * config/aarch64/aarch64-protos.h
19500 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19501 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19502 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19503 accordingly, using the same costs as for post_modify.
19504 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19505 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19506 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19507 (a64fx_addrcost_table): Likewise.
19508 (neoversev1_addrcost_table): New.
19509 (neoversev1_tunings): Use neoversev1_addrcost_table.
19510 (aarch64_address_cost): Use the new post_modify costs for CImode
19513 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19515 * config/aarch64/aarch64.opt
19516 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19517 * doc/invoke.texi: Document it.
19518 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19519 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19520 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19521 (aarch64_vec_issue_info): New structures.
19522 (cpu_vector_cost): Write comments above the variables rather
19524 (cpu_vector_cost::issue_info): New member variable.
19525 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19526 and tree-ssa-loop-niter.h.
19527 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19528 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19529 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19530 (thunderx3t110_vector_cost): Initialize issue_info to null.
19531 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19532 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19533 (neoversev1_vector_cost): Use them.
19534 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19535 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19536 (aarch64_vector_costs::num_vector_iterations): Likewise.
19537 (aarch64_vector_costs::scalar_ops): Likewise.
19538 (aarch64_vector_costs::advsimd_ops): Likewise.
19539 (aarch64_vector_costs::sve_ops): Likewise.
19540 (aarch64_vector_costs::seen_loads): Likewise.
19541 (aarch64_simd_vec_costs_for_flags): New function.
19542 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19543 Count the number of predicate operations required by SVE WHILE
19545 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19546 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19547 (aarch64_count_ops): Likewise.
19548 (aarch64_add_stmt_cost): Record whether see an SVE operation
19549 that cannot currently be implementing using Advanced SIMD.
19550 Record issue information about the scalar, Advanced SIMD
19551 and (where relevant) SVE versions of a loop.
19552 (aarch64_vec_op_count::dump): New function.
19553 (aarch64_sve_op_count::dump): Likewise.
19554 (aarch64_estimate_min_cycles_per_iter): Likewise.
19555 (aarch64_adjust_body_cost): If issue information is available,
19556 try to compare the issue rates of the various loop implementations
19557 and increase or decrease the vector body cost accordingly.
19559 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19561 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19562 Assume a zero cost for induction phis.
19564 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19566 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19568 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19569 vector comparisons.
19571 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19573 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19575 (aarch64_add_stmt_cost): Call it.
19577 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19579 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19580 New tuning parameter.
19581 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19582 (aarch64_estimated_sve_vq): New function.
19583 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19584 (aarch64_vector_costs::is_loop): Likewise.
19585 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19586 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19587 (aarch64_record_potential_advsimd_unrolling): New function.
19588 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19589 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19590 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19591 Detect whether we're vectorizing a loop for SVE that might be
19592 completely unrolled if it used Advanced SIMD instead.
19593 (aarch64_adjust_body_cost_for_latency): New function.
19594 (aarch64_finish_cost): Call it.
19596 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19598 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
19599 (aarch64_init_cost): New function.
19600 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
19601 the default unsigned[3].
19602 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
19603 (TARGET_VECTORIZE_INIT_COST): Override.
19604 (TARGET_VECTORIZE_FINISH_COST): Likewise.
19605 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
19607 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19609 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
19610 (neoversev1_sve_vector_cost): New cost structures.
19611 (neoversev1_vector_cost): Likewise.
19612 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
19614 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19616 * config/aarch64/aarch64-protos.h
19617 (sve_vec_cost::scatter_store_elt_cost): New member variable.
19618 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19619 accordingly, taking the cost from the cost of a scalar_store.
19620 (a64fx_sve_vector_cost): Likewise.
19621 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
19623 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19625 * config/aarch64/aarch64-protos.h
19626 (simd_vec_cost::store_elt_extra_cost): New member variable.
19627 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19628 accordingly, using the vec_to_scalar cost for the new field.
19629 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19630 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19631 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19632 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19633 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19634 (thunderx3t110_advsimd_vector_cost): Likewise.
19635 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
19637 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19639 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
19640 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
19641 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
19642 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19643 accordingly, using zero for the new costs.
19644 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19645 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19646 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19647 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19648 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19649 (thunderx3t110_advsimd_vector_cost): Likewise.
19650 (aarch64_ld234_st234_vectors): New function.
19651 (aarch64_adjust_stmt_cost): Likewise.
19652 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
19653 the new vector costs.
19655 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19657 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
19658 derived class of simd_vec_cost. Add information about CLAST[AB]
19659 and FADDA instructions.
19660 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19661 accordingly, using the vec_to_scalar costs for the new fields.
19662 (a64fx_sve_vector_cost): Likewise.
19663 (aarch64_reduc_type): New function.
19664 (aarch64_sve_in_loop_reduction_latency): Likewise.
19665 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
19666 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
19667 that occur in the loop body.
19668 (aarch64_add_stmt_cost): Update call accordingly.
19670 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19672 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
19674 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
19675 above the fields rather than to the right.
19676 (simd_vec_cost::reduc_i8_cost): New member variable.
19677 (simd_vec_cost::reduc_i16_cost): Likewise.
19678 (simd_vec_cost::reduc_i32_cost): Likewise.
19679 (simd_vec_cost::reduc_i64_cost): Likewise.
19680 (simd_vec_cost::reduc_f16_cost): Likewise.
19681 (simd_vec_cost::reduc_f32_cost): Likewise.
19682 (simd_vec_cost::reduc_f64_cost): Likewise.
19683 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19684 accordingly, using the vec_to_scalar_cost for the new fields.
19685 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19686 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19687 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19688 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19689 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19690 (thunderx3t110_advsimd_vector_cost): Likewise.
19691 (aarch64_use_new_vector_costs_p): New function.
19692 (aarch64_simd_vec_costs): New function, split out from...
19693 (aarch64_builtin_vectorization_cost): ...here.
19694 (aarch64_is_reduction): New function.
19695 (aarch64_detect_vector_stmt_subtype): Likewise.
19696 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
19697 using the new vector costs.
19699 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19702 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
19703 TLS declarations as public.
19705 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19707 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
19708 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
19709 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
19710 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
19711 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
19712 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
19713 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
19714 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
19715 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
19717 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19720 * config.gcc (*-*-cygwin*): Add winnt-d.o
19721 (*-*-mingw*): Likewise.
19722 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
19723 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
19724 * config/i386/t-cygming: Add winnt-d.o.
19725 * config/i386/winnt-d.c: New file.
19727 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19729 * config/freebsd-d.c: Include memmodel.h.
19731 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19734 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
19735 * config/t-openbsd: Add openbsd-d.o.
19736 * config/openbsd-d.c: New file.
19738 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
19740 PR tree-optimization/96974
19741 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
19742 with graceful exit.
19744 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19747 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19751 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19752 always_inline in system headers.
19754 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
19756 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
19758 2021-03-25 Jakub Jelinek <jakub@redhat.com>
19761 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
19762 * fold-const.c (operand_compare::operand_equal_p): Don't compare
19763 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
19765 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19769 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19770 always_inline in system headers.
19772 2021-03-25 Richard Biener <rguenther@suse.de>
19774 PR tree-optimization/99746
19775 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
19776 the scalar stmt as patterned. Instead set up required things
19779 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
19781 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
19784 2021-03-24 Martin Liska <mliska@suse.cz>
19787 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
19789 * config/i386/i386-options.c (ix86_option_override_internal):
19790 Add run-time assert.
19792 2021-03-24 Martin Jambor <mjambor@suse.cz>
19795 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
19796 parameters with unknown type.
19797 (ipacp_value_safe_for_type): New function.
19798 (propagate_vals_across_arith_jfunc): Verify that the constant type
19799 can be used for a type of the formal parameter.
19800 (propagate_vals_across_ancestor): Likewise.
19801 (propagate_scalar_across_jump_function): Likewise. Pass the type
19802 also to propagate_vals_across_ancestor.
19804 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
19807 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
19809 (movmisalign<mode>_mve_load): Likewise.
19811 2021-03-24 Jakub Jelinek <jakub@redhat.com>
19814 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
19815 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
19817 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19819 * doc/sourcebuild.texi (sysconf): New effective target.
19821 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19823 * config/i386/predicates.md (reg_or_const_vec_operand): New.
19824 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
19825 the now *-prefixed insn_and_split, turn the splitter const vec
19826 into an input for the insn, making it an ignored immediate for
19827 non-split cases, and loaded into the scratch register
19830 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
19833 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
19834 Use define_relaxed_memory_constraint for them.
19836 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
19839 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
19840 colon to the diagnostic message.
19842 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
19844 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
19846 (try_fwprop_subst_note): Use set_info instead of insn_info.
19847 (try_fwprop_subst_pattern): Likewise.
19848 (try_fwprop_subst_notes): Likewise.
19849 (try_fwprop_subst): Likewise.
19850 (forward_propagate_subreg): Likewise.
19851 (forward_propagate_and_simplify): Likewise.
19852 (forward_propagate_into): Likewise.
19853 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
19855 (set_info::single_nondebug_insn_use): Likewise.
19856 (set_info::single_phi_use): Likewise.
19857 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
19859 (set_info::single_nondebug_insn_use): Likewise.
19860 (set_info::single_phi_use): Likewise.
19862 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
19864 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
19866 2021-03-23 Jakub Jelinek <jakub@redhat.com>
19869 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
19870 expand_mult to perform an unsigned rather than a signed
19873 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
19876 * config/i386/cpuid.h (__cpuid): Add __volatile__.
19877 (__cpuid_count): Likewise.
19879 2021-03-23 Richard Biener <rguenther@suse.de>
19881 PR tree-optimization/99721
19882 * tree-vect-slp.c (vect_slp_analyze_node_operations):
19883 Make sure we can schedule the node.
19885 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19887 * config/riscv/riscv.c (riscv_subword): Take endianness into
19888 account when calculating the byte offset.
19890 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19892 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
19893 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
19894 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
19895 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
19896 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
19897 new predicate "subreg_lowpart_operator"
19899 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19901 * config/riscv/riscv.c (riscv_swap_instruction): New function
19902 to byteswap an SImode rtx containing an instruction.
19903 (riscv_trampoline_init): Byteswap the generated instructions
19906 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19908 * common/config/riscv/riscv-common.c
19909 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
19910 * config.gcc (riscv32be-*, riscv64be-*): Set
19911 TARGET_BIG_ENDIAN_DEFAULT to 1.
19912 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
19913 depending on default endianness.
19914 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
19915 * config/riscv/linux.h (LINK_SPEC): Likewise.
19916 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
19917 default endianness.
19918 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
19920 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19922 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
19923 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
19924 * config/riscv/linux.h (LINK_SPEC): Likewise.
19925 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
19927 (BYTES_BIG_ENDIAN): Handle big endian.
19928 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
19929 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
19931 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
19933 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
19935 * regcprop.c (find_oldest_value_reg): Ask target whether
19936 different mode is fine for replacement register.
19938 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
19940 PR tree-optimization/99296
19941 * value-range.cc (irange::irange_set_1bit_anti_range): New.
19942 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
19943 * value-range.h (irange::irange_set_1bit_anti_range): New.
19945 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
19948 * config/aarch64/constraints.md (UtQ): Use
19949 define_relaxed_memory_constraint for it.
19950 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
19951 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19952 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
19953 (have_relaxed_memory_constraints): New static var.
19954 (relaxed_memory_start, relaxed_memory_end): Ditto.
19955 (add_constraint): Add arg is_relaxed_memory. Check name for
19956 relaxed memory. Set up is_relaxed_memory in constraint_data and
19957 have_relaxed_memory_constraints. Adjust calls.
19958 (choose_enum_order): Process relaxed memory.
19959 (write_tm_preds_h): Ditto.
19960 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19961 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19962 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
19963 * ira-lives.c (single_reg_class): Use
19964 insn_extra_relaxed_memory_constraint.
19965 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
19966 * lra-constraints.c (valid_address_p): Use
19967 insn_extra_relaxed_memory_constraint instead of other memory
19969 (process_alt_operands): Process CT_RELAXED_MEMORY.
19970 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
19971 * recog.c (asm_operand_ok, preprocess_constraints): Process
19973 * reload.c (find_reloads): Ditto.
19974 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
19975 * stmt.c (parse_input_constraint): Use
19976 insn_extra_relaxed_memory_constraint.
19978 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
19981 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
19984 2021-03-22 Alex Coplan <alex.coplan@arm.com>
19987 * config/arm/arm-protos.h (neon_make_constant): Add generate
19988 argument to guard emitting insns, default to true.
19989 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
19990 CONST_VECTORs which neon_make_constant can't handle.
19991 (neon_vdup_constant): Add generate argument, avoid emitting
19992 insns if it's not set.
19993 (neon_make_constant): Plumb new generate argument through.
19994 * config/arm/constraints.md (Ui): New. Use it...
19995 * config/arm/mve.md (*mve_mov<mode>): ... here.
19996 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
19997 synthesize constants.
19999 2021-03-22 Richard Biener <rguenther@suse.de>
20001 * debug.h: Add deprecation warning.
20003 2021-03-22 Richard Biener <rguenther@suse.de>
20005 PR tree-optimization/99694
20006 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
20009 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
20012 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
20013 after type checking.
20015 2021-03-22 Jakub Jelinek <jakub@redhat.com>
20019 * dwarf2out.c (get_full_len): Use get_precision rather than
20021 (add_const_value_attribute): Make sure add_AT_wide argument has
20022 precision prec rather than some very wide one.
20024 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
20026 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
20027 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
20028 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
20029 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
20030 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
20031 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
20032 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
20033 *fix_trunc<mode>si2_internal): Fix empty split condition.
20034 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
20035 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
20036 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
20037 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
20039 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
20042 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20043 Convert idx to DImode.
20044 (rs6000_expand_vector_set_var_p8): Likewise.
20046 2021-03-21 Jakub Jelinek <jakub@redhat.com>
20049 * dwarf2out.c (insert_float): Change return type from void to
20050 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
20051 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
20054 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
20057 * config/i386/i386.c (construct_container): Check cfun != NULL
20058 before accessing silent_p.
20060 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
20062 * asan.c: Fix typos in comments.
20064 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
20066 PR rtl-optimization/99680
20067 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
20068 (process_address_1): Check empty constraint before using
20071 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
20073 * config/rs6000/rs6000.c (power10_cost): New.
20074 (rs6000_option_override_internal): Set Power10 costs.
20075 (rs6000_issue_rate): Set Power10 issue rate.
20076 * config/rs6000/power10.md: Rewrite for Power10.
20078 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
20081 * lra-constraints.c (process_address_1): Don't use unknown
20082 constraint for address constraint.
20084 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
20087 * config.gcc (powerpc-*-darwin8): Delete the reference to
20088 the now removed darwin8.h.
20090 2021-03-19 Olivier Hainque <hainque@adacore.com>
20093 * config/vxworksae.h (VX_CPU_PREFIX): Define.
20095 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
20097 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
20099 2021-03-19 Tamar Christina <tamar.christina@arm.com>
20101 PR tree-optimization/99656
20102 * tree-vect-slp-patterns.c (linear_loads_p,
20103 complex_add_pattern::matches, is_eq_or_top,
20104 vect_validate_multiplication, complex_mul_pattern::matches,
20105 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
20106 * tree-vectorizer.h: (complex_load_perm_t): Removed.
20107 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
20108 complex_load_perm_t.
20110 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
20113 * config/i386/i386-options.c (ix86_init_machine_status): Set
20115 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20117 (construct_container): Return early for return and argument
20118 errors if silent_p is true.
20119 * config/i386/i386.h (machine_function): Add silent_p.
20121 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20124 * config/arm/constraints.md (Ds): New constraint.
20125 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20126 constraint instead of w,Dm.
20128 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20130 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20133 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20135 PR middle-end/99641
20136 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20137 array type, do the computation of the current position in sizetype.
20139 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20142 * lra-constraints.c (process_address_1): Use lookup_constraint
20143 only for a single constraint.
20145 2021-03-18 Martin Sebor <msebor@redhat.com>
20147 PR middle-end/99502
20148 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20149 (inbounds_memaccess_p): ...to this. Check the ending offset of
20150 the accessed member.
20152 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20154 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20155 %> quote markers to error messages.
20156 (gcn_goacc_validate_dims): Likewise.
20157 (gcn_conditional_register_usage): Remove exclaimation mark from error
20159 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20161 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20163 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20166 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20167 Kito Cheng <kito.cheng@sifive.com>
20169 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20170 to unsigned HOST_WIDE_INT for parameter and local variable with
20171 HOST_WIDE_INT type.
20172 (riscv_adjust_block_mem): Ditto.
20173 (riscv_block_move_loop): Ditto.
20174 (riscv_expand_block_move): Ditto.
20176 2021-03-18 Nick Clifton <nickc@redhat.com>
20178 * config/v850/v850.c (construct_restore_jr): Increase static
20180 (construct_save_jarl): Likewise.
20181 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20183 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20185 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20186 (aarch64_override_options_internal): Use it.
20187 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20190 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20192 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20193 error message format issues.
20194 (nios2_option_override): Likewise.
20195 (nios2_expand_fpu_builtin): Likewise.
20196 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20197 truncation warning.
20198 (nios2_expand_custom_builtin): More error message format fixes.
20199 (nios2_expand_rdwrctl_builtin): Likewise.
20200 (nios2_expand_rdprs_builtin): Likewise.
20201 (nios2_expand_eni_builtin): Likewise.
20202 (nios2_expand_builtin): Likewise.
20203 (nios2_register_custom_code): Likewise.
20204 (nios2_valid_target_attribute_rec): Likewise.
20205 (nios2_add_insn_asm): Fix uninitialized variable warning.
20207 2021-03-17 Jan Hubicka <jh@suse.cz>
20209 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20210 of gather to match reality.
20211 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20213 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20215 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20216 to compare against CC_REG rather than NE.
20218 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20221 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20222 inline assembly statements.
20223 (ix86_print_operand): Update 'P' handling for -fno-plt.
20225 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20228 * config/aarch64/aarch64.c
20229 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20231 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20234 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20235 ordered and unordered for CCFPmode, if flag_finite_math_only.
20237 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20240 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
20241 rather than ASHIFT.
20242 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
20244 2021-03-16 Martin Liska <mliska@suse.cz>
20247 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
20248 cl_optimization_compare function.
20250 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
20252 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
20255 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20258 * config/i386/i386.h (struct machine_function): Add
20259 has_explicit_vzeroupper bitfield.
20260 * config/i386/i386-expand.c (ix86_expand_builtin): Set
20261 cfun->machine->has_explicit_vzeroupper when expanding
20262 IX86_BUILTIN_VZEROUPPER.
20263 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
20264 Do the mode switching only when TARGET_VZEROUPPER, expensive
20265 optimizations turned on and not optimizing for size.
20266 (pass_insert_vzeroupper::gate): Enable even when
20267 cfun->machine->has_explicit_vzeroupper is set.
20269 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20272 * config/aarch64/aarch64.c
20273 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20274 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20275 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20277 2021-03-15 Richard Biener <rguenther@suse.de>
20279 PR tree-optimization/98834
20280 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20281 subsetting by truncating the access size.
20283 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20285 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20286 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20289 2021-03-15 Martin Liska <mliska@suse.cz>
20291 * spellcheck.c: Add missing comma in initialization.
20293 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20295 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20296 alternative 2 and alternative 1 with alternative 3 using
20297 YW register constraint.
20298 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20299 using YW register constraint.
20300 (*vec_extractv16qi_zext): Ditto.
20301 (*vec_extractv4si): Merge alternatives 4 and 5
20302 using Yw register constraint.
20303 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20305 2021-03-13 Martin Sebor <msebor@redhat.com>
20307 PR tree-optimization/99489
20308 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20309 is not a call statement.
20311 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20313 PR tree-optimization/99544
20314 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20315 if for vector types multiplication can't be done in type's mode.
20317 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20320 * config/sparc/constraints.md (w): Rename to...
20321 (W): ... this and ditch previous implementation.
20322 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20323 (*movdf_insn_sp64): Likewise.
20324 (*mov<VM64:mode>_insn_sp64): Likewise.
20325 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20327 (atomic_compare_and_swap_leon3_1): Likewise.
20328 (*atomic_compare_and_swapdi_v8plus): Likewise.
20329 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20330 architecture and add missing address validity check during LRA.
20332 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20335 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20336 occuring for assumed-size arrays in use_device_{ptr,addr}.
20338 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20341 * config/i386/constraints.md (YW): New internal constraint.
20342 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20343 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20344 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20345 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20347 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20348 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20349 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20350 into one, use Yw instead of former x,v.
20351 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20352 the last alternative.
20353 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20354 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20355 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20356 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20357 into one, use <v_Yw> instead of former x,v.
20358 (avx2_interleave_highv32qi<mask_name>,
20359 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20360 constraints. Add && <mask_avx512bw_condition> to condition.
20361 (avx2_interleave_lowv32qi<mask_name>,
20362 vec_interleave_lowv16qi<mask_name>,
20363 avx2_interleave_highv16hi<mask_name>,
20364 vec_interleave_highv8hi<mask_name>,
20365 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20366 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20367 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20368 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20369 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20370 Yw instead of v in constraints.
20371 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20372 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20373 instead of Yv in constraints.
20374 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20375 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20376 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20377 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20378 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20380 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20381 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20382 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20383 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20384 two, one with just x, another isa avx512vl with v.
20386 2021-03-12 Martin Liska <mliska@suse.cz>
20388 * doc/invoke.texi: Add missing param documentation.
20390 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20393 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20394 analyzer/trimmed-graph.o.
20395 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20396 feasibility checking to reflect new implementation.
20397 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20399 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20401 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20403 * digraph.cc (selftest::test_shortest_paths): Update
20404 shortest_paths init for new param. Add test of
20405 SPS_TO_GIVEN_TARGET.
20406 * shortest-paths.h (enum shortest_path_sense): New.
20407 (shortest_paths::shortest_paths): Add "sense" param.
20408 Update for renamings. Generalize to use "sense" param.
20409 (shortest_paths::get_shortest_path): Rename param.
20410 (shortest_paths::m_sense): New field.
20411 (shortest_paths::m_prev): Rename...
20412 (shortest_paths::m_best_edge): ...to this.
20413 (shortest_paths::get_shortest_path): Update for renamings.
20414 Conditionalize flipping of path on sense of traversal.
20416 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20418 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20419 for paths from B and C.
20420 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20421 unreachable nodes, rather than asserting.
20423 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20426 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20427 xcoff_tbss_section_name.
20428 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20429 * xcoffout.c (xcoff_tbss_section_name): Delete.
20430 * xcoffout.h (xcoff_tbss_section_name): Delete.
20432 2021-03-11 Richard Biener <rguenther@suse.de>
20434 PR tree-optimization/99523
20435 * tree-cfg.c (dump_function_to_file): Dump SSA names
20436 w/o identifier to the decls section as well, not only those
20437 without a VAR_DECL.
20439 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20442 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20443 function calls with lhs fail if the lhs don't have compatible types.
20445 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20447 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20448 Change FRAME_POINTER_REGNUM to correspond to a new faked
20449 register faked_fp, part of GENNONACR_REGS like faked_ap.
20450 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20451 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20452 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20453 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20454 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20456 (CRIS_REAL_FP_REGNUM): New constant.
20457 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20458 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20459 (cris_initial_elimination_offset): Handle elimination changes
20460 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20461 and add one from FRAME_POINTER_REGNUM to
20462 HARD_FRAME_POINTER_REGNUM.
20463 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20464 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20466 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20469 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20470 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20472 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20475 * lra-constraints.c (process_address_1): Don't check unknown
20476 constraint, use X for empty constraint.
20478 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20480 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20481 Fix typo in comment describing "is_ha" argument.
20483 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20485 * doc/sourcebuild.texi: Document LRA target selector.
20487 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20489 * doc/ux.texi: Add subsection contrasting interactive versus
20490 batch usage of GCC.
20492 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20495 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20497 (vectorizable_load): Fix gather load mask check condition.
20499 2021-03-10 Richard Biener <rguenther@suse.de>
20501 PR tree-optimization/99510
20502 * tree.c (check_aligned_type): Check that the candidate
20503 has TYPE_USER_ALIGN set instead of matching with the
20506 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20508 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20509 float and vector integer modes only if the mode is not larger.
20511 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20513 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20515 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20517 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20519 * ira-lives.c (single_reg_class): Ditto.
20521 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20523 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20524 the architecture-specific rtems.h.
20525 (aarch64-*-rtems*): Likewise.
20526 (arm*-*-rtems*): Likewise.
20527 (epiphany-*-rtems*): Likewise.
20528 (riscv*-*-rtems*): Likewise.
20530 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20532 PR tree-optimization/99305
20533 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20534 before integer_all_onesp instead of vice versa.
20536 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20538 * common/config/arm/arm-common.c (arm_config_default): Change type
20539 of 'i' to unsigned.
20541 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20544 * lra-constraints.c (process_address_1): Process constraint 'g'
20545 separately and digital constraints containing more one digit.
20547 2021-03-09 Nick Clifton <nickc@redhat.com>
20549 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20550 (DWARF"_DEBUGGING_INFO): Define.
20552 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20555 * calls.c (initialize_argument_information): When the argument
20556 is passed by reference, do not make a copy in a thunk only if
20557 the argument is already in memory. Remove redundant test for
20558 the case of callee copy.
20560 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20563 * lra-constraints.c (process_address_1): Process 0..9 constraints
20564 in process_address_1.
20566 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20568 * config/s390/s390.c (struct s390_processor processor_table):
20569 Binutils name string must not be empty.
20571 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20573 * config/arc/arc.c (arc_attr_type): Remove function.
20575 2021-03-09 Martin Liska <mliska@suse.cz>
20578 * config/i386/i386-options.c (ix86_option_override_internal):
20579 Set isa_flags for OPTS argument and not for the global
20582 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20584 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20587 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20590 * config/rs6000/predicates.md (ds_form_mem_operand) New
20592 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20593 ds_form_mem_operand in ld/lwa patterns.
20594 * config/rs6000/fusion.md: Regenerate file.
20596 2021-03-08 Martin Sebor <msebor@redhat.com>
20598 PR middle-end/98266
20599 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
20600 (array_bounds_checker::check_array_bounds): Call it.
20602 2021-03-08 Martin Sebor <msebor@redhat.com>
20604 PR middle-end/97631
20605 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
20606 (handle_builtin_stxncpy_strncat): Rename locals. Determine
20607 destination size from allocation calls. Issue a more appropriate
20609 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
20610 (handle_builtin_memset): Same.
20612 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
20615 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
20616 to ensure we do not have an Altivec style address.
20617 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
20618 an Altivec style address.
20619 (*vsx_le_perm_store_<mode>): Likewise.
20620 (splitters after *vsx_le_perm_store_<mode>): Likewise.
20621 (vsx_load_<mode>): Disable special expander if passed an Altivec
20623 (vsx_store_<mode>): Likewise.
20625 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20628 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
20629 (aarch64_simd_shift_imm_vec_hi): Likewise.
20630 (aarch64_simd_shift_imm_vec_si): Likewise.
20631 (aarch64_simd_shift_imm_vec_di): Likewise.
20632 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
20633 predicate from above.
20634 (aarch64_shrn<mode>_insn_be): Likewise.
20635 (aarch64_rshrn<mode>_insn_le): Likewise.
20636 (aarch64_rshrn<mode>_insn_be): Likewise.
20637 (aarch64_shrn2<mode>_insn_le): Likewise.
20638 (aarch64_shrn2<mode>_insn_be): Likewise.
20639 (aarch64_rshrn2<mode>_insn_le): Likewise.
20640 (aarch64_rshrn2<mode>_insn_be): Likewise.
20642 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
20645 * lra-constraints.c (skip_contraint_modifiers): New function.
20646 (process_address_1): Use it before lookup_constraint call.
20648 2021-03-08 Martin Liska <mliska@suse.cz>
20651 * config/i386/i386-options.c (ix86_option_override_internal):
20652 Enable UINTR and HRESET for -march that supports it.
20654 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
20656 * config/s390/s390.c (f_constraint_p): New function.
20657 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
20658 (TARGET_MD_ASM_ADJUST): Likewise.
20660 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
20663 * tree-nested.c (convert_local_reference_stmt): Avoid calling
20664 lookup_field_for_decl for Fortran module (= namespace context).
20666 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
20668 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
20669 comparison with arithmetic right shift.
20670 (s390_expand_vcond): No need for a force_reg anymore.
20671 s390_vec_compare will do it.
20672 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
20673 immediate operands.
20675 2021-03-07 Jakub Jelinek <jakub@redhat.com>
20678 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
20679 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
20681 * config/i386/sse.md (v_Yw): New define_mode_attr.
20682 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
20683 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
20685 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
20686 xYw in constraints.
20688 2021-03-06 Julian Brown <julian@codesourcery.com>
20690 * tree-pretty-print.c (dump_generic_node): Emit non-generic
20691 address space info for aggregates.
20693 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
20695 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
20697 2021-03-05 Jakub Jelinek <jakub@redhat.com>
20699 PR middle-end/99322
20700 * tree-cfg.c (bb_to_omp_idx): New variable.
20701 (execute_build_cfg): Release the bb_to_omp_idx vector after
20702 cleanup_tree_cfg returns.
20703 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
20704 for bb_to_omp_idx being a vec<int> instead of pointer to array
20706 (make_edges): Remove bb_to_omp_idx local variable, don't pass
20707 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
20708 vec<int> instead of pointer to array of ints and don't free/release
20710 (remove_bb): When removing a bb and placing forced label somewhere
20711 else, ensure it is put into the same OpenMP region during cfg
20712 pass if possible or to entry successor as fallback. Unregister
20713 bb from bb_to_omp_idx.
20715 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
20718 * lra-constraints.c (process_address_1): Skip decomposing address
20719 for asm insn operand with unknown constraint.
20721 2021-03-05 Martin Jambor <mjambor@suse.cz>
20724 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
20725 corresponding speculative edges if we are about to resolve
20726 sepculation. Make edge direct (and so resolve speculations) before
20727 removing it from call_site_hash.
20728 (cgraph_edge::make_direct): Relax the initial assert to allow calling
20729 the function on speculative direct edges.
20731 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
20733 PR rtl-optimization/99376
20734 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
20735 of low-order zero bits is too large, set the result to 0 directly.
20737 2021-03-04 Jakub Jelinek <jakub@redhat.com>
20739 PR middle-end/93235
20740 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
20741 SUBREG and a SUBREG to op_mode can't be created.
20743 2021-03-04 Alex Coplan <alex.coplan@arm.com>
20746 * config/aarch64/aarch64-sve-builtins.cc
20747 (function_resolver::require_vector_type): Handle error_mark_node.
20749 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
20751 * cfgexpand.c (expand_asm_loc): Pass new parameter.
20752 (expand_asm_stmt): Likewise.
20753 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
20755 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20756 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
20757 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
20758 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
20759 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
20760 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
20761 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
20762 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
20763 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
20764 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
20765 * doc/tm.texi (md_asm_adjust): Likewise.
20766 * target.def (md_asm_adjust): Likewise.
20768 2021-03-04 Richard Biener <rguenther@suse.de>
20770 PR middle-end/97855
20771 * tree-pretty-print.c: Poison pp_printf.
20772 (dump_decl_name): Avoid use of pp_printf.
20773 (dump_block_node): Likewise.
20774 (dump_generic_node): Likewise.
20776 2021-03-04 Martin Sebor <msebor@redhat.com>
20778 PR middle-end/96963
20779 PR middle-end/94655
20780 * builtins.c (handle_array_ref): New helper.
20781 (handle_mem_ref): New helper.
20782 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
20783 into new helper functions. Correct a workaround for vectorized
20786 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
20788 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
20789 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
20790 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
20791 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
20792 attribute for Power10.
20793 * config/rs6000/mma.md (*movoo): Likewise.
20794 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
20795 (define_mode_attr bits): Add DD/TD modes.
20796 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
20797 store_conditionalpti): Update size attribute for Power10.
20799 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20802 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
20803 -Wuninitialized, -Wmaybe-uninitialized.
20804 (wide-int.o-warn): Likewise.
20806 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
20808 * common/config/arm/arm-common.c: Include configargs.h.
20809 (arm_config_default): New function.
20810 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
20811 processors that do not support Thumb. Take into account the
20812 --with-mode configuration setting for selecting the default.
20813 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
20814 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
20816 2021-03-03 Martin Liska <mliska@suse.cz>
20818 PR gcov-profile/97461
20819 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
20821 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
20824 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20825 point back the hard frame pointer to its default location when the
20826 frame is larger than SEH_MAX_FRAME_SIZE.
20828 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20831 * config/i386/predicates.md (logic_operator): New define_predicate.
20832 * config/i386/i386.md (mov + mem using comm arith peephole2):
20833 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
20834 and the inner mode is [QH]Imode.
20836 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20839 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
20840 (new_loc_list): Clear end_entry.
20841 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
20842 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
20844 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
20845 initialize also end_entry.
20847 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20850 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
20851 partitions, if in non-layout mode after reorder_blocks also move
20852 affected blocks to ensure a single partition transition.
20854 2021-03-03 Jason Merrill <jason@redhat.com>
20857 * cgraphunit.c (process_function_and_variable_attributes): Don't
20858 warn about flatten on an alias if the target also has it.
20859 * cgraph.h (symtab_node::get_alias_target_tree): New.
20861 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
20863 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
20864 period to symbol name.
20865 (tls_get_addr_internal<mode>): Same.
20867 2021-03-02 David Malcolm <dmalcolm@redhat.com>
20870 * diagnostic-show-locus.c
20871 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
20874 2021-03-02 Martin Sebor <msebor@redhat.com>
20876 PR middle-end/99276
20877 * builtins.c (warn_for_access): Remove stray warning text.
20879 2021-03-02 Martin Sebor <msebor@redhat.com>
20881 PR middle-end/99295
20882 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
20885 2021-03-02 Jakub Jelinek <jakub@redhat.com>
20888 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
20889 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
20890 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
20891 DW_MACRO_define_strx and DW_MACRO_undef_strx.
20892 (save_macinfo_strings): Use DW_MACRO_*_str* even with
20893 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
20894 DW_MACRO_undef_strx.
20896 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
20898 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
20900 (BT_FN_V8HI_V8HI_UINT): Likewise.
20901 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
20902 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
20903 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
20904 New builtin definitions.
20905 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
20906 vector extension version.
20907 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
20908 available with current -march level.
20909 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
20910 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
20911 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
20912 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
20913 (vec_extend_to_fp32_lo): Likewise.
20914 (vec_round_from_fp32): Likewise.
20915 (vec_convert_to_fp16): Likewise.
20916 (vec_convert_from_fp16): Likewise.
20917 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
20918 (vclfnls_v8hi): Likewise.
20919 (vcrnfs_v8hi): Likewise.
20920 (vcfn_v8hi): Likewise.
20921 (vcnf_v8hi): Likewise.
20923 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
20925 * common/config/s390/s390-common.c (processor_flags_table): New entry.
20926 * config.gcc: Enable arch14 for --with-arch and --with-tune.
20927 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
20928 arch14 for unknown CPU models.
20929 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
20930 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
20931 (s390_get_sched_attrmask): Likewise.
20932 (s390_get_unit_mask): Likewise.
20933 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
20934 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
20935 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
20936 (TARGET_NNPA_P): New macro definitions.
20937 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
20938 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
20940 2021-03-02 Jakub Jelinek <jakub@redhat.com>
20942 PR middle-end/95757
20943 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
20944 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
20945 != 1 comparisons if name is lhs of a comparison.
20947 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
20951 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
20952 * config/darwin.c (darwin_should_restore_cfa_state): New.
20953 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
20954 * doc/tm.texi: Regenerated.
20955 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
20956 * dwarf2cfi.c (connect_traces): If the target requests, restore
20957 the CFA expression after a DW_CFA_restore.
20958 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
20960 2021-03-01 Martin Liska <mliska@suse.cz>
20963 * optc-save-gen.awk: Add 4 more exceptions.
20965 2021-03-01 Nathan Sidwell <nathan@acm.org>
20968 * tree.h (TYPE_ALIGN_RAW): New accessor.
20969 (TYPE_ALIGN): Use it.
20971 2021-03-01 Jan Hubicka <jh@suse.cz>
20974 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
20976 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
20979 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20980 point the hard frame pointer to the SSE register save area instead
20981 of the general register save area. Perform only minimal adjustment
20982 for small frames if it is initially not correctly aligned.
20983 (ix86_expand_prologue): Remove early saves for a SEH target.
20984 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
20986 2021-02-28 Jakub Jelinek <jakub@redhat.com>
20989 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
20990 typo - referneced -> referenced.
20991 * tree.c (component_ref_size): Fix comment typo -
20992 refernce -> reference.
20993 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
20994 traling -> trailing.
20995 (aliasing_component_refs_p): Fix comment typos -
20996 refernce -> reference and refernece -> reference and
20997 traling -> trailing.
20998 (nonoverlapping_refs_since_match_p): Fix comment typo -
20999 referneces -> references.
21000 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
21001 referneces -> references.
21003 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
21005 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
21006 diagnostic message to avoid use of a contraction and format
21009 2021-02-27 Jakub Jelinek <jakub@redhat.com>
21012 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
21013 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
21014 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
21015 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
21018 2021-02-26 Richard Biener <rguenther@suse.de>
21020 PR middle-end/99281
21021 * expr.c (store_field): For calls with return-slot optimization
21022 and addressable return type expand the store directly.
21024 2021-02-26 Richard Biener <rguenther@suse.de>
21027 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
21029 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
21032 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
21035 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
21037 * config.gcc: Add rs6000-pcrel-opt.o.
21038 * config/rs6000/rs6000-pcrel-opt.c: New file.
21039 * config/rs6000/pcrel-opt.md: New file.
21040 * config/rs6000/predicates.md: Add d_form_memory predicate.
21041 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
21042 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
21043 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
21044 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
21045 and make_pass_pcrel_opt().
21046 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
21047 (rs6000_option_override_internal): Add pcrel-opt.
21048 (rs6000_delegitimize_address): Support pcrel-opt.
21049 (rs6000_opt_masks): Add pcrel-opt.
21050 (pcrel_opt_valid_mem_p): New function.
21051 (reg_to_non_prefixed): Make global.
21052 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
21053 (output_pcrel_opt_reloc): New function.
21054 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
21055 (pcrel_extern_addr): Set loads_extern_addr.
21056 Add include for pcrel-opt.md.
21057 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
21058 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
21061 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
21064 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
21065 If TARGET_64BIT and dest is SUBREG, we check the width, if it
21066 equal to SImode, we use SImode operation, just like what we are
21069 2021-02-26 Marek Polacek <polacek@redhat.com>
21071 * builtins.c (warn_for_access): Fix typos.
21073 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
21075 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
21076 mark in front of the immediate quantity.
21077 (<optab>_rolsi3_uxtw): Likewise.
21079 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
21082 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
21083 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
21084 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
21085 address and disable when the FPCXT is not available.
21086 (nonsecure_call_value_reg_thumb2): Likewise.
21088 2021-02-25 Nathan Sidwell <nathan@acm.org>
21091 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
21093 2021-02-25 Tamar Christina <tamar.christina@arm.com>
21095 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
21097 2021-02-25 Richard Biener <rguenther@suse.de>
21099 PR tree-optimization/99253
21100 * tree-vect-loop.c (check_reduction_path): First compute
21101 code, then verify out-of-loop uses.
21103 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21106 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
21108 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21110 PR tree-optimization/80635
21111 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
21112 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
21113 has mode precision.
21115 2021-02-25 Richard Biener <rguenther@suse.de>
21117 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21118 load_map population.
21119 (vect_match_slp_patterns_2): Revert part of last change.
21120 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21121 with pattern detection but do it afterwards. Dump the
21122 whole SLP graph after pattern recognition and load
21123 redistribution optimization finished.
21125 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21128 * omp-low.c (struct omp_context): Add teams_nested_p and
21129 nonteams_nested_p members.
21130 (scan_omp_target): Diagnose teams nested inside of target with other
21131 directives strictly nested inside of the same target.
21132 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21133 ctx->nonteams_nested_p as needed.
21135 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21137 PR inline-asm/99123
21138 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21140 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21142 * config/cris/cris.c (cris_expand_prologue): Set
21143 current_function_static_stack_size, if flag_stack_usage_info.
21145 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21147 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21148 (rs6000_final_prescan_insn): Adjust.
21149 (rs6000_asm_output_opcode): Likewise.
21151 2021-02-24 Martin Sebor <msebor@redhat.com>
21153 PR middle-end/97172
21154 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21155 from function arguments.
21157 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21159 PR tree-optimization/99220
21160 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21161 node from cache when it's about to be deleted.
21163 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21165 PR tree-optimization/99225
21166 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21167 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21168 build_int_cst (..., 1). Formatting fixes.
21170 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21172 PR tree-optimization/99149
21173 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21175 (vect_slp_reset_pattern): Remove.
21176 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21177 (complex_mul_pattern::build, complex_fma_pattern::build,
21178 complex_fms_pattern::build): Fix ref counts.
21179 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21180 when node is being deleted.
21181 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21182 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21184 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21186 2021-02-24 Matthias Klose <doko@ubuntu.com>
21189 2020-12-07 Matthias Klose <doko@ubuntu.com>
21191 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21192 and ENABLE_RTL_FLAG_CHECKING.
21194 2021-02-24 Richard Biener <rguenther@suse.de>
21197 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21199 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21201 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21202 (vsx_assemble_pair): ...to this.
21203 (*mma_assemble_pair): Rename from this...
21204 (*vsx_assemble_pair): ...to this.
21205 (mma_disassemble_pair): Rename from this...
21206 (vsx_disassemble_pair): ...to this.
21207 (*mma_disassemble_pair): Rename from this...
21208 (*vsx_disassemble_pair): ...to this.
21209 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21210 BU_COMPAT): New macros.
21211 (mma_assemble_pair): Rename from this...
21212 (vsx_assemble_pair): ...to this.
21213 (mma_disassemble_pair): Rename from this...
21214 (vsx_disassemble_pair): ...to this.
21215 (mma_assemble_pair): New compatibility built-in.
21216 (mma_disassemble_pair): Likewise.
21217 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21218 (RS6000_BUILTIN_COMPAT): Define.
21219 (bdesc_compat): New.
21220 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21221 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21222 and VSX_BUILTIN_ASSEMBLE_PAIR.
21223 (rs6000_init_builtins): Register compatibility built-ins.
21224 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21225 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21226 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21227 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21228 (__builtin_vsx_assemble_pair): ...to this.
21229 (__builtin_mma_disassemble_pair): Rename from this...
21230 (__builtin_vsx_disassemble_pair): ...to this.
21232 2021-02-23 Martin Liska <mliska@suse.cz>
21235 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
21236 with different alignment. That leads to an invalid red zone
21237 size allocated in runtime.
21239 2021-02-23 Jakub Jelinek <jakub@redhat.com>
21241 PR tree-optimization/99204
21242 * fold-const.c (fold_read_from_constant_string): Check that
21243 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
21245 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
21246 Kewen Lin <linkw@gcc.gnu.org>
21248 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
21249 (rotl<mode>3_insert_3): ...this.
21250 (plus_ior_xor): New code_iterator.
21251 (define_split for GPR rl*imi): New splitter.
21252 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
21253 for integer merging.
21255 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21257 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
21259 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
21260 into a register when the above is enabled.
21261 * config/aarch64/aarch64.c (neoversev1_tunings):
21262 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21263 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21265 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21267 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21268 valid operand from is now an addi mult-value to shift-value.
21269 * config/cris/cris.md (*addi): Change expression of scaled
21270 operand from mult to ashift.
21271 * config/cris/cris.md (*addi_reload): New insn_and_split.
21273 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21276 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21277 hook_bool_const_tree_hwi_hwi_const_tree_true.
21278 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21280 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21282 PR rtl-optimization/98791
21283 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21284 for unordered modes.
21286 2021-02-22 Martin Liska <mliska@suse.cz>
21288 * tree-inline.c (inline_forbidden_p): Set
21289 inline_forbidden_reason.
21291 2021-02-22 Richard Biener <rguenther@suse.de>
21293 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21296 2021-02-22 Richard Biener <rguenther@suse.de>
21298 PR tree-optimization/99165
21299 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21300 Accumulate changed to ret.
21302 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21305 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21307 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21309 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21312 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21314 (trunctf<DFP_ALL:mode>2): Likewise.
21315 (trunctdtf2_vr): Likewise.
21316 (trunctdtf2): Likewise.
21317 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21318 (extend<DFP_ALL:mode>tf2): Likewise.
21319 (extendtftd2_vr): Likewise.
21320 (extendtftd2): Likewise.
21322 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21324 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21325 add memory alternative.
21326 (tf_to_fprx2): New pattern.
21328 2021-02-19 Martin Sebor <msebor@redhat.com>
21331 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21332 (attr_access::free_lang_data): Remove a spurious test.
21334 2021-02-19 Nathan Sidwell <nathan@acm.org>
21336 * doc/invoke.texi (flang-info-module-read): Document.
21338 2021-02-19 Martin Liska <mliska@suse.cz>
21340 PR translation/99167
21341 * params.opt: Fix typo.
21343 2021-02-19 Richard Biener <rguenther@suse.de>
21345 PR middle-end/99122
21346 * tree-inline.c (inline_forbidden_p): Do not inline functions
21347 with VLA arguments or return value.
21349 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21352 * config/arm/arm.md (*stack_protect_combined_set_insn,
21353 *stack_protect_combined_test_insn): If force_const_mem result
21354 is not valid general operand, force its address into the destination
21357 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21360 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21361 pad or non-local label, put FORCED_LABELs from bb b after that label
21362 rather than before it.
21364 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21367 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21368 expand_vector_broadcast' to emit the vec_duplicate operand.
21370 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21372 PR rtl-optimization/96264
21373 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21376 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21379 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21380 looking up the retain attribute.
21381 (resolve_unique_section): Likewise.
21382 (get_variable_section): Likewise.
21383 (switch_to_section): Likewise. Warn when a symbol without the
21384 retain attribute and a symbol with the retain attribute are
21385 placed in the section with the same name, instead of the used
21387 * doc/extend.texi: Document the "retain" attribute.
21389 2021-02-18 Nathan Sidwell <nathan@acm.org>
21392 * doc/invoke.texi (flang-info-include-translate): Document header
21395 2021-02-18 Richard Biener <rguenther@suse.de>
21397 PR middle-end/99122
21398 * ipa-fnsummary.c (analyze_function_body): Set
21399 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21400 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21401 error_mark_node values.
21402 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21404 (setup_one_parameter): Delay force_value_to_type until when
21407 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21409 PR tree-optimization/99142
21410 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21412 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21414 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21415 wide_int_bitmask::wide_int_bitmask (uint64_t),
21416 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21417 wide_int_bitmask::operator ~ () const,
21418 wide_int_bitmask::operator | (wide_int_bitmask) const,
21419 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21421 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21422 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21423 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21424 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21425 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21426 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21427 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21428 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21429 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21430 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21431 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21432 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21433 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21434 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21435 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21436 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21437 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21438 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21439 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21440 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21441 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21442 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21443 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21444 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21446 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21448 PR middle-end/99109
21449 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21450 (build_printable_array_type): ... this. Add nelts argument. For
21451 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21452 nelts, call build_array_type_nelts.
21453 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21454 instead of build_zero_elt_array_type and build_array_type_nelts.
21456 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21459 * config/i386/i386.c (distance_non_agu_define): Don't call
21460 extract_insn_cached here.
21461 (ix86_lea_outperforms): Save and restore recog_data around call
21462 to distance_non_agu_define and distance_agu_use.
21463 (ix86_ok_to_clobber_flags): Remove.
21464 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21465 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21466 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21467 into define_insn. Move the splitting to define_peephole2 and
21468 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21470 2021-02-17 Julian Brown <julian@codesourcery.com>
21472 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21475 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21478 * config/mips/mips.c (mips_symbol_insns): Do not use
21479 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21481 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21483 PR inline-asm/98096
21484 * stmt.c (resolve_operand_name_1): Take inout operands into account
21485 for access to labels by names.
21486 * doc/extend.texi: Describe counting operands for accessing labels.
21488 2021-02-16 Richard Biener <rguenther@suse.de>
21490 PR tree-optimization/38474
21491 * tree-ssa-structalias.c (variable_info::address_taken): New.
21492 (new_var_info): Initialize address_taken.
21493 (process_constraint): Set address_taken.
21494 (solve_constraints): Use the new address_taken flag rather
21495 than is_reg_var for sorting variables.
21496 (dump_constraint): Dump the variable number if the name
21499 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21502 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21503 multiply by 4096 and for inbranch by 8192.
21504 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21505 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21507 2021-02-15 Maya Rashish <coypu@sdf.org>
21509 * config/aarch64/aarch64.c (aarch64_init_builtins):
21510 Call SUBTARGET_INIT_BUILTINS.
21512 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21514 PR rtl-optimization/98872
21515 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21516 if CONST0_RTX is NULL.
21518 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21520 PR rtl-optimization/98863
21521 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21522 (function_info::build_info): Turn into a declaration, moving the
21523 definition to internals.h.
21524 (function_info::bb_walker): Declare.
21525 (function_info::create_reg_use): Likewise.
21526 (function_info::calculate_potential_phi_regs): Take a build_info
21528 (function_info::place_phis, function_info::create_ebbs): Declare.
21529 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21530 (function_info::populate_backedge_phis): Delete.
21531 (function_info::start_block, function_info::end_block): Declare.
21532 (function_info::populate_phi_inputs): Delete.
21533 (function_info::m_potential_phi_regs): Move information to build_info.
21534 * rtl-ssa/internals.h: New file.
21535 (function_info::bb_phi_info): New class.
21536 (function_info::build_info): Moved from functions.h.
21537 Add a constructor and destructor.
21538 (function_info::build_info::ebb_use): Delete.
21539 (function_info::build_info::ebb_def): Likewise.
21540 (function_info::build_info::bb_live_out): Likewise.
21541 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21542 (function_info::build_info::potential_phi_regs): Likewise.
21543 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21544 (function_info::build_info::ebb_def_regs): Likewise.
21545 (function_info::build_info::bb_phis): Likewise.
21546 (function_info::build_info::bb_mem_live_out): Likewise.
21547 (function_info::build_info::bb_to_rpo): Likewise.
21548 (function_info::build_info::def_stack): Likewise.
21549 (function_info::build_info::old_def_stack_limit): Likewise.
21550 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21551 Remove the regno argument. Push the previous definition onto the
21552 definition stack where necessary.
21553 * rtl-ssa/accesses.cc: Include internals.h.
21554 * rtl-ssa/changes.cc: Likewise.
21555 * rtl-ssa/blocks.cc: Likewise.
21556 (function_info::build_info::build_info): Define.
21557 (function_info::build_info::~build_info): Likewise.
21558 (function_info::bb_walker): New class.
21559 (function_info::bb_walker::bb_walker): Define.
21560 (function_info::add_live_out_use): Convert a logarithmic-complexity
21561 test into a linear one. Allow the same definition to be passed
21563 (function_info::calculate_potential_phi_regs): Moved from
21564 functions.cc. Take a build_info parameter and store the
21565 information there instead.
21566 (function_info::place_phis): New function.
21567 (function_info::add_entry_block_defs): Update call to record_reg_def.
21568 (function_info::calculate_ebb_live_in_for_debug): New function.
21569 (function_info::add_phi_nodes): Use bb_phis to decide which
21570 registers need phi nodes and initialize ebb_def_regs accordingly.
21571 Do not add degenerate phis here.
21572 (function_info::add_artificial_accesses): Use create_reg_use.
21573 Assert that all definitions are listed in the DF LR sets.
21574 Update call to record_reg_def.
21575 (function_info::record_block_live_out): Record live-out register
21576 values in the phis of successor blocks. Use the live-out set
21577 when processing the last block in an EBB, instead of always
21578 using the live-in sets of successor blocks. AND the live sets
21579 with the set of registers that have been defined in the EBB,
21580 rather than with all potential phi registers. Cope correctly
21581 with branches back to the start of the current EBB.
21582 (function_info::start_block): New function.
21583 (function_info::end_block): Likewise.
21584 (function_info::populate_phi_inputs): Likewise.
21585 (function_info::create_ebbs): Likewise.
21586 (function_info::process_all_blocks): Rewrite into a multi-phase
21588 * rtl-ssa/functions.cc: Include internals.h.
21589 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21590 (function_info::init_function_data): Remove caller.
21591 * rtl-ssa/insns.cc: Include internals.h
21592 (function_info::create_reg_use): New function. Lazily any
21593 degenerate phis needed by the linear RPO view.
21594 (function_info::record_use): Use create_reg_use. When processing
21595 debug uses, use potential_phi_regs and test it before checking
21596 whether the register is live on entry to the current EBB. Lazily
21597 calculate ebb_live_in_for_debug.
21598 (function_info::record_call_clobbers): Update call to record_reg_def.
21599 (function_info::record_def): Likewise.
21601 2021-02-15 Martin Liska <mliska@suse.cz>
21603 * toplev.c (init_asm_output): Free output of
21604 gen_command_line_string function.
21605 (process_options): Likewise.
21607 2021-02-15 Martin Liska <mliska@suse.cz>
21609 * params.opt: Add 2 missing Param keywords.
21611 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
21613 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
21615 2021-02-15 Jakub Jelinek <jakub@redhat.com>
21617 PR tree-optimization/99079
21618 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
21619 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
21620 require both type and TREE_TYPE (@1) to be integral types and either
21621 type having smaller or equal precision, or TREE_TYPE (@1) being
21622 unsigned type, or type being signed type. If TREE_TYPE (@1)
21623 doesn't have wrapping overflow, perform the subtraction of one in
21626 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
21627 Richard Biener <rguether@suse.de>
21630 * ipa-reference.c (ipa_init): Only conditinally initialize
21631 reference_vars_to_consider.
21632 (propagate): Conditionally deninitialize reference_vars_to_consider.
21633 (ipa_reference_write_optimization_summary): Sanity check that
21634 reference_vars_to_consider is not allocated.
21636 2021-02-13 Levy Hsu <admin@levyhsu.com>
21639 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
21640 extend parameter to get_si_mem_base_reg declaration.
21641 (get_si_mem_base_reg): Add extend parameter. Set it.
21642 (analyze): Pass extend arg to get_si_mem_base_reg.
21643 (transform): Likewise. Use it when rewriting mems.
21644 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
21645 loads and emit sign/zero extending load followed by subreg move.
21647 2021-02-13 Jim Wilson <jimw@sifive.com>
21650 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
21651 exit when !reload_completed. Only perform check for compressed reg
21652 if reload_completed.
21653 (riscv_rtx_costs): In MEM case, when optimizing for size and
21654 shorten memrefs, if not compressible, then increase cost.
21656 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21658 PR rtl-optimization/98439
21659 * recog.c (pass_split_before_regstack::gate): Enable even when
21660 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
21663 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21666 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
21667 swap of V2SImode elements in memory into DImode memory rotate by 32.
21669 2021-02-12 Martin Sebor <msebor@redhat.com>
21671 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
21673 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21675 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
21676 m_temp_obstack rather than m_obstack to allocate the temporary use.
21678 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21680 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
21681 as read-modify operations.
21683 2021-02-12 Richard Biener <rguenther@suse.de>
21685 PR middle-end/38474
21686 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
21687 fbi->aa_walk_budget is bigger than zero. Update
21688 fbi->aa_walk_budget.
21689 (param_change_prob): Likewise.
21690 * ipa-prop.c (detect_type_change_from_memory_writes):
21691 Properly account walk_aliased_vdefs.
21692 (parm_preserved_before_stmt_p): Canonicalize updates.
21693 (parm_ref_data_preserved_p): Likewise.
21694 (parm_ref_data_pass_through_p): Likewise.
21695 (determine_known_aggregate_parts): Account own alias queries.
21697 2021-02-12 Martin Liska <mliska@suse.cz>
21699 * opts-common.c (decode_cmdline_option): Release werror_arg.
21700 * opts.c (gen_producer_string): Release output of
21701 gen_command_line_string.
21703 2021-02-12 Richard Biener <rguenther@suse.de>
21705 PR tree-optimization/38474
21706 * params.opt (-param=max-store-chains-to-track=): New param.
21707 (-param=max-stores-to-track=): Likewise.
21708 * doc/invoke.texi (max-store-chains-to-track): Document.
21709 (max-stores-to-track): Likewise.
21710 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
21712 (pass_store_merging::m_n_stores): Likewise.
21713 (pass_store_merging::terminate_and_process_chain): Update
21714 m_n_stores and m_n_chains.
21715 (pass_store_merging::process_store): Likewise. Terminate
21716 oldest chains if the number of stores or chains get too large.
21717 (imm_store_chain_info::terminate_and_process_chain): Dump
21720 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
21722 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
21723 the cold section, emit a nop before the directive if the previous
21724 active instruction can throw.
21726 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
21729 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
21730 memory addresses that are legal for quad word accesses.
21732 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
21735 * config/arm/thumb2.md (*doloop_end_internal): Generate
21736 alternative sequence to handle long range branches.
21738 2021-02-11 Joel Hutton <joel.hutton@arm.com>
21740 PR tree-optimization/98772
21741 * optabs-tree.c (supportable_half_widening_operation): New function
21742 to check for supportable V8QI->V8HI widening patterns.
21743 * optabs-tree.h (supportable_half_widening_operation): New function.
21744 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
21745 to create promotion stmts for V8QI->V8HI widening patterns.
21746 (vectorizable_conversion): Add case for V8QI->V8HI.
21748 2021-02-11 Richard Biener <rguenther@suse.de>
21750 * sparseset.h (SPARSESET_ELT_BITS): Remove.
21751 (SPARSESET_ELT_TYPE): Use unsigned int.
21752 * fwprop.c: Do not include sparseset.h.
21754 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21757 * varasm.c (declare_weak): For -fsyntax-only, allow even
21758 TREE_ASM_WRITTEN function decls.
21760 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21763 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
21764 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
21765 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
21766 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
21767 calling simplify_gen_subreg on it.
21769 2021-02-10 Martin Liska <mliska@suse.cz>
21771 * config/nvptx/nvptx.c (nvptx_option_override): Use
21772 flag_patchable_function_entry instead of the removed
21773 function_entry_patch_area_size.
21775 2021-02-10 Martin Liska <mliska@suse.cz>
21777 PR tree-optimization/99002
21778 PR tree-optimization/99026
21779 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
21780 leak when adjacent cases are merged.
21781 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
21783 (make_pass_lower_switch): Remove trailing whitespace.
21784 * tree-switch-conversion.h (release_clusters): New.
21786 2021-02-10 Richard Biener <rguenther@suse.de>
21788 PR rtl-optimization/99054
21789 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
21790 (fixup_partitions): Adjust.
21791 (rtl_verify_edges): Likewise.
21793 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21795 PR middle-end/99007
21796 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
21797 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
21800 2021-02-10 Richard Biener <rguenther@suse.de>
21803 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
21806 2021-02-10 Richard Biener <rguenther@suse.de>
21808 PR tree-optimization/99024
21809 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
21810 clear loop->aux if it is associated with the destroyed loop_vinfo.
21812 2021-02-10 Martin Liska <mliska@suse.cz>
21814 PR tree-optimization/99002
21815 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
21818 2021-02-10 Martin Liska <mliska@suse.cz>
21821 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
21822 a reference exists.
21824 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21827 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
21828 at class scope for DWARF5+.
21830 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
21832 PR rtl-optimization/96015
21833 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
21834 (relax_delay_slots): When deleting a jump to the next active
21835 instruction over a barrier, first delete the barrier if the
21836 jump is the only way to reach the target label.
21838 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
21840 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
21841 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
21842 vector multiplies and vect.alu for SSRA.
21843 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
21844 vect.mul cost field.
21845 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
21846 * config/arm/arm.c: Likewise.
21848 2021-02-09 Richard Biener <rguenther@suse.de>
21850 PR tree-optimization/98863
21851 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
21852 * tree-ssa-sccvn.c (last_pushed_avail): New global.
21853 (rpo_elim::eliminate_push_avail): Chain pushed avails.
21854 (unwind_state::avail_top): Add.
21855 (do_unwind): Rewrite unwinding of avail entries.
21856 (do_rpo_vn): Initialize last_pushed_avail and
21857 avail_top of the undo state.
21859 2021-02-09 Jakub Jelinek <jakub@redhat.com>
21861 PR middle-end/99004
21862 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
21863 const char * to char * and free those pointers after use.
21865 2021-02-09 Richard Biener <rguenther@suse.de>
21867 PR tree-optimization/99017
21868 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
21869 zero vector cost entries.
21871 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
21873 PR middle-end/98974
21874 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
21875 parameter in vectorizable_condition.
21877 2021-02-08 Richard Biener <rguenther@suse.de>
21880 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
21882 2021-02-08 Martin Liska <mliska@suse.cz>
21885 * cfgexpand.c (pass_expand::execute): Parse per-function option
21886 flag_patchable_function_entry and use it.
21887 * common.opt: Remove function_entry_patch_area_size and
21888 function_entry_patch_area_start global variables.
21889 * opts.c (parse_and_check_patch_area): New function.
21890 (common_handle_option): Use it.
21891 * opts.h (parse_and_check_patch_area): New function.
21892 * toplev.c (process_options): Parse and use
21893 function_entry_patch_area_size.
21895 2021-02-08 Martin Sebor <msebor@redhat.com>
21897 * doc/extend.texi (attribute malloc): Correct typos.
21899 2021-02-05 Nathan Sidwell <nathan@acm.org>
21902 * gcc.c (driver::maybe_run_linker): Check for input file
21903 accessibility if not linking.
21905 2021-02-05 Richard Biener <rguenther@suse.de>
21907 PR tree-optimization/98855
21908 * tree-vectorizer.h (add_stmt_cost): New overload.
21909 * tree-vect-slp.c (li_cost_vec_cmp): New.
21910 (vect_bb_slp_scalar_cost): Cost individual loop regions
21911 separately. Account for the scalar instance root stmt.
21913 2021-02-05 Tom de Vries <tdevries@suse.de>
21916 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
21918 (bit_test_cluster::emit): Reuse location_t for newly created
21920 (switch_decision_tree::try_switch_expansion): Preserve
21922 * tree-switch-conversion.h: Change function signatures.
21924 2021-02-05 Jakub Jelinek <jakub@redhat.com>
21927 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
21928 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
21929 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
21930 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
21932 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21934 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
21935 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
21936 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
21937 (vget_high_f16): Reimplement using new builtin.
21938 (vget_high_f32): Likewise.
21939 (vget_high_f64): Likewise.
21940 (vget_high_p8): Likewise.
21941 (vget_high_p16): Likewise.
21942 (vget_high_p64): Likewise.
21943 (vget_high_s8): Likewise.
21944 (vget_high_s16): Likewise.
21945 (vget_high_s32): Likewise.
21946 (vget_high_s64): Likewise.
21947 (vget_high_u8): Likewise.
21948 (vget_high_u16): Likewise.
21949 (vget_high_u32): Likewise.
21950 (vget_high_u64): Likewise.
21952 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21954 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
21955 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
21956 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
21957 (vget_low_f16): Reimplement using new builtin.
21958 (vget_low_f32): Likewise.
21959 (vget_low_f64): Likewise.
21960 (vget_low_p8): Likewise.
21961 (vget_low_p16): Likewise.
21962 (vget_low_p64): Likewise.
21963 (vget_low_s8): Likewise.
21964 (vget_low_s16): Likewise.
21965 (vget_low_s32): Likewise.
21966 (vget_low_s64): Likewise.
21967 (vget_low_u8): Likewise.
21968 (vget_low_u16): Likewise.
21969 (vget_low_u32): Likewise.
21970 (vget_low_u64): Likewise.
21972 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
21974 * gcc.c (print_multilib_info): Check all required argument is provided
21977 2021-02-05 liuhongt <hongtao.liu@intel.com>
21980 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
21981 generate integer mask comparison for 128/256-bits vector when
21982 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
21983 delete redundant !maskcmp condition.
21984 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
21986 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
21987 condition directly to if (maskcmp), add extra check for
21988 cmpmode, it should be MODE_INT.
21989 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
21990 parameters op_true/op_false.
21991 (ix86_use_mask_cmp_p): New.
21993 2021-02-05 liuhongt <hongtao.liu@intel.com>
21996 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
21997 Remove m_GENERIC from ~list.
21998 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
22000 2021-02-04 David Malcolm <dmalcolm@redhat.com>
22003 * diagnostic-show-locus.c (compatible_locations_p): Require
22004 locations in the same macro map to be either both from the
22005 macro definition, or both from the macro arguments.
22007 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22009 * config/aarch64/aarch64-simd-builtins.def: Add
22010 [su]mull_hi_lane[q] builtin generator macros.
22011 * config/aarch64/aarch64-simd.md
22012 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
22013 (aarch64_<su>mull_hi_lane<mode>): Define.
22014 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
22015 (aarch64_<su>mull_hi_laneq<mode>): Define.
22016 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
22017 builtin instead of inline asm.
22018 (vmull_high_lane_s32): Likewise.
22019 (vmull_high_lane_u16): Likewise.
22020 (vmull_high_lane_u32): Likewise.
22021 (vmull_high_laneq_s16): Likewise.
22022 (vmull_high_laneq_s32): Likewise.
22023 (vmull_high_laneq_u16): Likewise.
22024 (vmull_high_laneq_u32): Liekwise.
22026 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22028 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
22029 builtin generator macros.
22030 * config/aarch64/aarch64-simd.md
22031 (aarch64_<su>mull_hi_n<mode>_insn): Define.
22032 (aarch64_<su>mull_hi_n<mode>): Define.
22033 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
22034 instead of inline asm.
22035 (vmull_high_n_s32): Likewise.
22036 (vmull_high_n_u16): Likewise.
22037 (vmull_high_n_u32): Likewise.
22039 2021-02-04 Richard Biener <rguenther@suse.de>
22041 PR tree-optimization/98855
22042 * tree-vect-loop.c (vectorizable_phi): Do not cost
22043 single-argument PHIs.
22044 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
22045 * tree-vect-stmts.c (vectorizable_bswap): Also perform
22046 costing for SLP operation.
22048 2021-02-04 Martin Liska <mliska@suse.cz>
22050 * doc/extend.texi: Mention -mprefer-vector-width in target
22053 2021-02-03 Martin Sebor <msebor@redhat.com>
22055 PR tree-optimization/98937
22056 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
22057 Flush pointer_query cache.
22059 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22061 * config/rs6000/genfusion.pl (gen_2logical): Add missing
22062 fixes based on patch review.
22063 * config/rs6000/fusion.md: Regenerate file.
22065 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22067 * config/rs6000/t-rs6000: Comment out auto generation of
22070 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
22072 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
22073 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
22074 (output_file_start): Add gfx908.
22075 * config/gcn/gcn.opt (gpu_type): Add gfx908.
22076 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
22077 (MULTILIB_DIRNAMES): Add gfx908.
22078 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
22079 (main): Recognize gfx908.
22080 * config/gcn/t-omp-device: Add gfx908.
22082 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22084 * config/aarch64/aarch64-simd-builtins.def: Add
22085 [su]mlsl_hi_lane[q] builtin macro generators.
22086 * config/aarch64/aarch64-simd.md
22087 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
22088 (aarch64_<su>mlsl_hi_lane<mode>): Define.
22089 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
22090 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
22091 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
22092 builtin instead of inline asm.
22093 (vmlsl_high_lane_s32): Likewise.
22094 (vmlsl_high_lane_u16): Likewise.
22095 (vmlsl_high_lane_u32): Likewise.
22096 (vmlsl_high_laneq_s16): Likewise.
22097 (vmlsl_high_laneq_s32): Likewise.
22098 (vmlsl_high_laneq_u16): Likewise.
22099 (vmlsl_high_laneq_u32): Likewise.
22100 (vmlal_high_laneq_u32): Likewise.
22102 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22104 * config/aarch64/aarch64-simd-builtins.def: Add
22105 [su]mlal_hi_lane[q] builtin generator macros.
22106 * config/aarch64/aarch64-simd.md
22107 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
22108 (aarch64_<su>mlal_hi_lane<mode>): Define.
22109 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
22110 (aarch64_<su>mlal_hi_laneq<mode>): Define.
22111 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
22112 builtin instead of inline asm.
22113 (vmlal_high_lane_s32): Likewise.
22114 (vmlal_high_lane_u16): Likewise.
22115 (vmlal_high_lane_u32): Likewise.
22116 (vmlal_high_laneq_s16): Likewise.
22117 (vmlal_high_laneq_s32): Likewise.
22118 (vmlal_high_laneq_u16): Likewise.
22119 (vmlal_high_laneq_u32): Likewise.
22121 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22123 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22124 builtin generator macros.
22125 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22127 (aarch64_<su>mlsl_hi_n<mode>): Define.
22128 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22129 instead of inline asm.
22130 (vmlsl_high_n_s32): Likewise.
22131 (vmlsl_high_n_u16): Likewise.
22132 (vmlsl_high_n_u32): Likewise.
22134 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22136 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22137 builtin generator macros.
22138 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22140 (aarch64_<su>mlal_hi_n<mode>): Define.
22141 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22142 instead of inline asm.
22143 (vmlal_high_n_s32): Likewise.
22144 (vmlal_high_n_u16): Likewise.
22145 (vmlal_high_n_u32): Likewise.
22147 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22149 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22151 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22153 (aarch64_<su>mlal_hi<mode>_insn): This.
22154 (aarch64_<su>mlal_hi<mode>): Define.
22155 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22156 instead of inline asm.
22157 (vmlal_high_s16): Likewise.
22158 (vmlal_high_s32): Likewise.
22159 (vmlal_high_u8): Likewise.
22160 (vmlal_high_u16): Likewise.
22161 (vmlal_high_u32): Likewise.
22163 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22165 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22166 after calling alter_subreg() on a (mem).
22168 2021-02-03 Martin Liska <mliska@suse.cz>
22171 * lto-streamer-out.c (produce_lto_section): Fill up missing
22173 * lto-streamer.h (struct lto_section): Add _padding field.
22175 2021-02-03 Richard Biener <rguenther@suse.de>
22177 * lto-streamer.c (lto_get_section_name): Free temporary
22179 * tree-loop-distribution.c
22180 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22182 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22184 PR middle-end/97487
22185 * ifcvt.c (noce_can_force_operand): New function.
22186 (noce_emit_move_insn): Use it.
22187 (noce_try_sign_mask): Likewise. Formatting fix.
22189 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22191 PR middle-end/97971
22192 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22193 fatal_insn, but instead return false.
22195 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22197 PR tree-optimization/98287
22198 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22201 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22203 PR tree-optimization/98928
22204 * tree-vect-loop.c (vect_analyze_loop_2): Change
22205 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22206 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22207 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22208 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22210 2021-02-02 Richard Biener <rguenther@suse.de>
22212 * gimple-loop-interchange.cc (prepare_data_references):
22214 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22215 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22216 * tree-vect-stmts.c (vectorizable_condition): Do not
22218 (vectorizable_comparison): Likewise.
22220 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22222 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22223 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22224 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22225 (vrsqrteq_u32): Likewise.
22227 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22229 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22230 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22231 (aarch64_sqxtun2<mode>_be): Likewise.
22232 (aarch64_sqxtun2<mode>): Likewise.
22233 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22234 (vqmovun_high_s32): Likewise.
22235 (vqmovun_high_s64): Likewise.
22236 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
22238 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22240 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
22242 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
22244 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22246 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
22247 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
22248 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
22249 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
22250 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
22251 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
22252 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
22253 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
22255 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22257 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
22258 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
22259 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
22261 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22263 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
22264 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
22266 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22268 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22269 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22270 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22271 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22272 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22273 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22274 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22275 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22276 NONE builtin flags.
22278 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22280 PR tree-optimization/98848
22281 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22282 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22284 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22287 * expr.c: Check mode before calling store_expr.
22289 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22291 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22293 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22294 instruction using expression ior.
22295 (mve_vornq_u<mode>): New expander.
22296 (mve_vornq_f<mode>): Use ior code instead of unspec.
22297 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22299 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22301 * tree-nested.c (convert_nonlocal_reference_op): Move
22302 current_function_decl restore after re-gimplification.
22303 (convert_local_reference_op): Likewise.
22305 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22307 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22309 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22311 (aarch64_rshrn<mode>_insn_be): Likewise.
22312 (aarch64_rshrn<mode>): Likewise.
22313 (aarch64_rshrn2<mode>_insn_le): Likewise.
22314 (aarch64_rshrn2<mode>_insn_be): Likewise.
22315 (aarch64_rshrn2<mode>): Likewise.
22316 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22317 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22319 (vrshrn_high_n_s32): Likewise.
22320 (vrshrn_high_n_s64): Likewise.
22321 (vrshrn_high_n_u16): Likewise.
22322 (vrshrn_high_n_u32): Likewise.
22323 (vrshrn_high_n_u64): Likewise.
22324 (vrshrn_n_s16): Likewise.
22325 (vrshrn_n_s32): Likewise.
22326 (vrshrn_n_s64): Likewise.
22327 (vrshrn_n_u16): Likewise.
22328 (vrshrn_n_u32): Likewise.
22329 (vrshrn_n_u64): Likewise.
22331 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22333 PR tree-optimization/98499
22334 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22335 conservatively and assume all possible side-effects.
22337 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22339 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22340 vec_unpacku_hi_): Define builtins.
22341 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22343 (vmovl_high_s16): Likewise.
22344 (vmovl_high_s32): Likewise.
22345 (vmovl_high_u8): Likewise.
22346 (vmovl_high_u16): Likewise.
22347 (vmovl_high_u32): Likewise.
22349 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22351 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22353 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22355 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22357 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22359 (vabdl_s16): Likewise.
22360 (vabdl_s32): Likewise.
22361 (vabdl_u8): Likewise.
22362 (vabdl_u16): Likewise.
22363 (vabdl_u32): Likewise.
22364 * config/aarch64/iterators.md (ABDL): New int iterator.
22365 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22367 2021-02-01 Martin Sebor <msebor@redhat.com>
22369 * tree.h (BLOCK_VARS): Add comment.
22370 (BLOCK_SUBBLOCKS): Same.
22371 (BLOCK_SUPERCONTEXT): Same.
22372 (BLOCK_ABSTRACT_ORIGIN): Same.
22373 (inlined_function_outer_scope_p): Same.
22375 2021-02-01 Martin Sebor <msebor@redhat.com>
22377 PR middle-end/97172
22378 * attribs.c (attr_access::free_lang_data): Define new function.
22379 * attribs.h (attr_access::free_lang_data): Declare new function.
22381 2021-02-01 Richard Biener <rguenther@suse.de>
22383 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22385 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22387 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22389 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22390 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22392 2021-02-01 Richard Biener <rguenther@suse.de>
22394 PR rtl-optimization/98863
22395 * config/i386/i386-features.c (convert_scalars_to_vector):
22396 Set DF_RD_PRUNE_DEAD_DEFS.
22398 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22400 * system.h (SIZE_MAX): Define if not already defined.
22402 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22404 * config/rs6000/genfusion.pl (gen_2logical): New function to
22405 generate patterns for logical-logical fusion.
22406 * config/rs6000/fusion.md: Regenerated patterns.
22407 * config/rs6000/rs6000-cpus.def: Add
22408 OPTION_MASK_P10_FUSION_2LOGICAL.
22409 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22410 Enable logical-logical fusion for p10.
22411 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22413 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22415 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22417 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22419 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22420 (mabi=vec-default): New.
22421 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22422 __EXTABI__ for AIX Vector extended ABI.
22423 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22425 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22427 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22429 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22431 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22432 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22434 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22437 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22440 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22442 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22443 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22444 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22446 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22448 PR tree-optimization/98866
22449 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22450 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22451 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22452 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22453 (gori_map::is_export_p): Ditto.
22454 (gori_map::calculate_gori): Ditto.
22455 (gori_compute::set_range_invariant): New.
22456 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22457 invariant for pointers evaluating to [1, +INF].
22459 2021-01-29 Richard Biener <rguenther@suse.de>
22461 PR rtl-optimization/98863
22462 * config/i386/i386-features.c (remove_partial_avx_dependency):
22463 Do not perform DF analysis.
22464 (pass_data_remove_partial_avx_dependency): Remove
22467 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22469 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22470 builtin generator macros.
22471 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22473 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22474 instead of inline asm.
22475 (vmull_n_s32): Likewise.
22476 (vmull_n_u16): Likewise.
22477 (vmull_n_u32): Likewise.
22479 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22481 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22483 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22485 (aarch64_<sur>abdl2<mode>): ... This.
22486 (<sur>sadv16qi): Adjust use of above.
22487 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22489 (vabdl_high_s16): Likewise.
22490 (vabdl_high_s32): Likewise.
22491 (vabdl_high_u8): Likewise.
22492 (vabdl_high_u16): Likewise.
22493 (vabdl_high_u32): Likewise.
22495 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22497 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22499 (uabal2): Likewise.
22500 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22502 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22504 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22506 (vabal_high_s16): Likewise.
22507 (vabal_high_s32): Likewise.
22508 (vabal_high_u8): Likewise.
22509 (vabal_high_u16): Likewise.
22510 (vabal_high_u32): Likewise.
22511 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22512 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22514 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22516 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22519 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22521 (aarch64_<sur>abal<mode>): ... This
22522 (<sur>sadv16qi): Adust use of the above.
22523 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22525 (vabal_s16): Likewise.
22526 (vabal_s32): Likewise.
22527 (vabal_u8): Likewise.
22528 (vabal_u16): Likewise.
22529 (vabal_u32): Likewise.
22531 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22533 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22535 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22537 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22539 (vaddlv_s16): Likewise.
22540 (vaddlv_u8): Likewise.
22541 (vaddlv_u16): Likewise.
22542 (vaddlvq_s8): Likewise.
22543 (vaddlvq_s16): Likewise.
22544 (vaddlvq_s32): Likewise.
22545 (vaddlvq_u8): Likewise.
22546 (vaddlvq_u16): Likewise.
22547 (vaddlvq_u32): Likewise.
22548 (vaddlv_s32): Likewise.
22549 (vaddlv_u32): Likewise.
22550 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22551 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22552 (Vwstype): New mode attribute.
22554 (VWIDE_S): Likewise.
22555 (USADDLV): New int iterator.
22556 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22558 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22560 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22561 builtin generator macros.
22562 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22564 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22565 instead of inline asm.
22566 (vmlsl_lane_s32): Likewise.
22567 (vmlsl_lane_u16): Likewise.
22568 (vmlsl_lane_u32): Likewise.
22569 (vmlsl_laneq_s16): Likewise.
22570 (vmlsl_laneq_s32): Likewise.
22571 (vmlsl_laneq_u16): Likewise.
22572 (vmlsl_laneq_u32): Likewise.
22574 2021-01-29 Richard Biener <rguenther@suse.de>
22576 * doc/invoke.texi (--param max-gcse-memory): Document unit
22578 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22579 * params.opt (--param max-gcse-memory): Adjust default and
22580 document unit of size.
22582 2021-01-29 Richard Biener <rguenther@suse.de>
22584 PR rtl-optimization/98863
22585 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22586 HOST_WIDE_INT for the memory estimate.
22588 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22589 Richard Biener <rguenther@suse.de>
22591 PR tree-optimization/97627
22592 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22593 Do not analyze fake edges.
22595 2021-01-29 Richard Biener <rguenther@suse.de>
22597 PR rtl-optimization/98144
22598 * df.h (df_mir_bb_info): Add con_visited member.
22599 * df-problems.c (df_mir_alloc): Initialize con_visited,
22600 do not fully populate IN and OUT.
22601 (df_mir_reset): Likewise.
22602 (df_mir_confluence_0): Set con_visited.
22603 (df_mir_confluence_n): Properly handle implicitely
22604 fully populated IN and OUT as designated by con_visited
22605 and update con_visited accordingly.
22607 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22610 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
22611 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
22612 && !TARGET_REALLY_IWMMXT to conditions.
22614 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22617 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
22620 2021-01-28 Marek Polacek <polacek@redhat.com>
22623 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
22624 the main variant, maybe reset it in its variants too.
22625 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
22626 (check_aligned_type): Check if TYPE_USER_ALIGN match.
22628 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
22631 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
22632 of constant zero for comparisons.
22634 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
22636 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
22637 support for mapping built-in function names for long double
22638 built-in functions if long double is IEEE 128-bit.
22640 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22642 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
22643 builtin generator macros.
22644 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
22646 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
22647 instead of inline asm.
22648 (vmlsl_n_s32): Likewise.
22649 (vmlsl_n_u16): Likewise.
22650 (vmlsl_n_u32): Likewise.
22652 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22654 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
22655 builtin generator macros.
22656 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
22658 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
22659 instead of inline asm.
22660 (vmlal_n_s32): Likewise.
22661 (vmlal_n_u16): Likewise.
22662 (vmlal_n_u32): Likewise.
22664 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22666 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
22668 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
22670 (aarch64_shrn2<mode>_insn_be): Likewise.
22671 (aarch64_shrn2<mode>): Likewise.
22672 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
22674 (vshrn_high_n_s32): Likewise.
22675 (vshrn_high_n_s64): Likewise.
22676 (vshrn_high_n_u16): Likewise.
22677 (vshrn_high_n_u32): Likewise.
22678 (vshrn_high_n_u64): Likewise.
22680 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22682 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
22684 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
22686 (aarch64_shrn<mode>_insn_be): Likewise.
22687 (aarch64_shrn<mode>): Likewise.
22688 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
22690 (vshrn_n_s32): Likewise.
22691 (vshrn_n_s64): Likewise.
22692 (vshrn_n_u16): Likewise.
22693 (vshrn_n_u32): Likewise.
22694 (vshrn_n_u64): Likewise.
22695 * config/aarch64/iterators.md (vn_mode): New mode attribute.
22697 2021-01-28 Richard Biener <rguenther@suse.de>
22699 PR rtl-optimization/80960
22700 * dse.c (check_mem_read_rtx): Call get_addr on the
22703 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
22704 David Edelsohn <dje.gcc@gmail.com>
22707 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22708 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
22710 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22712 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
22713 wrapper call rs6000_expand_vector_set_var for cleanup. Call
22714 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
22716 (rs6000_expand_vector_set_var): Delete.
22717 (rs6000_expand_vector_set_var_p9): Make static.
22718 (rs6000_expand_vector_set_var_p8): Make static.
22720 2021-01-28 Xing GUO <higuoxing@gmail.com>
22722 * common/config/riscv/riscv-common.c
22723 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
22724 when `p` extension exists.
22726 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
22728 PR rtl-optimization/97684
22729 * ira.c (ira): Call ira_set_pseudo_classes before
22730 update_equiv_regs when it is necessary.
22732 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22735 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
22736 %w0, %w1 and %2 instead of %0, %1 and %2.
22738 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
22740 * config/rs6000/genfusion.pl: New script to generate
22741 define_insn_and_split patterns so combine can arrange fused
22742 instructions next to each other.
22743 * config/rs6000/fusion.md: New file, generated fused instruction
22744 patterns for combine.
22745 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
22746 (non_update_memory_operand): New predicate.
22747 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
22748 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
22750 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
22752 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22753 Automatically set OPTION_MASK_P10_FUSION and
22754 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
22755 (rs600_opt_masks): Allow -mpower10-fusion
22756 in function attributes.
22757 (address_is_non_pfx_d_or_x): New function.
22758 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
22759 * config/rs6000/rs6000.md: Include fusion.md.
22760 * config/rs6000/rs6000.opt: Add -mpower10-fusion
22761 and -mpower10-fusion-ld-cmpi.
22762 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
22764 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22766 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
22767 builtin generator macros.
22768 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
22770 (aarch64_<su>mlal<mode>): This.
22771 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
22772 instead of inline asm.
22773 (vmlal_s16): Likewise.
22774 (vmlal_s32): Likewise.
22775 (vmlal_u8): Likewise.
22776 (vmlal_u16): Likewise.
22777 (vmlal_u32): Likewise.
22779 2021-01-27 Richard Biener <rguenther@suse.de>
22781 PR tree-optimization/98854
22782 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
22783 PHIs from scalars when the number of CTORs matches the
22784 number of children.
22786 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22788 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
22790 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
22792 (aarch64_mls_n<mode>): This.
22793 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
22795 (vmls_n_s32): Likewise.
22796 (vmls_n_u16): Likewise.
22797 (vmls_n_u32): Likewise.
22798 (vmlsq_n_s16): Likewise.
22799 (vmlsq_n_s32): Likewise.
22800 (vmlsq_n_u16): Likewise.
22801 (vmlsq_n_u32): Likewise.
22803 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22805 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
22807 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
22809 (vmls_s16): Likewise.
22810 (vmls_s32): Likewise.
22811 (vmls_u8): Likewise.
22812 (vmls_u16): Likewise.
22813 (vmls_u32): Likewise.
22814 (vmlsq_s8): Likewise.
22815 (vmlsq_s16): Likewise.
22816 (vmlsq_s32): Likewise.
22817 (vmlsq_u8): Likewise.
22818 (vmlsq_u16): Likewise.
22819 (vmlsq_u32): Likewise.
22821 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22823 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
22825 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
22827 (aarch64_mla_n<mode>): This.
22828 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
22830 (vmla_n_s32): Likewise.
22831 (vmla_n_u16): Likewise.
22832 (vmla_n_u32): Likewise.
22833 (vmlaq_n_s16): Likewise.
22834 (vmlaq_n_s32): Likewise.
22835 (vmlaq_n_u16): Likewise.
22836 (vmlaq_n_u32): Likewise.
22838 2021-01-27 liuhongt <hongtao.liu@intel.com>
22841 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
22842 (*sse2_eq<mode>3): Ditto.
22844 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22846 * tree-pass.h (PROP_trees): Rename to ...
22847 (PROP_gimple): ... this.
22848 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
22849 * passes.c (execute_function_dump, execute_function_todo,
22850 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
22851 * varpool.c (ctor_for_folding): Likewise.
22853 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22855 PR tree-optimization/97260
22856 * varpool.c: Include tree-pass.h.
22857 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
22858 non-TREE_SIDE_EFFECTS automatic variables.
22860 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
22862 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
22864 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
22865 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
22866 (gen_compile_unit_die): Recognise C++23.
22868 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22871 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
22874 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22877 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
22878 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
22879 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
22881 2021-01-26 Richard Biener <rguenther@suse.de>
22883 * gimple-pretty-print.c (dump_binary_rhs): Handle
22884 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
22886 2021-01-26 Richard Biener <rguenther@suse.de>
22888 PR middle-end/98726
22889 * tree.h (vector_cst_int_elt): Remove.
22890 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
22893 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
22895 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
22896 for V64DFmode min/max reductions.
22898 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22900 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
22901 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
22902 two halves, one with x and the other with const0_rtx, ordered
22903 depending on endianity.
22905 2021-01-26 Alexandre Oliva <oliva@adacore.com>
22907 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
22908 temporaries not seen in binding block, and not about to be
22909 added as gimple variables.
22911 2021-01-25 Martin Sebor <msebor@redhat.com>
22914 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
22916 2021-01-25 Martin Liska <mliska@suse.cz>
22918 * value-prof.c (get_nth_most_common_value): Use %s instead
22921 2021-01-25 Jakub Jelinek <jakub@redhat.com>
22924 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
22925 readelf -wi is able to read the emitted .debug_info back.
22926 * configure: Regenerated.
22928 2021-01-25 Martin Liska <mliska@suse.cz>
22930 PR gcov-profile/98739
22931 * common.opt: Add missing sign symbol.
22932 * value-prof.c (get_nth_most_common_value): Restore handling
22933 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
22934 PROFILE_REPRODUCIBILITY_MULTITHREADED.
22936 2021-01-25 Richard Biener <rguenther@suse.de>
22938 PR middle-end/98807
22939 * tree.c (vector_element_bits): Always use precision of
22940 the element type for boolean vectors.
22942 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
22944 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
22945 (ENDFILE_SPEC): Evaluate qnolinkcmds.
22947 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
22949 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
22950 nostartfiles handling since this is already done by
22951 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
22952 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
22953 is already done by LINK_COMMAND_SPEC.
22954 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
22955 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
22958 2021-01-25 Jakub Jelinek <jakub@redhat.com>
22961 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
22962 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
22963 value rather than host size_t.
22964 (fold_const_call): Change type of s2 from size_t to
22965 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
22966 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
22967 instead of s2 as last argument.
22969 2021-01-25 Tamar Christina <tamar.christina@arm.com>
22971 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
22972 VCMLA_OP, VCMUL_OP): New.
22973 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
22974 * config/arm/neon.md (cmul<conj_op><mode>3): New.
22975 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
22976 UNSPEC_VCMUL_CONJ): New.
22977 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
22978 cml<fcmac1><conj_op><mode>4): New.
22980 2021-01-23 Jakub Jelinek <jakub@redhat.com>
22983 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
22985 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
22987 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
22989 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
22991 (vmla_s16): Likewise.
22992 (vmla_s32): Likewise.
22993 (vmla_u8): Likewise.
22994 (vmla_u16): Likewise.
22995 (vmla_u32): Likewise.
22996 (vmlaq_s8): Likewise.
22997 (vmlaq_s16): Likewise.
22998 (vmlaq_s32): Likewise.
22999 (vmlaq_u8): Likewise.
23000 (vmlaq_u16): Likewise.
23001 (vmlaq_u32): Likewise.
23003 2021-01-22 David Malcolm <dmalcolm@redhat.com>
23005 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
23008 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23011 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
23012 filenames to emit, still emit the required 0 index directory and
23013 filename entries that match DW_AT_comp_dir and DW_AT_name of the
23016 2021-01-22 Marek Polacek <polacek@redhat.com>
23019 * doc/invoke.texi: Update C++ ABI Version 15 description.
23021 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23023 PR tree-optimization/98766
23024 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
23025 comparing against type size with param_avoid_fma_max_bits.
23027 2021-01-22 Richard Biener <rguenther@suse.de>
23029 PR middle-end/98793
23030 * tree.c (vector_element_bits): Key single-bit bool vector on
23031 integer mode rather than not vector mode.
23033 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23036 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23037 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
23039 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
23040 to call different path for P8 and P9.
23041 (rs6000_expand_vector_set_var_p9): New function.
23042 (rs6000_expand_vector_set_var_p8): New function.
23044 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23048 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23049 Ajdust variable index vec_insert from address dereference to
23050 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
23051 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23053 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
23055 2021-01-22 Martin Liska <mliska@suse.cz>
23057 PR gcov-profile/98739
23058 * profile.c (compute_value_histograms): Drop time profile for
23059 -fprofile-reproducible=multithreaded.
23061 2021-01-22 Nathan Sidwell <nathan@acm.org>
23063 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
23066 2021-01-22 Richard Biener <rguenther@suse.de>
23068 PR middle-end/98773
23069 * tree-data-ref.c (initalize_matrix_A): Revert previous
23070 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
23072 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23074 PR tree-optimization/90248
23075 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
23076 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
23078 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
23079 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
23081 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23083 PR tree-optimization/98255
23084 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
23085 extend index - low_bound from sizetype's precision rather than index
23087 (get_addr_base_and_unit_offset_1): Likewise.
23088 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
23089 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
23091 2021-01-22 Richard Biener <rguenther@suse.de>
23093 PR tree-optimization/98786
23094 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
23095 adding new uses of abnormals. Verify we deal with a conditional
23098 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23101 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
23103 2021-01-22 liuhongt <hongtao.liu@intel.com>
23107 * config/i386/sse.md (VI_128_256): New mode iterator.
23108 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
23109 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
23110 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
23111 define_insn_and_split to lower avx512 vector comparison to avx
23112 version when dest is vector.
23113 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
23114 define_insn_and_split for negating the comparison result.
23115 * config/i386/predicates.md (float_vector_all_ones_operand):
23117 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23118 general NOT operator without UNSPEC_MASKOP.
23120 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23122 PR rtl-optimization/98777
23123 * lra-int.h (lra_pmode_pseudo): New extern.
23124 * lra.c (lra_pmode_pseudo): New global.
23126 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23128 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23130 * fwprop.c (fwprop_propagation::classify_result): Allow
23131 (subreg (mem)) simplifications.
23133 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23135 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23137 (aarch64_sqdmlal<mode>): ... This...
23138 (aarch64_sqdmlsl<mode>): ... And this.
23139 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23140 (aarch64_sqdmlal_lane<mode>): ... This...
23141 (aarch64_sqdmlsl_lane<mode>): ... And this.
23142 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23143 (aarch64_sqdmlsl_laneq<mode>): ... This...
23144 (aarch64_sqdmlal_laneq<mode>): ... And this.
23145 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23146 (aarch64_sqdmlsl_n<mode>): ... This...
23147 (aarch64_sqdmlal_n<mode>): ... And this.
23148 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23149 (aarch64_sqdmlal2<mode>_internal): ... This...
23150 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23152 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23154 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23156 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23159 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23161 2021-01-21 liuhongt <hongtao.liu@intel.com>
23163 PR rtl-optimization/98694
23164 * regcprop.c (copy_value): If SRC had been assigned a mode
23165 narrower than the copy, we can't link DEST into the chain even
23166 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23169 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23171 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23172 Convert to define_insn_and_split. Split into simple move when moving
23175 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23177 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23178 Adjust comment. Simplify code.
23180 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23183 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23184 with DW_FORM_line_strp form.
23185 (prune_unused_types_update_strings): Don't add into debug_str_hash
23186 indirect strings with DW_FORM_line_strp form.
23187 (adjust_name_comp_dir): New function.
23188 (dwarf2out_finish): Call it on CU DIEs after resetting
23189 debug_line_str_hash.
23191 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23193 PR rtl-optimization/98722
23194 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23195 has no 3-op add insn to transform insns containing two pluses.
23197 2021-01-20 Richard Biener <rguenther@suse.de>
23199 * hwint.h (add_hwi): New function.
23200 (mul_hwi): Likewise.
23201 * tree-data-ref.c (initialize_matrix_A): Properly translate
23202 tree constants and avoid HOST_WIDE_INT_MIN.
23203 (lambda_matrix_row_add): Avoid undefined integer overflow
23204 and return true on such overflow.
23205 (lambda_matrix_right_hermite): Handle overflow from
23206 lambda_matrix_row_add gracefully. Simplify previous fix.
23207 (analyze_subscript_affine_affine): Likewise.
23209 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23211 PR tree-optimization/96674
23212 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23213 x >= y && y != XXX_MIN --> x > y - 1
23215 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23217 PR tree-optimization/98535
23218 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23219 If the high and low permutes are the same, remove the high permutes
23220 from the working set and only continue with the low ones.
23222 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23224 PR tree-optimization/98721
23225 * builtins.c (access_ref::inform_access): Don't assume
23226 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23227 object whenever allocfn is NULL, rather than only when DECL_P
23228 is true. Use %qE instead of %qD for that. Formatting fixes.
23230 2021-01-20 Richard Biener <rguenther@suse.de>
23232 PR tree-optimization/98758
23233 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23234 (lambda_matrix_right_hermite): Avoid undefinedness with
23235 signed integer abs and multiplication.
23236 (analyze_subscript_affine_affine): Use lambda_int.
23238 2021-01-20 David Malcolm <dmalcolm@redhat.com>
23241 * dwarf2out.c (output_line_info): Rename static variable
23242 "generation", moving it out of the function to...
23243 (output_line_info_generation): New.
23244 (init_sections_and_labels): Likewise, renaming the variable to...
23245 (init_sections_and_labels_generation): New.
23246 (dwarf2out_c_finalize): Reset the new variables.
23248 2021-01-19 Martin Sebor <msebor@redhat.com>
23250 PR middle-end/98664
23251 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
23252 all functions, even if they're not declared artificial or inline.
23253 * tree.c (tree_inlined_location): Use macro expansion location
23254 only if scope traversal fails to expose one.
23256 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
23258 PR rtl-optimization/92294
23259 * alias.c (compare_base_symbol_refs): Take an extra parameter
23260 and add the distance between two symbols to it. Enshrine in
23261 comments that -1 means "either 0 or 1, but we can't tell
23262 which at compile time".
23263 (memrefs_conflict_p): Update call accordingly.
23264 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23267 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23269 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23270 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23271 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23272 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23273 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23274 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23275 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23276 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23277 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23278 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23279 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23280 sqmovun): Use NONE flags.
23282 2021-01-19 Richard Biener <rguenther@suse.de>
23285 * ipa-modref.c (analyze_stmt): Only record a summary for a
23288 2021-01-19 Richard Biener <rguenther@suse.de>
23290 PR middle-end/98638
23291 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23293 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23295 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23296 built-in define __FIX_LEON3FT_TN0018.
23298 2021-01-19 Richard Biener <rguenther@suse.de>
23301 * tree-inline.c (tree_function_versioning): Set input_location
23302 to UNKNOWN_LOCATION throughout the function.
23304 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23307 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23309 2021-01-19 Martin Jambor <mjambor@suse.cz>
23312 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23313 whether non-call exceptions allow removal of a statement.
23314 (isra_analyze_call): Pass the appropriate function to
23315 ssa_name_only_returned_p.
23317 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23319 * config/riscv/arch-canonicalize (longext_sort): New function for
23320 sorting 'multi-letter'.
23321 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23322 'alts'. The 'arch' may not be the first of 'alts'.
23323 (_expand_combination): Add underline for the 'ext' without '*'.
23324 This is because, a single-letter extension can always be treated well
23325 with a '_' prefix, but it cannot be separated out if it is appended
23328 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23331 * ira.c (ira): Skip abnormal critical edge splitting.
23333 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23335 PR tree-optimization/98727
23336 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23337 second .MUL_OVERFLOW operand for signed multiplication with overflow
23338 checking if the second operand of multiplication is not constant.
23340 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23342 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23343 defaults to version 4.
23345 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23347 * attribs.h (fndecl_dealloc_argno): New decl.
23348 * builtins.c (call_dealloc_argno): Split out second half of
23350 (fndecl_dealloc_argno): New.
23351 * doc/extend.texi (Common Function Attributes): Document the
23352 interaction between the analyzer and the malloc attribute.
23353 * doc/invoke.texi (Static Analyzer Options): Likewise.
23355 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23357 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23358 dwarf_version to 4.
23359 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23361 2021-01-17 Martin Jambor <mjambor@suse.cz>
23364 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23367 2021-01-17 Mark Wielaard <mark@klomp.org>
23369 * common.opt (gdwarf-): Init(5).
23370 * doc/invoke.texi (-gdwarf): Document default to 5.
23372 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23374 * builtin-types.def
23375 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23377 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23378 ...this. Add extra argument.
23379 * gimplify.c (omp_default_clause): Ensure that event handle is
23380 firstprivate in a task region.
23381 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23382 (gimplify_adjust_omp_clauses): Likewise.
23383 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23384 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23385 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23386 if detach clause specified. Add detach argument when generating
23388 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23390 (finish_taskreg_scan): Move field for variable containing the event
23391 handle to the front of the struct.
23392 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23394 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23395 OMP_CLAUSE_DETACH clause.
23396 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23397 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23398 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23400 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23402 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23404 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23406 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23407 multilib directory names. Use MULTILIB_REQUIRED instead of
23408 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23409 -mcustom-fpu-cfg=fph2 multilib.
23411 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23413 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23414 (nios2_init_fpu_configs): Provide register values for new
23415 -mcustom-fpu-cfg=fph2 option variant.
23416 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23419 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23421 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23422 custom instruction warnings.
23424 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23426 PR tree-optimization/96669
23427 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23429 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23431 PR tree-optimization/96271
23432 * passes.def: Pass false argument to first two pass_cd_dce
23433 instances and true to last instance. Add comment that
23434 last instance rewrites no longer addressed locals.
23435 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23437 (pass_cd_dce::set_pass_param): New method.
23438 (pass_cd_dce::execute): Return TODO_update_address_taken from
23439 last cd_dce instance.
23441 2021-01-15 Carl Love <cel@us.ibm.com>
23443 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23445 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23446 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23447 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23448 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23449 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23450 Add builtin define.
23451 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23452 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23453 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23454 New overloaded definitions.
23455 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23456 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23457 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23458 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23459 P10V_BUILTIN_MULHU_V4SI]: Add case
23460 statement for builtins.
23461 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23462 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23463 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23464 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23465 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23466 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23467 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23468 Add define_insn, mode is VIlong.
23469 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23470 Add builtin descriptions.
23472 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23474 * final.c (final_start_function_1): Reset force_source_line.
23476 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23478 PR tree-optimization/96669
23479 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23480 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23481 possibly different power of two constants and to right shift too.
23483 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23485 PR tree-optimization/96681
23486 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23487 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23488 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23489 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23491 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23493 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23495 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23497 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23498 cmul<conj_op><mode>3): New.
23499 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23500 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23501 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23502 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23503 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23504 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23505 (rot_op): Renamed to conj_op.
23506 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23507 cmul<conj_op><mode>3): New.
23508 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23509 cmul<conj_op><mode>3): New.
23511 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23515 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23516 Escape the tempfile name when constructing the expected output.
23518 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23520 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23522 (aarch64_<su>mlsl_hi<mode>): ... This.
23523 (aarch64_<su>mlsl_hi<mode>): Define.
23524 (*aarch64_<su>mlsl<mode): Rename to...
23525 (aarch64_<su>mlsl<mode): ... This.
23526 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23527 smlsl_hi, umlsl_hi): Define builtins.
23528 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23529 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23530 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23531 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23533 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23535 * config/i386/i386-c.c (ix86_target_macros):
23536 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23538 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23541 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23542 * Makefile.in (RTL_SSA_H): New variable.
23543 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23544 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23545 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23547 * config/aarch64/aarch64-cc-fusion.cc: New file.
23549 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23551 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23552 calling cancel_changes for changes that no longer exist.
23554 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23556 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23557 (function_info::reg_defs): ...this.
23558 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23559 (function_info::reg_defs): ...this.
23561 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23564 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23566 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23569 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23572 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23574 2021-01-15 Richard Biener <rguenther@suse.de>
23576 PR tree-optimization/96376
23577 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23578 for VMAT_INVARIANT.
23580 2021-01-15 Martin Liska <mliska@suse.cz>
23582 * doc/install.texi: Document that some tests need pytest module.
23583 * doc/sourcebuild.texi: Likewise.
23585 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23588 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23590 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23592 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23593 (mve_vshrq_n_u<mode>_imm): Likewise.
23594 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23595 * config/arm/vec-common.md: ... here.
23597 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23599 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
23601 * config/arm/neon.md (vashl<mode>3): Delete.
23602 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
23603 (vasl<mode>3): New expander.
23605 2021-01-15 Richard Biener <rguenther@suse.de>
23607 PR tree-optimization/98685
23608 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
23609 of vector extern defs.
23611 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23614 * diagnostic.c (diagnostic_kind_text): Break out this array
23616 (diagnostic_build_prefix): ...here.
23617 (fancy_abort): Detect when diagnostic_initialize has not yet been
23618 called and fall back to a minimal implementation of printing the
23619 ICE, rather than segfaulting in internal_error.
23621 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23623 * diagnostic.c (diagnostic_initialize): Eliminate
23624 parseable_fixits_p in favor of initializing extra_output_kind from
23625 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23626 (convert_column_unit): New function, split out from...
23627 (diagnostic_converted_column): ...this.
23628 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
23629 Use them to call convert_column_unit on the column values.
23630 (diagnostic_report_diagnostic): Eliminate conditional on
23631 parseable_fixits_p in favor of a switch statement on
23632 extra_output_kind, passing the appropriate values to the new
23633 params of print_parseable_fixits.
23634 (selftest::test_print_parseable_fixits_none): Update for new
23635 params of print_parseable_fixits.
23636 (selftest::test_print_parseable_fixits_insert): Likewise.
23637 (selftest::test_print_parseable_fixits_remove): Likewise.
23638 (selftest::test_print_parseable_fixits_replace): Likewise.
23639 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23641 (selftest::diagnostic_c_tests): Call it.
23642 * diagnostic.h (enum diagnostics_extra_output_kind): New.
23643 (diagnostic_context::parseable_fixits_p): Delete field in favor
23645 (diagnostic_context::extra_output_kind): ...this new field.
23646 * doc/invoke.texi (Environment Variables): Add
23647 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23648 * opts.c (common_handle_option): Update handling of
23649 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
23652 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23654 * tree-vect-slp-patterns.c (class complex_operations_pattern,
23655 complex_operations_pattern::matches,
23656 complex_operations_pattern::recognize,
23657 complex_operations_pattern::build): New.
23658 (slp_patterns): Use it.
23660 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23662 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
23663 * optabs.def (cmls_optab, cmls_conj_optab): New.
23664 * doc/md.texi: Document them.
23665 * tree-vect-slp-patterns.c (class complex_fms_pattern,
23666 complex_fms_pattern::matches, complex_fms_pattern::recognize,
23667 complex_fms_pattern::build): New.
23669 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23671 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
23672 * optabs.def (cmla_optab, cmla_conj_optab): New.
23673 * doc/md.texi: Document them.
23674 * tree-vect-slp-patterns.c (vect_match_call_p,
23675 class complex_fma_pattern, vect_slp_reset_pattern,
23676 complex_fma_pattern::matches, complex_fma_pattern::recognize,
23677 complex_fma_pattern::build): New.
23679 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23681 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
23682 * optabs.def (cmul_optab, cmul_conj_optab): New.
23683 * doc/md.texi: Document them.
23684 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
23685 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
23686 vect_build_combine_node, class complex_mul_pattern,
23687 complex_mul_pattern::matches, complex_mul_pattern::recognize,
23688 complex_mul_pattern::build): New.
23690 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23692 * tree-vect-slp.c (optimize_load_redistribution_1): New.
23693 (optimize_load_redistribution, vect_is_slp_load_node): New.
23694 (vect_match_slp_patterns): Use it.
23696 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23698 * tree-vect-slp-patterns.c (complex_add_pattern::build):
23701 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
23703 * config/gcn/mkoffload.c (main): Create an offload image only in
23704 64-bit configurations.
23706 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
23709 * config/i386/i386-options.c (ix86_option_override_internal):
23710 Issue an error for -fcf-protection with CF_BRANCH when compiling
23711 for 32-bit non-TARGET_CMOV targets.
23713 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23716 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
23717 Remove declaration and initialization of shadow variable "ret".
23718 (ix86_option_override_internal): Remove delcaration of
23719 shadow variable "i". Redeclare shadowed variable to unsigned.
23720 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
23721 * config/i386/i386-builtins.c (get_builtin_code_for_version):
23722 Update for redeclaration.
23723 * config/i386/i386.h (pta_size): Ditto.
23725 2021-01-14 Richard Biener <rguenther@suse.de>
23727 PR tree-optimization/98674
23728 * tree-data-ref.c (base_supports_access_fn_components_p): New.
23729 (initialize_data_dependence_relation): For two bases without
23730 possible access fns resort to type size equality when determining
23731 shape compatibility.
23733 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23736 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
23737 <=, >= operators in vcle and vcge intrinsics respectively.
23738 * config/arm/arm_neon_builtins.def: Remove entry for
23741 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23744 * config/i386/i386-options.c (ix86_function_specific_save):
23745 Remove redundant assignment to opts->x_ix86_branch_cost.
23746 * config/i386/i386.c (ix86_prefetch_sse):
23747 Rename from x86_prefetch_sse. Update all uses.
23748 * config/i386/i386.h: Update for rename.
23749 * config/i386/i386-options.h: Ditto.
23751 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23754 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
23755 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
23756 Use Bm instead of m for non-avx. Add isa attribute.
23758 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23760 PR tree-optimization/96688
23761 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
23762 ~X can be simplified.
23764 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
23766 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
23767 IFN_LOAD_LANES results.
23769 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23771 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
23773 (aarch64_xtn<mode>): Likewise.
23774 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
23777 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
23779 (vmovl_s16): Likewise.
23780 (vmovl_s32): Likewise.
23781 (vmovl_u8): Likewise.
23782 (vmovl_u16): Likewise.
23783 (vmovl_u32): Likewise.
23784 (vmovn_s16): Likewise.
23785 (vmovn_s32): Likewise.
23786 (vmovn_s64): Likewise.
23787 (vmovn_u16): Likewise.
23788 (vmovn_u32): Likewise.
23789 (vmovn_u64): Likewise.
23791 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23793 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
23795 (aarch64_<su>qxtn2<mode>_be): Likewise.
23796 (aarch64_<su>qxtn2<mode>): Likewise.
23797 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
23799 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
23800 (su): Handle ss_truncate and us_truncate.
23801 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
23803 (vqmovn_high_s32): Likewise.
23804 (vqmovn_high_s64): Likewise.
23805 (vqmovn_high_u16): Likewise.
23806 (vqmovn_high_u32): Likewise.
23807 (vqmovn_high_u64): Likewise.
23809 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23811 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
23813 (aarch64_xtn2<mode>_be): Likewise.
23814 (aarch64_xtn2<mode>): Likewise.
23815 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
23817 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
23819 (vmovn_high_s32): Likewise.
23820 (vmovn_high_s64): Likewise.
23821 (vmovn_high_u16): Likewise.
23822 (vmovn_high_u32): Likewise.
23823 (vmovn_high_u64): Likewise.
23825 2021-01-13 Stafford Horne <shorne@gmail.com>
23827 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
23829 2021-01-13 Stafford Horne <shorne@gmail.com>
23831 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
23833 2021-01-13 Stafford Horne <shorne@gmail.com>
23835 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
23836 define for __or1k_hard_float__.
23838 2021-01-13 Stafford Horne <shorne@gmail.com>
23840 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
23841 (PROFILE_HOOK): Define to call _mcount.
23842 (FUNCTION_PROFILER): Change from abort to no-op.
23844 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23846 PR tree-optimization/96691
23847 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
23848 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
23849 (~D ^ C) or (D ^ C) can be simplified.
23851 2021-01-13 Richard Biener <rguenther@suse.de>
23853 PR tree-optimization/92645
23854 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
23855 until after vector lowering.
23857 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23859 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
23861 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
23862 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
23864 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23866 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
23868 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
23869 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
23871 2021-01-13 Richard Biener <rguenther@suse.de>
23873 PR tree-optimization/92645
23874 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
23875 BIT_FIELD_REF argument.
23876 (vect_build_slp_tree_2): Record the desired vector type
23877 on the external vector def.
23878 (vectorizable_slp_permutation): Handle required punning
23879 of existing vector defs.
23881 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23883 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
23885 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23887 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
23889 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
23891 * config.gcc [$target == *-*-gnu*]: Enable
23892 'default_gnu_indirect_function'.
23894 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23897 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
23898 registers before calling targetm.vectorize.vec_perm_const, only after
23900 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
23901 two argument permutation when one operand is zero vector and only
23902 after that force operands into registers.
23903 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
23904 define_insn_and_split pattern.
23905 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
23906 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
23907 (*avx2_zero_extendv8hiv8si2_1): Likewise.
23908 (*avx512f_zero_extendv8siv8di2_1): Likewise.
23909 (*avx2_zero_extendv4siv4di2_1): Likewise.
23910 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
23912 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
23913 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
23914 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
23915 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
23916 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
23917 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
23919 2021-01-13 Martin Liska <mliska@suse.cz>
23921 PR tree-optimization/98455
23922 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
23923 Record also virtual PHIs.
23924 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
23927 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
23929 * doc/invoke.texi (C++ Modules): Fix typos.
23931 2021-01-13 Richard Biener <rguenther@suse.de>
23933 PR tree-optimization/98640
23934 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
23935 handle plus or minus from a truncated operand to be
23938 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23941 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
23942 define_insn_and_split patterns.
23943 (splitter after *btr<mode>_2): New splitter.
23945 2021-01-13 Martin Liska <mliska@suse.cz>
23948 * cgraphunit.c (analyze_functions): Remove dead code.
23950 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
23952 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
23953 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
23954 (a64fx_regmove_cost, a64fx_vector_cost): New.
23955 (a64fx_tunings): Use the new added cost tables.
23957 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23960 * config/i386/predicates.md (pmovzx_parallel): New predicate.
23961 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
23962 define_insn_and_split pattern.
23963 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
23964 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
23966 2021-01-13 Julian Brown <julian@codesourcery.com>
23968 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
23969 to fix v0 register.
23971 2021-01-13 Julian Brown <julian@codesourcery.com>
23973 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
23976 2021-01-13 Julian Brown <julian@codesourcery.com>
23978 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
23979 for reciprocal-approximation instructions.
23980 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
23981 refinement and division result.
23982 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
23984 2021-01-13 Julian Brown <julian@codesourcery.com>
23986 * config/gcn/gcn-valu.md (subdf): Rename to...
23989 2021-01-12 Martin Liska <mliska@suse.cz>
23991 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
23993 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
23995 * function-abi.h: Fix typo.
23997 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24001 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
24002 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
24003 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
24004 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
24005 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
24006 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
24007 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
24008 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
24009 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
24010 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
24011 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
24012 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
24013 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
24014 (ARM_HAVE_V2DI_LDST): Likewise.
24015 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
24016 (*movmisalign<mode>_mve_load): New pattern.
24017 * config/arm/neon.md (movmisalign<mode>): Move to ...
24018 * config/arm/vec-common.md: ... here.
24020 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
24023 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
24024 of pattern 'plus (plus (hard reg, const), pseudo)'.
24026 2021-01-12 Richard Biener <rguenther@suse.de>
24028 PR tree-optimization/98550
24029 * tree-vect-slp.c (vect_record_max_nunits): Check whether
24030 the group size is a multiple of the vector element count.
24031 (vect_build_slp_tree_1): When we need to fail because
24032 the vector type choosen causes unrolling do so lazily
24033 without affecting matches only at the end to guide group splitting.
24035 2021-01-12 Martin Liska <mliska@suse.cz>
24038 * optc-save-gen.awk: Compare also n_target_save vars with
24041 2021-01-12 Martin Liska <mliska@suse.cz>
24043 * gcov.c (source_info::debug): New.
24044 (print_usage): Add --debug (-D) option.
24045 (process_args): Likewise.
24046 (generate_results): Call src->debug after
24047 accumulate_line_counts.
24048 (read_graph_file): Properly assign id for EXIT_BLOCK.
24049 * profile.c (branch_prob): Dump function body before it is
24052 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24054 PR tree-optimization/98629
24055 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
24056 unless returning non-zero.
24058 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24060 PR tree-optimization/95731
24061 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
24062 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
24063 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
24064 only after optimize_range_tests_var_bound.
24066 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24068 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
24069 * configure: Regenerated.
24071 2021-01-12 liuhongt <hongtao.liu@intel.com>
24074 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
24076 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
24079 2021-01-12 Alexandre Oliva <oliva@adacore.com>
24081 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
24083 (auto_end_imm_use_stmt_traverse): New struct.
24084 (FOR_EACH_IMM_USE_STMT): Use it.
24085 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
24087 * gimple-ssa-strength-reduction.c: ... here, ...
24088 * graphite-scop-detection.c: ... here, ...
24089 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
24090 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
24091 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
24092 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
24093 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
24094 * tree-vect-slp.c: ... and here, ...
24095 * doc/tree-ssa.texi: ... and the example here.
24097 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24099 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
24100 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
24101 (*sdiv_pow2<mode>3): New pattern.
24102 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
24103 Wrap the ASRD in an UNSPEC_PRED_X.
24104 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
24105 predicate with a constant PTRUE, if it isn't already.
24106 (*cond_<sve_int_op><mode>_z): Replace with...
24107 (*cond_<sve_int_op><mode>_any): ...this new pattern.
24109 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24111 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
24112 SVE_FULL_I to SVE_I.
24113 (*cond_bic<mode>_any): Likewise.
24115 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24117 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24118 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24121 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24123 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24124 SVE_FULL_I to SVE_I.
24125 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24126 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24127 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24128 for the max and min but not for the minus.
24129 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24131 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24133 * config/aarch64/iterators.md (SVE_24I): New iterator.
24134 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24135 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24137 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24139 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24140 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24142 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24143 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24144 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24145 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24147 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24149 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24150 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24151 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24154 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24156 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24157 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24158 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24160 2021-01-11 Martin Liska <mliska@suse.cz>
24163 * symtab-clones.h (clone_info::release): Release
24164 symtab::m_clones with ggc_delete as it's a GGC memory.
24166 2021-01-11 Matthias Klose <doko@ubuntu.com>
24168 * Makefile.in (LINK_PROGRESS): Show the link target.
24170 2021-01-11 Richard Biener <rguenther@suse.de>
24172 PR tree-optimization/91403
24173 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24174 single-element interleaving group size at 4096 elements.
24176 2021-01-11 Richard Biener <rguenther@suse.de>
24178 PR tree-optimization/98526
24179 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24180 of the actual reduction op for the regular case.
24181 (vectorizable_reduction): Cost the stmts
24182 vect_transform_reduction produces here.
24184 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24186 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24187 big-endian, use UNPACK[_FLOAT]_HI.
24189 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24191 * tree-vect-slp-patterns.c (class complex_pattern,
24192 class complex_add_pattern): Add parameters to matches.
24193 (complex_add_pattern::build): Free memory.
24194 (complex_add_pattern::matches): Move validation end of match.
24195 (complex_add_pattern::recognize): Likewise.
24197 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24199 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24201 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24203 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24205 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24207 PR tree-optimization/95867
24208 * tree-ssa-math-opts.h: New header.
24209 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24210 (powi_as_mults): No longer static. Use build_one_cst instead of
24211 build_real. Formatting fix.
24212 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24213 (attempt_builtin_powi): Handle multiplication reassociation without
24214 powi_fndecl using powi_as_mults.
24215 (reassociate_bb): For integral types don't require
24216 -funsafe-math-optimizations to call attempt_builtin_powi.
24218 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24220 PR tree-optimization/95852
24221 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24222 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24223 allow in the bb any of the stmts in that vector, div_stmt and
24224 up to 3 cast stmts.
24225 (arith_cast_equal_p): New function.
24226 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24227 multiply overflow checks.
24228 (match_arith_overflow): Adjust caller. Handle signed multiply
24231 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24233 PR tree-optimization/95852
24234 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24235 (uaddsub_overflow_check_p): Renamed to ...
24236 (arith_overflow_check_p): ... this. Handle also multiplication
24237 with overflow check.
24238 (match_uaddsub_overflow): Renamed to ...
24239 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
24240 also multiplication with overflow check. Adjust function comment.
24241 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
24242 match_arith_overflow also for MULT_EXPR.
24244 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24246 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24247 __builtin_convertvector.
24248 (vmovl_s16): Likewise.
24249 (vmovl_s32): Likewise.
24250 (vmovl_u8): Likewise.
24251 (vmovl_u16): Likewise.
24252 (vmovl_u32): Likewise.
24253 (vmovn_s16): Likewise.
24254 (vmovn_s32): Likewise.
24255 (vmovn_s64): Likewise.
24256 (vmovn_u16): Likewise.
24257 (vmovn_u32): Likewise.
24258 (vmovn_u64): Likewise.
24260 2021-01-11 Martin Liska <mliska@suse.cz>
24262 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
24263 (if_chain::is_beneficial): Delete clusters
24264 (find_conditions): Make second argument of conditions_in_bbs a
24265 pointer so that we control over it's lifetime.
24266 (pass_if_to_switch::execute): Delete them.
24268 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24270 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24273 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24275 * config/vax/vax.md (cc): Remove mode attribute.
24276 (subst_<cc>, subst_f<cc>): Rename to...
24277 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24278 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24279 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24280 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24282 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24284 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24285 `const_double_zero'.
24287 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24289 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24290 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24293 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24295 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24297 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24298 with `const_double_zero'.
24299 * doc/rtl.texi (Constant Expression Types): Document it.
24301 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24304 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24305 POINTER_DIFF_EXPR to be any integral type.
24307 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24309 PR rtl-optimization/98603
24310 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24311 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24312 if any, set ASM_OPERANDS mode to VOIDmode and change
24313 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24315 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24318 * final.c (notice_source_line): Narrow down the condition to
24319 skip a line-0 marker.
24321 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24323 * ipa-modref.c (merge_call_side_effects): Fix
24324 linebreak split by reordering two print calls.
24326 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24328 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24329 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24330 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24331 consistency, fix constraint.
24333 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24335 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24336 callables instead of mask values.
24337 (struct target_flag_set_p): New predicate.
24338 (s390_cpu_cpp_builtins_internal): Define or undefine
24339 __LONG_DOUBLE_VX__ macro.
24341 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24344 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24345 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24348 2021-01-08 Richard Biener <rguenther@suse.de>
24350 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24352 2021-01-08 Richard Biener <rguenther@suse.de>
24354 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24355 (vect_build_slp_tree): On cache hit release the matched
24356 scalar stmts vector.
24357 * tree-vect-stmts.c (vectorizable_store): Properly free
24358 vec_oprnds before possibly gathering them again.
24360 2021-01-08 Richard Biener <rguenther@suse.de>
24362 PR tree-optimization/98544
24363 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24364 permutes at a permute node.
24366 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24369 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24370 mcount in large model. Sorry for large model with PIC.
24372 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24375 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24376 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24377 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24378 TargetSave and initialize for variables with enum types.
24379 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24380 mstack-protector-guard-symbol=): Add Save.
24381 * config/i386/i386-options.c (ix86_function_specific_save,
24382 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24383 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24384 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24385 x_ix86_veclibabi_type.
24387 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24389 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24390 SVE_FULL_I to SVE_I.
24391 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24393 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24395 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24396 SVE_FULL_I to SVE_I.
24397 (*cond_uxt<mode>_any): Likewise.
24399 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24401 * config/aarch64/iterators.md (Vwhalf): New iterator.
24402 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24404 (aarch64_<sur>adalp<mode>): ... This. Make more
24406 (<sur>sadv16qi): Adjust callsite of the above.
24407 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24409 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24411 (vpadal_s16): Likewise.
24412 (vpadal_u8): Likewise.
24413 (vpadal_u16): Likewise.
24414 (vpadalq_s8): Likewise.
24415 (vpadalq_s16): Likewise.
24416 (vpadalq_s32): Likewise.
24417 (vpadalq_u8): Likewise.
24418 (vpadalq_u16): Likewise.
24419 (vpadalq_u32): Likewise.
24421 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24423 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24425 (aarch64_<su>abd<mode>): ... This.
24426 (<sur>sadv16qi): Adjust callsite of the above.
24427 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24429 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24431 (vabd_s16): Likewise.
24432 (vabd_s32): Likewise.
24433 (vabd_u8): Likewise.
24434 (vabd_u16): Likewise.
24435 (vabd_u32): Likewise.
24436 (vabdq_s8): Likewise.
24437 (vabdq_s16): Likewise.
24438 (vabdq_s32): Likewise.
24439 (vabdq_u8): Likewise.
24440 (vabdq_u16): Likewise.
24441 (vabdq_u32): Likewise.
24443 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24445 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24447 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24448 (vaba_s16): Likewise.
24449 (vaba_s32): Likewise.
24450 (vaba_u8): Likewise.
24451 (vaba_u16): Likewise.
24452 (vaba_u32): Likewise.
24453 (vabaq_s8): Likewise.
24454 (vabaq_s16): Likewise.
24455 (vabaq_s32): Likewise.
24456 (vabaq_u8): Likewise.
24457 (vabaq_u16): Likewise.
24458 (vabaq_u32): Likewise.
24460 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24462 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24463 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24464 Change RTL pattern to match.
24466 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24468 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24469 * config/riscv/riscv-c.c (riscv-subset.h): New.
24470 (INCLUDE_STRING): Define.
24471 (riscv_cpu_cpp_builtins): Add new style architecture extension
24473 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24474 (riscv_subset_list::end): New.
24475 (riscv_current_subset_list): New.
24477 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24479 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24480 Move to riscv-subset.h.
24481 (struct riscv_subset_t): Ditto.
24482 (class riscv_subset_list): Ditto.
24483 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24484 from riscv-common.c.
24485 (struct riscv_subset_t): Ditto.
24486 (class riscv_subset_list): Ditto.
24487 * config/riscv/t-riscv ($(common_out_file)): Add file
24490 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24493 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24494 New define_insn patterns.
24496 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24498 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24499 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24500 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24502 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24504 PR tree-optimization/98560
24505 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24506 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24508 (vec_cond_direct): Likewise argument 2.
24509 (vec_condu_direct, vec_condeq_direct): Delete.
24510 (expand_vect_cond_optab_fn): Rename to...
24511 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24512 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24513 (expand_vect_cond_mask_optab_fn): Rename to...
24514 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24515 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24517 (direct_vec_cond_optab_supported_p): Likewise.
24518 (direct_vec_condu_optab_supported_p): Delete.
24519 (direct_vec_condeq_optab_supported_p): Delete.
24520 * gimple-isel.cc: Include internal-fn.h.
24521 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24524 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24526 PR tree-optimization/98560
24527 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24528 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24530 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24532 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24533 rotate_insn and optab code attributes.
24534 Update all uses to merged code attribute.
24535 * config/i386/sse.md: Update all uses to merged code attribute.
24536 * config/i386/mmx.md: Update all uses to merged code attribute.
24538 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24540 PR tree-optimization/98568
24541 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24542 (bswap_replace): Use it.
24544 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24546 PR rtl-optimization/97978
24547 * lra-int.h (lra_hard_reg_split_p): New external.
24548 * lra.c (lra_hard_reg_split_p): New global.
24549 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24550 * lra-assigns.c (lra_assign): Don't check allocation correctness
24551 after hard reg splitting.
24553 2021-01-06 Martin Sebor <msebor@redhat.com>
24556 * builtins.c (new_delete_mismatch_p): New overload.
24557 (new_delete_mismatch_p (tree, tree)): Call it.
24559 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24561 * Makefile.in (T_GLIMITS_H): New.
24562 (stmp-int-hdrs): Depend on it, use it.
24563 * config/t-vxworks (T_GLIMITS_H): Override it.
24564 (vxw-glimits.h): New.
24566 2021-01-06 Richard Biener <rguenther@suse.de>
24568 PR tree-optimization/98513
24569 * value-range.cc (intersect_ranges): Compare the upper bounds
24570 for the expected relation.
24572 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24575 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24577 * doc/standards.texi (HSAIL): Remove section.
24579 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24581 * configure: Re-generate.
24583 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24585 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24586 ISO 14882:2020 standard.
24587 * doc/standards.texi: Likewise.
24589 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24591 PR tree-optimization/94802
24592 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24593 * expr.c: Include tree-pretty-print.h and flags.h.
24594 (maybe_optimize_sub_cmp_0): New function.
24595 (do_store_flag): Use it.
24596 * cfgexpand.c (expand_gimple_cond): Likewise.
24598 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24600 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
24601 * rtlanal.c (simple_regno_set): Tweak description to clarify the
24604 2021-01-05 Richard Biener <rguenther@suse.de>
24606 PR tree-optimization/98516
24607 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
24608 lanes when materializing on a VEC_PERM node.
24609 (vectorizable_slp_permutation): Dump the permute properly.
24611 2021-01-05 Richard Biener <rguenther@suse.de>
24613 * tree-vect-slp.c (vect_slp_region): Move debug counter
24614 to cover individual subgraphs.
24616 2021-01-05 Richard Biener <rguenther@suse.de>
24618 PR tree-optimization/98428
24619 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
24620 vector lane extracts for loop vectorization.
24622 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24624 PR tree-optimization/98514
24625 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
24627 (operand_rank): Change type from hash_map<tree, long> to
24628 hash_map<tree, int64_t>.
24629 (phi_rank): Change return type from long to int64_t.
24630 (loop_carried_phi): Change block_rank variable type from long to
24632 (propagate_rank): Change return type, rank parameter type and
24633 op_rank variable type from long to int64_t.
24634 (find_operand_rank): Change return type from long to int64_t
24635 and change slot variable type from long * to int64_t *.
24636 (insert_operand_rank): Change rank parameter type from long to
24638 (get_rank): Change return type and rank variable type from long to
24639 int64_t. Use PRId64 instead of ld to print the rank.
24640 (init_reassoc): Change rank variable type from long to int64_t
24641 and adjust correspondingly bb_rank and operand_rank initialization.
24643 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24645 PR tree-optimization/96928
24646 * tree-ssa-phiopt.c (xor_replacement): New function.
24647 (tree_ssa_phiopt_worker): Call it.
24649 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24651 PR tree-optimization/96930
24652 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
24653 from narrower value which has the same type as 1 << B, perform
24654 the right shift on the narrower value followed by extension.
24656 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24658 PR tree-optimization/96239
24659 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
24661 (get_status_for_store_merging): Don't return BB_INVALID for blocks
24662 with potential bswap optimizable CONSTRUCTORs.
24663 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
24666 2021-01-05 Richard Biener <rguenther@suse.de>
24668 PR tree-optimization/98381
24669 * tree.c (vector_element_bits): Properly compute bool vector
24671 * tree-vect-loop.c (vectorizable_live_operation): Properly
24672 compute the last lane bit offset.
24674 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24677 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
24678 Clear the top 64 bytes of the input XMM register.
24679 (sse_cvttps2pi): Ditto.
24681 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24684 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
24686 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
24689 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
24692 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
24694 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
24695 (umaddsidi4_split): Likewise.
24697 2021-01-05 liuhongt <hongtao.liu@intel.com>
24700 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
24701 define_insn_and_split for zero_extend of subreg HI of pmovskb
24703 (*sse2_pmovskb_zexthisi): Add new combine splitters for
24704 zero_extend of not of subreg HI of pmovskb result.
24706 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24709 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
24711 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
24712 convert_memory_address to convert symbolic immediates to ptr_mode
24713 before forcing them to memory.
24715 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24717 PR rtl-optimization/97144
24718 * recog.c (constrain_operands): Initialize matching_operand
24719 for each alternative, rather than only doing it once.
24721 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24723 PR rtl-optimization/98403
24724 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
24725 why we don't remove call clobbers.
24726 (function_info::apply_changes_to_insn): Don't attempt to add
24727 call clobbers here.
24729 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24731 PR tree-optimization/98371
24732 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
24733 (vect_analyze_loop): If an epilogue loop appears to be cheaper
24734 than the main loop, re-analyze it as a main loop before adopting
24737 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24740 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
24741 * aclocal.m4, configure: Regenerate.
24742 * Makefile.in (NETLIBS): Define.
24743 (BACKEND): Remove $(CODYLIB).
24745 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24747 PR rtl-optimization/98334
24748 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
24749 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
24751 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
24753 * tree-inline.c (expand_call_inline): Restore input_location.
24754 Return result from recursive call.
24756 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24758 PR tree-optimization/95401
24759 * config/aarch64/aarch64-sve-builtins.cc
24760 (gimple_folder::load_store_cookie): Use bits rather than bytes
24761 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
24762 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
24763 * tree-vect-stmts.c (vectorizable_store): Likewise.
24764 (vectorizable_load): Likewise.
24766 2021-01-04 Richard Biener <rguenther@suse.de>
24768 PR tree-optimization/98308
24769 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
24772 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24774 PR tree-optimization/95771
24775 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
24776 with precision smaller than int's precision and types with precision
24777 twice as large as long long. Formatting fixes.
24779 2021-01-04 Richard Biener <rguenther@suse.de>
24781 PR tree-optimization/98464
24782 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
24783 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
24784 (process_bb): Adjust.
24786 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
24789 * doc/invoke.texi (-fsanitize=address): Fix wording describing
24790 clash with -fsanitize=hwaddress.
24792 2021-01-04 Richard Biener <rguenther@suse.de>
24794 PR tree-optimization/98282
24795 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
24796 invariants as VN_NARY.
24798 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24801 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
24802 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
24803 to handle zero operands.
24805 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24807 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
24808 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
24809 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
24812 2021-01-04 Richard Biener <rguenther@suse.de>
24814 PR tree-optimization/98393
24815 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
24816 when hitting the limit.
24818 2021-01-04 Richard Biener <rguenther@suse.de>
24820 PR tree-optimization/98291
24821 * tree-vect-loop.c (vectorizable_reduction): Bypass
24822 associativity check for SLP reductions with VF 1.
24824 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24826 PR tree-optimization/96782
24827 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
24829 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
24831 * collect-utils.c (collect_execute): Check dumppfx.
24832 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
24833 to collect_execute.
24834 (do_link): Add new parameter atsuffix.
24835 (main): Handle -dumpdir option. Skip one argument for
24836 -o, -isystem and -B options.
24837 * gcc.c (make_at_file): New helper function.
24838 (close_at_file): Use it.
24840 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24842 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
24843 Amend handling for LD64_VERSION fallback defaults.
24845 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24847 * config.gcc: Compute default version information
24848 from the configured target. Likewise defaults for
24850 * config/darwin10.h: Removed.
24851 * config/darwin12.h: Removed.
24852 * config/darwin9.h: Removed.
24853 * config/rs6000/darwin8.h: Removed.
24855 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24857 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
24859 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24861 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
24862 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
24864 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24866 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
24868 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
24870 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24872 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
24873 for the Darwin10 unwinder stub from here ...
24874 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
24876 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24878 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
24879 (ASM_DEBUG_SPEC):Only define if the assembler supports
24881 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
24882 (DARWIN_PREFER_DWARF): Define.
24883 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
24884 (DARWIN_PREFER_DWARF): Likewise
24885 (DSYMUTIL_SPEC): Likewise.
24886 (COLLECT_RUN_DSYMUTIL): Likewise.
24887 (ASM_DEBUG_SPEC): Likewise.
24888 (ASM_DEBUG_OPTION_SPEC): Likewise.
24890 2021-01-02 Jan Hubicka <jh@suse.cz>
24892 * cfg.c (free_block): ggc_free bb.
24894 2021-01-01 Jakub Jelinek <jakub@redhat.com>
24896 * gcc.c (process_command): Update copyright notice dates.
24897 * gcov-dump.c (print_version): Ditto.
24898 * gcov.c (print_version): Ditto.
24899 * gcov-tool.c (print_version): Ditto.
24900 * gengtype.c (create_file): Ditto.
24901 * doc/cpp.texi: Bump @copying's copyright year.
24902 * doc/cppinternals.texi: Ditto.
24903 * doc/gcc.texi: Ditto.
24904 * doc/gccint.texi: Ditto.
24905 * doc/gcov.texi: Ditto.
24906 * doc/install.texi: Ditto.
24907 * doc/invoke.texi: Ditto.
24909 2021-01-01 Jakub Jelinek <jakub@redhat.com>
24911 * ChangeLog-2020: Rotate ChangeLog. New file.
24914 Copyright (C) 2021 Free Software Foundation, Inc.
24916 Copying and distribution of this file, with or without modification,
24917 are permitted in any medium without royalty provided the copyright
24918 notice and this notice are preserved.